CampGain Blog

Ecommerce Email Marketing: The Complete 2026 Guide

Arpit Anand — Fri, 03 Jul 2026 00:00:00 GMT

Most stores do not have an email problem. They have a sequencing problem: they send a newsletter when someone remembers to, they never built the automations that run while they sleep, and they cannot tell which of the two made money last month.

This guide walks through building an ecommerce email programme in the order that actually works — sending infrastructure first, then consent, then the handful of automations that carry most of the revenue, then personalization and measurement. Every section is meant to be actionable on a store you are running today, not a theoretical one with a data team behind it.

Campaigns and flows are two different products

Almost every mistake in ecommerce email traces back to blurring these two things together.

A campaign is a one-time send. You pick a segment, write a message, and schedule it — Black Friday, a product drop, a restock announcement. Campaigns create spikes.
A flow is always-on. It has no send date and no audience list. It reacts to an event, like a cart abandonment or a first order, and enrolls whoever triggers it. Flows create a baseline.

They also fail for different reasons. A campaign fails because the offer or the segment was wrong, and you find out within a day. A flow fails quietly — it keeps firing at people who already bought, or it never enrolls anyone because the events reaching it carry no email address. Diagnosing one with the other's playbook wastes weeks.

Build both, but build the flows first. They compound.

Set up sending before you write a word

Decide where your mail physically comes from

Your marketing platform handles campaign logic; a sending provider handles the actual SMTP or API delivery. Sometimes they are the same company, often not. What matters is what you are optimizing for:

Dedicated sending providers such as SendGrid or Resend — built for volume, with bounce and complaint handling, suppression lists, and delivery webhooks. This is what a real store should send marketing from.
Gmail or Google Workspace SMTP — fine for your first few hundred sends while you validate an offer. It has hard daily limits and no bulk reputation tooling, so plan to move off it.

The switch is far easier before you have a list. Do it early.

Authenticate your domain

Four DNS records decide whether mailbox providers trust you at all:

SPF — declares which servers may send on behalf of your domain.
DKIM — cryptographically signs each message so tampering is detectable.
DMARC — tells mailbox providers what to do when SPF or DKIM fails, and where to send reports. Google, Yahoo, and Microsoft all require a DMARC record from bulk senders.
BIMI — optional, and only works once DMARC is enforcing. It puts your logo beside your messages in supporting clients.

Get these wrong and nothing else in this guide matters. The email deliverability guide covers the full setup, including the bulk-sender rules Gmail and Yahoo enforce and how to warm a new sending domain without burning it.

Collect consent you can prove

Every subscriber should be traceable to a moment where they agreed to hear from you, with a timestamp and a source. That is a legal requirement in most markets and a deliverability requirement everywhere — bought and scraped lists generate exactly the complaints and spam-trap hits that end sending programmes.

The practical version: an opt-in at checkout, a signup form that states what you will send and roughly how often, and an unsubscribe link that works on the first click. Never pre-tick the box.

The five flows that carry the revenue

These cover the large majority of what a store needs. Build them in this order.

1. Welcome series

Triggered when someone subscribes. This is the moment of maximum intent — they just raised their hand. Three to five emails across the first week: who you are, what you sell and why it is different, evidence that other people buy it, and a reason to place a first order.

2. Abandoned cart and abandoned checkout

Triggered when a shopper adds to cart or starts checkout and does not finish. Usually the single highest-revenue automation in an ecommerce account. Two or three messages: a low-key reminder within a few hours, a follow-up that answers the real objection (shipping cost, sizing, returns policy), and optionally a final nudge.

The non-negotiable part is an exit condition on purchase. A flow that keeps chasing someone who already checked out is worse than no flow at all — it reads as incompetence, and it costs you the complaint.

3. Browse abandonment

Triggered when an identified shopper views a product repeatedly without adding it to cart. Lower intent than a cart, so keep it to one message and make it useful rather than pushy.

One constraint worth planning around: browse flows can only reach shoppers you can identify. An anonymous visitor with no email or phone on file never enrolls, however many pages they view. If your browse flow reaches almost nobody, that is usually the reason.

4. Post-purchase

Triggered on order placement. Order confirmation, shipping updates, a how-to-use-it email timed to arrival, then a review request. This is the cheapest loyalty you will ever buy, and it deflects a meaningful share of "where is my order" support tickets.

5. Win-back

Triggered when a customer crosses your definition of lapsed — and that definition should come from your own purchase cycle, not a default setting. If people reorder every six weeks, 90 days of silence is already late. If you sell furniture, 90 days is nothing.

Personalization past "Hi {{first_name}}"

Merge variables are table stakes. Real personalization changes what the email is about, not just who it greets.

Three levels, in increasing order of effort

Merge fields — name, order number, coupon code, the product they left behind. Cheap, and it makes a message feel addressed rather than broadcast.
Segment-level relevance — a different send to first-time buyers than to repeat customers. No dynamic content needed; you simply send two campaigns instead of one.
Catalog-driven content — product blocks pulled from your live catalog, so names, prices, and links are current at send time rather than whatever they were when you built the template.

Most stores jump straight to level three and never do level two properly. Level two is where the easy lift is. Our guide to email personalization goes deeper on each level.

One rule that saves you

Every merge variable needs a fallback. "Hi ," is the classic failure — it happens the moment a contact has no first name, which means it goes out to precisely the people you know least about. Decide what an empty field renders as before you send, not after.

A worked example: a welcome series for a Shopify store

Say you sell skincare and your average order is around $60. A workable first version looks like this:

Immediately — "Here is your 10% off code." Confirm the discount, state what you send and how often, link to a bestsellers collection.
Day 2 — "How to pick the right routine." Educational, no hard sell. It solves the actual blocker for skincare, which is not knowing what to buy.
Day 4 — Social proof. Reviews of two or three specific products, ideally the ones with the strongest repeat rate.
Day 6 — "Your code expires tomorrow." One clear deadline, one call to action.

Four emails, each with a job. Then look at revenue per recipient step by step. If email three earns a fraction of the others, rewrite it or remove it — do not add a fifth just because the template gallery had five.

Measure the metrics that survive contact with reality

Rank them roughly like this:

Revenue per recipient — the number that tells you whether the send was worth sending, and the one that stays comparable across list sizes when total revenue does not.
Conversion rate — orders attributed to the send, divided by recipients.
Click-through rate — whether the content and the call to action landed.
Unsubscribe and complaint rate — your early warning system. A rising complaint rate is a deliverability emergency, not a content note.
Open rate — useful mainly for comparing two subject lines against each other. Apple Mail Privacy Protection pre-fetches images and registers opens that never happened, so treat the absolute number as noise.

One more distinction that catches people out: "sent" is not "delivered". A sending provider accepting your message means it has taken responsibility for it, not that it reached an inbox. Judge campaigns on delivered counts and what happens after. The campaign analytics guide covers attribution and reporting cadence in detail.

Common mistakes worth avoiding

Sending to everyone, every time. Mailing unengaged contacts to hit a volume number damages the reputation that gets you into your engaged subscribers' inboxes.
No exit conditions on flows. Being nagged about something you already bought is the fastest route to an unsubscribe.
Discounting reflexively. If every email carries a code, you have trained your list to wait for one. Save discounts for the steps where the data shows they change the outcome.
Never cleaning the list. Suppress hard bounces immediately, and sunset contacts who have not engaged in six to twelve months. A smaller list that lands in inboxes beats a big one that does not.
Treating email as a silo. Someone who ignored two emails may well answer a WhatsApp message. Sequencing across channels beats sending more email into the same inbox.

Where CampGain fits

CampGain is built for the multi-channel version of this. You connect your sending provider — SendGrid, Resend, or Gmail SMTP — alongside Twilio for SMS and RCS and your WhatsApp Business account, then build a campaign once as a sequence of steps with waits in between. Each step targets whichever channel suits that moment in the journey.

In practice that means:

Segments that combine sources — your own contact rules plus Shopify conditions in a single audience, with duplicate people collapsed by email or phone before anything sends.
A synced catalog — products pull from your live Shopify data, so merge variables resolve to current names, prices, and links.
Per-step and per-contact delivery data — sent, delivered, opened, clicked, and the specific failure reason on anything that did not make it, so you fix the template instead of guessing.
Test sends before launch — fire the whole workflow at a handful of contacts and confirm it renders and delivers before it reaches the full segment.

See how the campaign builder works if you want the specifics.

Where to start

If you do nothing else this month: authenticate your domain, build the welcome and abandoned-cart flows with a purchase exit condition, and start reporting revenue per recipient instead of open rate. That is the shortest path from "we send a newsletter sometimes" to an email programme that produces revenue on a schedule nobody has to keep manually.

Everything after that — browse abandonment, replenishment timing, cross-channel sequencing — is an optimization on a foundation that is already earning.

WhatsApp Business API: A Practical Guide for Ecommerce

Dipikesh Kumar — Sun, 28 Jun 2026 00:00:00 GMT

WhatsApp is the primary messaging channel for a large share of the world, and for ecommerce it does something email and SMS cannot: the customer can reply, and you can answer. That single property makes it the best channel for removing purchase objections.

It is also the channel with the most rules. Templates need approval, messages are priced by category, and there is a 24-hour clock that governs when you can say what you like. This guide explains how the platform actually works, what it costs, how to set it up, and how to design around the constraints rather than fight them.

The two things that govern everything

1. The 24-hour customer service window

When a customer sends you a message, a 24-hour window opens. Inside it you can reply freely — any content, any format, no template, no charge from Meta. When it closes, you can no longer send free-form messages to that person.

Every good WhatsApp programme is organised around this. A message that prompts a reply does not just measure engagement; it opens a window in which your team can have a real conversation at no message cost. That is why "reply YES for sizing help" outperforms a message with only a link.

2. Template messages

To reach someone outside an open window, you need a template Meta has reviewed and approved. Templates have fixed structure with variable slots — you cannot improvise.

Every template carries a category, and the category sets the price:

Marketing — promotions, offers, announcements, re-engagement. The most expensive category, deliberately.
Utility — messages tied to a specific transaction the customer initiated: order confirmations, shipping updates, delivery notifications, appointment reminders. Cheaper than marketing.
Authentication — one-time passcodes and login codes. Priced separately again.
Service — your free-form replies inside a customer-initiated window. Free.

Meta bills per delivered template message, with rates varying by category and by the recipient's country. The practical consequence for planning: your WhatsApp cost is a function of how many marketing templates you send. A programme weighted toward utility messages and conversations costs a fraction of one built on promotional blasts, and usually performs better.

Getting set up

What you need

A Meta Business account, with business verification completed.
A phone number that is not currently active on the WhatsApp consumer app or the Business app. Migrating a number off the Business app is possible but it is a one-way move — plan it deliberately.
A display name that meets Meta's naming policy.
A platform to connect through.

How CampGain connects

CampGain connects to your own WhatsApp Business Account rather than sending on behalf of a shared one. Three routes:

Embedded Signup — a Meta-hosted popup that creates or selects your WhatsApp Business Account, picks the phone number, subscribes the account to webhooks, and registers the number for the Cloud API. This is the standard production path, and it takes minutes rather than days.
Meta test number — connect Meta's free test number manually to try the flow end to end before committing a real number.
Twilio — if you already run WhatsApp through your own Twilio account, connect that instead.

Because the WhatsApp Business Account is yours, your quality rating, your templates, and your number stay with you.

Opt-in is required

Meta requires that you have opt-in before messaging someone, and it must be clear that the opt-in is for WhatsApp specifically from your business. An email subscriber is not a WhatsApp subscriber. The opt-in can be collected almost anywhere — checkout, a website form, a chat entry point — but it has to be explicit and recorded.

Getting templates approved

Template rejection is the most common early frustration, and it is nearly always one of a short list of causes.

Category mismatch. Submitting promotional content as utility because it is cheaper is the single most frequent rejection. Meta reads the content, and repeated mismatches affect your account standing.
Variables with no sample values. Every placeholder needs an example so a reviewer can see what the message looks like filled in.
Variables at the very start or end. A template that begins or ends with a placeholder can render as an incomplete message and is rejected.
Adjacent variables with nothing between them, for the same reason.
Policy breaches — prohibited categories, misleading claims, or unrelated link destinations.
Formatting problems — placeholder text left in, broken structure, unsupported characters.

Two habits that save time. Write templates so they read correctly when every variable is empty or unusually long — that is how they get reviewed and how they eventually render for a real contact. And submit early: approval is not instant, and discovering a rejection the morning of a launch is avoidable.

Quality rating and messaging limits

Meta tracks how recipients respond to your messages and assigns your number a quality rating. Blocks and "report business" actions drive it down; low ratings reduce your messaging limits and can restrict your template categories.

Alongside that sits a messaging limit — the number of unique customers you can message in a rolling day. It increases as you send at good quality, and it drops if quality falls. Which means WhatsApp is not a channel you can scale in one step; you build capacity by sending well.

What protects the rating:

Send fewer marketing templates than you think you should. Frequency is the main driver of blocks.
Make utility messages genuinely useful. Shipping updates are welcome; a "utility" message that is really a promotion is not.
Answer replies. A business that messages and never responds gets reported.
Honour opt-outs immediately, however they are worded. WhatsApp has no universal STOP keyword, so you have to handle this yourself.

What WhatsApp is actually good at

Ranked by how well they usually perform:

Order and shipping updates. High open rates, genuinely wanted, cheap utility category, and they deflect support tickets. Start here.
Cart recovery with a conversation opening. Name the product and invite a question. The reply opens a free window in which your team can close the objection.
Support and pre-sales questions. The channel's native strength — sizing, availability, delivery, returns.
Back-in-stock alerts. The recipient explicitly asked for this message, which makes it the rare promotional send nobody resents.
Post-delivery follow-up. A short check-in that produces reviews and catches problems before they become refunds.

What performs badly: broadcast promotions to your whole list. Expensive per message, hardest on your quality rating, and it teaches people to block you.

Measuring WhatsApp honestly

One measurement trap catches nearly everyone. A WhatsApp read receipt only arrives if the recipient has read receipts enabled — and many people switch them off. Your WhatsApp "open rate" is therefore structurally lower than your email open rate, and the difference says nothing about how the channel performed.

Compare across channels on clicks, replies, and attributed revenue. Use delivery rate to catch real problems: a drop usually means template or policy trouble, not disinterest.

Also worth watching: accepted is not delivered. Meta can accept a message and still not deliver it — a per-recipient marketing cap, for example, will silently drop it. The reason is available in the delivery status, and it is the difference between fixing something and guessing at it. Our campaign analytics guide covers this pattern across channels.

What CampGain provides

Your own WhatsApp Business Account, connected through Meta's Embedded Signup or your Twilio account.
A template builder plus curated presets written against Meta's category rules — correct category, sample values on every variable, and no placeholders stranded at the start or end.
WhatsApp as a step in multi-channel campaigns, with per-recipient template variables drawn from your contact data and synced Shopify catalog.
A shared team inbox. Threads can be assigned to an individual or a whole team, with canned replies and a composer that knows whether the 24-hour window is open — so nobody writes a free-form reply that cannot be sent.
Automated first responses on a visual canvas, with handoff to a human when the conversation needs one.
Owner, admin, and agent roles so a support team can work the inbox without access to campaigns and billing.
Delivery and read status per message, with the provider's stated reason on failures.

See how the WhatsApp features fit together.

A sensible first 30 days

Week 1 — connect your account, submit utility templates for order confirmation and shipping updates, and add a WhatsApp opt-in at checkout.
Week 2 — turn on transactional messages. They are cheap, welcome, and they build the quality rating you will need later.
Week 3 — set up the inbox. Decide who answers replies and how fast. This is the step people skip, and it is the one that determines whether WhatsApp works for you.
Week 4 — add one marketing use case, ideally cart recovery with a reply invitation. Measure replies and attributed revenue, not opens.

Ordering it this way means your first marketing message goes out from a number with a track record, to people who have already had a good experience with you on the channel. That is worth more than any template you could write.

For how WhatsApp sequences alongside email and SMS, see our multi-channel drip campaign guide; for the rules on the SMS side, the SMS compliance guide.

How to Build Multi-Channel Drip Campaigns That Convert

Dipikesh Kumar — Mon, 22 Jun 2026 00:00:00 GMT

Adding channels does not by itself add revenue. Sending the same message by email, then SMS, then WhatsApp is not a multi-channel campaign — it is the same campaign three times, and your unsubscribe rate will say so.

A drip campaign that works across channels does something different: it assigns each step to the channel that suits that moment, and it lets the sequence build. This guide covers how to decide which channel does what, how to time the waits, what each channel actually costs, and the design mistakes that quietly ruin cross-channel sequences.

Campaigns and flows: get this straight first

Two different things share the word "drip", and mixing them up leads to the wrong build.

A scheduled campaign targets an audience you choose, starting when you launch it. The audience is fixed at launch — a product drop announcement, a sale sequence, a re-engagement push.
An event-triggered flow has no audience and no start date. It waits for a shopper to do something and enrolls them individually — cart abandonment, first purchase, browse.

Both can be multi-step and multi-channel. But a campaign's audience is a snapshot, so someone who joins the segment tomorrow is not swept into a run that started today. If you want everyone who ever abandons a cart to get the sequence, you need a flow, not a campaign. Our marketing automation playbook covers the always-on side; this guide is mostly about the sequence design, which applies to both.

What each channel is actually for

Channel choice comes down to three variables: cost per message, how much context it carries, and how much it interrupts.

Email

Cheapest per message, highest capacity for detail, lowest interruption. It is the workhorse: product education, comparisons, imagery, longer stories, and anything where the reader needs to consider before acting. Because it is cheap, it can carry the steps that only pay off occasionally.

Weakness: it is easy to ignore, and a message that must be seen today often is not.

SMS and RCS

Read almost immediately, which makes it the right tool for genuine time pressure — a sale ending, a low-stock alert, a delivery update. It costs meaningfully more than email per message and it is the channel people resent most when misused, so the bar for sending is higher.

RCS is the richer successor to SMS: images, buttons, branded sender identity, and delivery and read receipts. The practical catch is that not every recipient's device and carrier support it, so an RCS message has to be able to fall back to plain SMS — which means your card content needs to still make sense as text, with any button links written into the fallback body. A card whose only call to action is a button becomes a dead end for every recipient who receives the SMS version.

SMS is also the most heavily regulated channel here. Consent, sending hours, and — in the US — carrier registration all apply before your first send. Our SMS compliance guide covers what is required.

The only channel in this list where the customer can reply and have a real conversation, which makes it uniquely good at removing purchase objections. Rich media, buttons, and a thread that persists.

Two rules shape every WhatsApp step you design:

Outbound messages must use a template Meta has approved, in a category (marketing, utility, authentication) that determines what it costs. You cannot improvise a promotional message.
Free-form replies are only allowed for 24 hours after the customer messages you. Inside that window, conversation is free and unrestricted. Outside it, you are back to templates.

That second rule is a design opportunity, not just a constraint: a WhatsApp step that invites a reply opens a 24-hour window in which your team can answer questions freely. Our WhatsApp Business API guide covers setup, categories, and pricing in detail.

Sequencing: match the channel to the intent

The reliable pattern is to escalate. Start on the cheap, low-interruption channel and move to the expensive, high-attention ones only as intent proves itself or a deadline approaches.

Worked example: cart recovery

Hour 1 — Email. The cart contents, a clear path back, no discount. Most recoveries happen here, and they cost you nothing but a send.
Hour 24 — WhatsApp. A short template that names the product and invites a question: "Still deciding? Reply here if you want sizing help." Now the shopper can raise the actual objection, and any reply opens a 24-hour window where your team answers freely.
Day 3 — SMS. Only if the cart value justifies it. A deadline and a link. One sentence.

Three steps, three jobs, escalating cost. Compare that to sending the same "you left something behind" three times.

Worked example: winning back a lapsed customer

Day 0 — Email. What is new since they last bought. No offer yet; the point is to see who still cares.
Day 4 — Email. A reason to return aimed at the category they used to buy from.
Day 8 — WhatsApp or SMS, but only to the people who clicked one of the first two. Interest earned the expensive channel.

Gating the paid channels on engagement from the free one is the single highest-leverage habit in multi-channel campaigns. It cuts spend and it stops you interrupting people who have already told you no.

Timing the waits

There is no universal cadence, but there is a rule: the wait should match the decay rate of the intent you are responding to.

Cart abandonment — hours. Intent evaporates fast; a two-day delay is a different campaign entirely.
Browse abandonment — 12 to 24 hours. Real interest, not yet a decision.
Welcome series — days. You are building a relationship, not closing today.
Win-back — days to weeks. The customer is not in-market right now, and pressure reads as desperation.

Two practical adjustments. Respect quiet hours on SMS and WhatsApp — a 3am message is a lost subscriber and, in some jurisdictions, a violation. And do not let a wait be so long that the context is gone: an email about a cart from three weeks ago is confusing rather than persuasive.

Design mistakes that break cross-channel campaigns

No exit condition. If someone buys at step one, steps two and three must not send. This matters more with every channel you add — a mistimed email is annoying, a mistimed WhatsApp message about something already purchased is worse and costs you money to deliver.
Duplicate contacts across sources. The same person imported from two systems receives everything twice. Deduplicate on email and phone before you launch, not after someone complains.
Unresolved merge variables. A message that goes out reading "Hi {{first_name}}" or with a broken product link is worse than no message. Give every variable a fallback and send a test first.
Treating a WhatsApp read receipt like an email open. Read receipts only arrive when the recipient has them switched on, so WhatsApp "open" numbers are structurally lower than email's. That is a measurement artefact, not weaker performance — compare clicks and conversions instead.
Ignoring per-channel consent. Someone who subscribed to your newsletter has not agreed to receive texts. Consent is per channel and needs to be recorded that way.
Assuming the same length works everywhere. A paragraph that reads well in an email is a wall of text on a phone. Rewrite per channel; do not truncate.

Measuring a multi-channel campaign

Per-channel metrics will mislead you here, because the channels are supposed to work together. Judge the sequence at two levels:

The whole campaign — revenue and conversions attributed to the run, against its total sending cost across all channels. This is the number that answers "was this worth it".
Each step — delivered, clicked, and conversions after that step, so you can tell which one is carrying the sequence and which is filler.

Two specifics worth watching. First, "sent" is not "delivered" — a provider accepting a message means it took responsibility, not that it arrived, and the gap between the two is where template rejections and bad phone numbers hide. Second, look at where conversions cluster: if 80% land after step one, your later steps are costing money to annoy people. Cut them. Our campaign analytics guide covers attribution properly.

How CampGain runs this

CampGain is built specifically for this shape of campaign: one workflow, multiple channels, waits in between.

Build the sequence once. Add steps, pick a channel and a template per step, and set the wait between them. Email goes through your own SendGrid, Resend, or Gmail connection; SMS and RCS through Twilio; WhatsApp through Meta or Twilio.
One audience across every step. Segments combine your own contact rules with Shopify conditions, and duplicates are collapsed by email or phone before the run starts, so nobody gets the sequence twice.
Merge variables that resolve per recipient, drawing on your synced Shopify catalog so product names, prices, and links are current at send time.
RCS with automatic SMS fallback, including carrying the card's button links into the fallback text so recipients on plain SMS still get somewhere to click.
Launch, pause, resume, or cancel a run mid-flight, and restart a finished campaign as a fresh copy with its own statistics rather than rebuilding it.
Per-step and per-contact results, with the provider's stated reason on every failure — so a step that underperforms tells you whether it was the template, the audience, or the channel.
Test the whole workflow first by sending it to a handful of contacts and checking every step renders and delivers before the segment sees it.

See how a campaign gets built, from connecting channels to launching the run.

Start with two channels

Take one sequence you already run over email — cart recovery is the usual candidate — and add a single non-email step, gated on people who did not convert from the first message. Compare revenue per recipient against the email-only version over a few weeks.

That one comparison teaches you more about your customers' channel preferences than any benchmark will, and it costs a fraction of building a full cross-channel programme you have not yet validated.

Ecommerce Customer Segmentation: A Practical Guide

Laxman Anand — Mon, 15 Jun 2026 00:00:00 GMT

Most stores segment on the wrong axis. They split a list by country or by gender, send two versions of the same email, and see no difference — because neither group was defined by anything that predicts buying.

Useful segmentation starts from a different question: what would I say differently to this group? If the answer is nothing, the segment should not exist. This guide covers the segments worth building, how to layer them from simplest to most sophisticated, and the practical traps — duplicates, stale definitions, over-segmentation — that make segmentation cost more than it returns.

A segment is only real if it changes the message

The test is simple. Write the campaign you would send to the segment, then write the campaign you would send to everyone else. If they are the same email with a different name in the greeting, you have not segmented — you have added maintenance work.

This is why "customers in California" is usually a bad segment and "customers who bought once, 45 days ago, and have not returned" is a good one. The second implies a message. The first does not.

Practically, that means starting from campaigns and working backwards to segments, rather than building a segmentation model and hoping campaigns emerge from it.

Layer 1: engagement tiers

The first split every store should have, because it protects deliverability as much as it improves relevance.

Engaged — opened, clicked, or ordered in the last 30 days.
Cooling — last engagement 30 to 90 days ago.
At risk — 90 to 180 days of silence.
Dormant — no engagement in 180 days or more.

What you do with these matters more than the boundaries. Engaged contacts get your regular sends. Cooling contacts get your best offers. At-risk contacts get a win-back attempt. Dormant contacts get one re-permission campaign and then get suppressed — continuing to mail them is the single most common cause of the deliverability problems described in our deliverability guide.

Adjust the windows to your purchase cycle. A coffee subscription and a mattress store should not use the same definition of dormant.

Layer 2: RFM scoring

RFM is the most useful segmentation model in ecommerce because it needs only order data — which you already have — and it maps cleanly onto actions.

How to build it without a data team

Recency — days since last order. Sort customers, split into five equal groups, score 5 (most recent) down to 1.
Frequency — total orders. Same process, same 1–5 scale.
Monetary — total spend. Same again.

Scoring by quintile rather than by fixed thresholds matters: it makes the model relative to your own customer base, so it works whether your average order is $20 or $2,000, and it keeps working as the business grows.

The groups that fall out, and what to send them

Champions (high on all three) — early access, new arrivals, referral asks. Never discount to this group; they were going to buy anyway.
Loyal (high frequency, mid recency) — replenishment reminders and cross-sells based on what they already own.
Potential loyalists (recent, one or two orders) — the highest-leverage group in most stores. A well-timed second-purchase nudge is worth more than almost any acquisition spend.
At risk (previously high frequency, recency falling) — a genuine win-back with a real reason to return.
Lost (low on everything) — one last attempt, then suppress. Sending forever costs reputation and returns nothing.

Notice how different those five messages are. That is what a working segmentation model looks like.

Layer 3: behavioural segments from on-site events

Order data tells you what people bought. Event data tells you what they were about to buy — which is where the timely campaigns live.

Useful behavioural conditions, all derived from ordinary storefront tracking:

Has an abandoned cart — added to cart, no purchase since.
Viewed a product more than once without adding to cart — interest without commitment, usually a question you have not answered.
Started checkout and stopped — the highest-intent group on your whole site.
Category affinity — repeated views or purchases in one category, which makes new-arrival campaigns actually relevant.
Time since last product view — separates "went quiet as a customer" from "went quiet entirely".

One constraint worth being honest about: you can only act on behaviour for shoppers you can identify. An anonymous visitor who browses ten products and never gives you an email or phone number cannot be enrolled in anything. If a behavioural segment looks suspiciously small, the cause is usually identification, not tracking.

Layer 4: predictive-style segments without a model

You do not need machine learning to act on the pattern. Most of what predictive segmentation delivers can be approximated with conditions you can write down:

Likely to buy soon — an active cart, or several product views in the last few days.
Churn risk — order frequency slowing relative to their own history, not to the store average.
High-value trajectory — a second order inside the window where your best customers placed theirs.
Due for replenishment — days since purchase approaching the typical reorder interval for that product.

Replenishment timing in particular is worth computing properly: take the median gap between repeat orders per product category and trigger a few days before it. That single segment often outperforms every clever idea in the account.

A starter set: seven segments most stores should build

New subscribers, never purchased — feeds the welcome flow.
First-time buyers, 14–45 days ago — the second-purchase push.
Repeat customers — cross-sell and loyalty.
Champions — early access, referral asks, review requests.
Cart or checkout abandoners — recovery.
At risk — win-back before they are gone.
Engaged-only sending audience — the safe list for high-volume moments and for warming a new sending domain.

Seven segments, each with an obvious campaign attached. That is a complete programme for most stores, and it is achievable in an afternoon.

Mistakes that make segmentation expensive

Building segments nobody sends to. Every unused segment is a definition that will silently drift out of date and eventually mislead someone.
Slicing too thin. Splitting a 400-person segment into four means you can never tell whether a result is real. Keep segments large enough that the numbers mean something.
Static lists that never refresh. A CSV exported in March is wrong by April. Segments should be conditions that re-evaluate, not snapshots.
Ignoring duplicates. The same person appears as a Shopify customer and as a manually imported contact under a different email or a differently formatted phone number. They then receive the campaign twice — the single fastest way to look careless.
Segmenting on data you do not have. A condition on a field that is empty for 80% of contacts is not a segment; it is a way to accidentally exclude most of your list.

How CampGain handles this

CampGain's segment builder is designed around the two problems above — conditions that stay current, and people who exist in more than one system.

One segment, multiple sources. A single segment can combine your own contact conditions with Shopify conditions, so "lapsed Shopify buyers who are also on my newsletter list" is one audience rather than two lists you reconcile by hand.
AND/OR condition groups. Nest rules to express the actual definition instead of approximating it.
Cross-source deduplication. Members are matched on normalised email or phone, so the same person arriving from two sources is counted once. The segment shows you the duplicate count rather than hiding it.
Refresh on demand. Re-sync a segment and every source re-evaluates, with member counts and last-synced time visible per source.
Drill into members. Open a segment and see the actual people in it before you send — the fastest sanity check there is on a condition you just wrote.

Segments then feed straight into campaigns, including multi-channel sequences that reach the same audience over email, SMS, and WhatsApp. Our multi-channel drip campaign guide covers how to sequence those steps, and the campaign analytics guide covers how to tell which segments are actually producing revenue.

Start with three

If this feels like a lot, build three segments this week: engaged contacts, first-time buyers from the last 45 days, and cart abandoners. Attach one campaign to each. Measure revenue per recipient against your usual send-to-everyone baseline.

That comparison — same store, same products, different audience definition — is the argument for segmentation, and you will have it in a fortnight. See how segments work in CampGain when you are ready to build them.

Ecommerce Marketing Automation: The Flows Playbook

Laxman Anand — Wed, 10 Jun 2026 00:00:00 GMT

The usual pitch for marketing automation is that it saves time. That is true and it is the least interesting thing about it.

The real argument is timing. A shopper abandons a cart at 11pm on a Tuesday. Nobody is going to notice, look up what they left, and write to them — but an automation will, an hour later, with the right products in the message. Automation is the only way to be present at the moment intent is highest, for every customer, indefinitely.

This guide covers how flows are actually constructed, the six worth building first, and the production details — exit conditions, re-entry rules, identity — that decide whether a flow earns money or quietly embarrasses you.

The anatomy of a flow

Every automation, in every tool, is the same five parts. Understanding them separately is what lets you debug one.

1. Trigger

The event that starts enrollment: added to cart, started checkout, placed an order, subscribed, viewed a product. A flow does nothing until an event arrives that matches its trigger.

2. Entry conditions

Filters applied at the moment of the trigger. "Started checkout and cart value is over $100" behaves very differently from "started checkout", and lets you justify an expensive channel for high-value carts only.

3. Delays

The wait between steps. This is where most of the tuning happens, and the rule is that the delay should match how fast the intent decays — hours for a cart, days for a welcome series, weeks for a win-back.

4. Branches

Conditional splits that send different people down different paths. The genuinely useful ones are simple: did they open the previous message, is this their first order, is the cart above your threshold. Two-way splits on facts you can rely on beat elaborate trees on data you cannot.

5. Exit conditions

The events that remove someone mid-flow. Most flows need at least one, and getting this wrong is the most damaging mistake in the whole category — enough that it gets its own section below.

The six flows to build first

1. Abandoned cart

Trigger: added to cart, no purchase since. Exit: purchase.

Baymard Institute's aggregate of published studies puts cart abandonment at roughly seven in ten, which makes this the largest single pool of recoverable revenue most stores have. Two or three messages: a reminder with the actual cart contents within a few hours, a follow-up that answers the real objection, and optionally a final nudge with a deadline.

Resist the urge to lead with a discount. A meaningful share of abandonment is distraction, not price — those people convert on a reminder, and discounting them costs you margin you did not need to spend.

2. Abandoned checkout

Trigger: checkout started, not completed. Exit: purchase.

Worth separating from cart abandonment because the intent is much higher and the failure is usually mechanical — shipping cost at the last step, a payment method that was not offered, a forced account creation. Address those directly in the message rather than repeating the product pitch.

3. Welcome series

Trigger: subscribed. Exit: first purchase, optionally.

Three to five messages over the first week. Set expectations, explain what makes your product different, show proof, then give a reason to buy. If someone purchases at message two, either exit them or branch them into post-purchase — continuing to sell the first order to someone who placed it is a small failure people notice.

4. Post-purchase

Trigger: order placed. Exit: order cancelled or refunded.

Order confirmation, shipping update, a usage or care email timed to arrival, then a review request a week or two later. This flow does double duty: it builds the repeat purchase and it deflects a real share of "where is my order" support tickets.

WhatsApp is often the better channel here. Order and shipping updates are exactly the kind of message people are glad to receive on it, and utility-category messages are cheaper to send than marketing ones.

5. Browse abandonment

Trigger: product viewed repeatedly, no cart add. Exit: cart add or purchase.

One message, helpful in tone. Someone looking at the same product three times usually has an unanswered question — sizing, materials, delivery time, returns. Answer it rather than repeating the price.

6. Win-back

Trigger: days since last order crosses your lapsed threshold. Exit: purchase.

Set the threshold from your own repeat purchase interval, not a default. Two or three messages spread over a few weeks: what is new, then a reason to come back, then a genuine last attempt before you suppress them for good.

Exit conditions: the part everyone skips

An abandoned-cart flow with no exit condition will, reliably, message customers about the thing they just bought. It is the fastest way to look like you are not paying attention, and it happens because exits are easy to forget: the flow works perfectly in testing, where you never complete a purchase.

Rules worth applying:

Every recovery flow exits on purchase. Cart, checkout, browse — no exceptions.
Post-purchase flows exit on cancellation or refund. "How are you enjoying your order?" to someone who returned it is worse than silence.
Win-back exits on any order. The moment they are back, they belong in the normal programme.
Consider a global suppression for recent buyers. Someone who ordered yesterday probably should not receive a promotional flow today, whatever triggered it.

Test exits deliberately: enroll yourself, then complete the goal action, then confirm the remaining messages do not arrive. That test takes ten minutes and is the highest-value one in this entire guide.

Re-entry rules and duplicate triggers

Two production problems that only appear once a flow is live.

Duplicate events. Platform webhooks are not guaranteed to fire exactly once. A fulfilment webhook can arrive twice, and a naive flow sends two shipping notifications. The fix is a short re-entry cooldown — an hour is usually enough — so a repeat trigger for the same contact is absorbed rather than acted on.

Re-entry policy. If a shopper abandons a cart in March and again in June, should they get the sequence twice? For cart recovery, yes — it is a new intent. For a welcome series, absolutely not. Set this per flow rather than accepting a default, and set a cooldown even where re-entry is allowed, or a shopper who repeatedly adds and removes items will get messaged repeatedly.

Identity: who can actually be enrolled

This is the constraint that surprises people most, and it explains most "why is my flow reaching nobody" questions.

A flow can only enroll a shopper the triggering event can identify — the event has to carry an email address or a phone number. An anonymous visitor who browses, adds to cart, and leaves without ever giving you either cannot be enrolled, no matter how well the flow is built.

What raises the identified share:

Capture email early — a newsletter signup, an account, or a first-step checkout field, rather than only at payment.
Make sure your tracking passes identity through. An event that fires but carries no email is invisible to enrollment even though it appears in your analytics.
Stitch sessions. If a shopper identifies themselves at checkout, earlier events in the same session can be attributed to them retrospectively — which turns a browsing session into an enrollable one.

Check this number before you blame the flow. Events arriving and events carrying identity are two different measurements, and only the second one produces enrollments.

Common mistakes

Building ten flows before validating one. Each live flow is something that can go wrong silently. Get two right, then add.
No exit conditions. Covered above, and worth repeating because it remains the most common failure.
Delays copied from a template. A template's timing was tuned for someone else's purchase cycle. Yours is different.
Branching on data you rarely have. A split on a field that is empty for most contacts sends almost everyone down the default path — and you will not notice, because it still looks like it is working.
Never auditing live flows. Catalogs change, templates get archived, integrations get disconnected. A flow that worked in January can be silently failing in June. Review each one quarterly.
Overlapping flows. A shopper who abandons a cart during a welcome series can end up in both. Decide which one wins and enforce it, or they get four messages in a day.

How CampGain runs flows

In CampGain, a flow is an event-triggered campaign built on a visual canvas, separate from scheduled blast campaigns because the two behave differently.

Start from a template or from scratch. Ready-made flows cover abandoned cart, abandoned checkout, welcome series, post-purchase, review request, win-back, and a transactional WhatsApp set for order confirmation, shipping, and delivery — each shipping with sensible timing and exit events already configured.
Drag-and-drop graph builder. Message nodes, delays, and conditional splits on a canvas, so the flow you see is the flow that runs.
Exit events on the trigger. Purchase exits a cart flow; cancellation and refund exit a transactional one. Configured on the flow, not bolted on per message.
Re-entry cooldowns so duplicate platform webhooks do not produce duplicate messages.
A readiness panel that checks a flow before you activate it — message nodes with no template chosen, channels with no connected provider, nodes nothing can reach, and abandonment flows with no exit event. Clicking a warning jumps to the node that caused it.
Mixed channels in one journey. A flow can send email, then WhatsApp, then SMS, using the same merge variables and the same synced Shopify catalog throughout.
Event ingestion built in — a storefront pixel and Shopify webhooks feed the events that trigger enrollment, with a live feed so you can confirm they are arriving and carrying identity before you go live.

See how flows and campaigns fit together.

Start with one

Build abandoned cart. Two messages, an exit on purchase, and a test where you enroll yourself and then complete the purchase to confirm the second message does not arrive.

Let it run for a month and measure revenue per enrollment. That single flow usually justifies the whole exercise — and it teaches you the mechanics you will reuse on every flow after it. For how these flows fit alongside scheduled sends, see our multi-channel campaign guide; for the audiences behind them, the segmentation guide.

Campaign Analytics: The Metrics That Actually Predict Revenue

Laxman Anand — Fri, 05 Jun 2026 00:00:00 GMT

Most campaign reporting measures the wrong things carefully. Open rate to two decimal places, click-through rate charted over twelve months, and no answer to the only question that matters: which of these sends should we do again?

This guide sorts campaign metrics by how much they actually tell you, explains where the standard numbers mislead, covers how attribution works and where it stops working, and lays out a reporting structure you can act on.

The metric hierarchy

Rank metrics by how directly they connect to a decision.

Tier 1 — outcome metrics

Revenue per recipient — attributed revenue divided by delivered messages. The best single number for comparing campaigns, because it is unaffected by audience size.
Attributed orders and revenue — the raw totals. Useful for the business, less useful for comparison.
Contribution margin per send — revenue per recipient minus cost per message. This is what separates channels: an email step and a WhatsApp step with the same revenue per recipient are not equally good.

Tier 2 — conversion metrics

Conversion rate — attributed orders divided by delivered messages.
Click-to-conversion rate — of the people who clicked, how many bought. This isolates the landing experience from the message. A high click rate with a low click-to-conversion rate means the email wrote a cheque the product page could not cash.
Average order value from the campaign — reveals whether a discount bought you volume at the cost of margin.

Tier 3 — diagnostic metrics

Delivery rate — the health check on your list and your templates.
Click-through rate — whether the content and call to action worked.
Unsubscribe and complaint rate — the cost side of frequency. Rising complaints is an emergency, not a footnote.
Open rate — subject line A/B comparison within a single send, and nothing else.

Report tier one. Investigate with tiers two and three.

Two numbers that lie to you

"Sent" is not "delivered"

When a provider returns success, it means the message was accepted for delivery — it took responsibility. It does not mean anything arrived. Messages can still bounce, be silently dropped, hit a per-recipient cap, or fail because a template was rejected.

The practical consequence: calculate every rate against delivered. A campaign with 10,000 sent and 9,100 delivered has a conversion rate 10% better than the one you would report against sent — and, more importantly, a 900-message problem you should be looking at.

This matters most on the channels people check least. Email bounces are visible; a WhatsApp template silently failing a per-recipient limit, or an RCS message that quietly downgraded to SMS, are the ones that go unnoticed for months.

Open rate is a measurement artefact

Opens are tracked with an invisible image. Apple Mail Privacy Protection loads that image on the recipient's behalf regardless of whether they read anything, and other clients block it entirely. So your open rate is a blend of real opens, phantom opens, and invisible real opens.

It remains useful in exactly one situation: comparing two subject lines in the same send to the same audience at the same time, where the distortion applies equally to both. Any other use — trend lines, benchmarks against other brands, leadership reporting — is measuring your list's device mix.

Channel-specific version of the same trap: a WhatsApp read receipt only arrives if the recipient has read receipts enabled. WhatsApp "open rates" are therefore structurally lower than email's, which says nothing about performance. Compare clicks and conversions across channels; never compare opens.

How attribution actually works

Attribution answers: when this order arrived, which campaign should get the credit?

Why URL parameters are not enough

The intuitive approach is UTM tags: a shopper clicks a tagged link, the parameters land in the URL, and the order records them. It falls apart in ordinary use. The shopper clicks the link, browses, refreshes, navigates to a different product — and the parameters are gone. Later they return directly and buy. From the URL's point of view, that purchase came from nowhere.

Robust attribution records the touch server-side: when a contact clicks a campaign link, that click is stored against the contact. When an order later arrives — including via a webhook with no browser context at all — it is matched back to the most recent qualifying touch inside your lookback window.

Last-touch, and being honest about it

Most platforms, including CampGain, use last-touch attribution: the most recent qualifying interaction gets the credit. It is the right default because it is simple, stable, and comparable across campaigns.

It is also an estimate. Real customers see an email, a social ad, and a WhatsApp message before buying. Last-touch hands all the credit to the final click. Multi-touch models spread credit across the journey, but they replace one arbitrary assumption with several.

The workable position: use last-touch consistently for comparing campaigns to each other, and never present attributed revenue as the store's total revenue. They are different numbers measuring different things, and conflating them is how marketing teams end up claiming more revenue than the business made.

The lookback window

How long after a click can an order still be credited? A short window under-credits considered purchases; a long one credits campaigns for orders they had nothing to do with. Match it to your purchase cycle — a few days for impulse categories, longer for high-consideration products — and then leave it alone, because changing it retroactively rewrites your history.

The funnel is where diagnosis happens

Campaign metrics tell you a send underperformed. The on-site funnel tells you why. Track four stages:

Product viewed — traffic arrived and looked at something.
Added to cart — interest became intent.
Checkout started — intent became commitment.
Purchased — commitment survived checkout.

The drop-off between consecutive stages localises the problem in a way no email metric can:

Views but few cart adds — the message promised something the product page did not deliver. Wrong products, wrong price expectation, or a landing page mismatch.
Cart adds but few checkouts — shipping cost or delivery time is doing the damage. Classic, and fixable by stating both in the email.
Checkouts started but few purchases — payment options, a forced account creation, or an unexpected total.

Each of those is a different fix, and none of them is "rewrite the subject line" — which is where teams reflexively go when a campaign underperforms.

Build a dashboard around decisions

Design each view around the decision its reader is about to make.

Daily, for whoever runs sends — did yesterday's campaigns deliver, are complaints or failures spiking, is anything stuck. Operational, scanned in a minute.
Weekly, for the marketing lead — revenue per recipient by campaign and by channel, top and bottom performers, funnel drop-off. This is where next week's plan comes from.
Monthly, for the business — attributed revenue by channel, list growth net of unsubscribes, repeat purchase rate. Trends, not events.

Two rules keep dashboards useful. Every metric needs a comparison — versus last period, versus the campaign average — because a number with nothing beside it cannot be judged. And every headline metric should have an owner who would change something if it moved.

Testing: fewer, bigger tests

Most campaign A/B testing produces noise, because ecommerce list sizes cannot detect the size of difference being tested.

Test one variable at a time, and make it a real difference — two versions of an offer, not two synonyms in a subject line.
Decide the metric before you send. If you test subject lines and evaluate on revenue, you are measuring the whole email, not the subject line.
Run to a pre-committed sample size. Stopping when one variant looks ahead manufactures winners that never repeat.
Retest anything that mattered. A result that does not survive a second run was noise.

Test structural things — send timing, discount versus no discount, three steps versus five, channel order. Those produce differences large enough for a normal list to detect.

What CampGain measures

CampGain reports on two layers: what your campaigns did, and what shoppers did on your store.

Per-step and per-contact campaign results — sent, delivered, opened, and clicked for every step and every recipient, with the provider's stated reason on each failure. A step that underperforms tells you whether it was the audience, the template, or the channel.
Server-side attribution. Campaign clicks are recorded against the contact, so an order arriving later by webhook — with no URL parameters anywhere — is still matched to the campaign that earned it.
Attributed revenue by campaign, by step, and by channel, plus a daily series, so you can see whether the WhatsApp step or the email step carried the sequence.
A storefront funnel — product views, cart adds, checkouts started, and purchases over your chosen period, with conversion rate between each stage and the top products by view and cart rate.
A live event feed for checking that tracking is working before you trust a report built on it.

Segments feed campaigns and campaigns feed these reports, so "which segment produces revenue" is answerable rather than inferred. Our segmentation guide covers building the audiences, and the multi-channel campaign guide covers sequencing the sends. See the analytics features for the full list.

The short version

Report revenue per recipient, calculated on delivered rather than sent. Use open rate only for subject line comparisons within a single send. Pick last-touch attribution with a lookback window that matches your purchase cycle, and never present attributed revenue as total revenue. Diagnose underperformance in the funnel rather than in the email.

Do those five things and your reporting stops being a monthly ritual and starts being the reason next month's campaigns are better than this month's.

Email Deliverability: How to Land in the Inbox, Not Spam

Arpit Anand — Sat, 30 May 2026 00:00:00 GMT

Deliverability is the one part of email marketing where doing nothing has a cost. Your best-written campaign converts nothing from the spam folder, and by the time you notice — revenue quietly down, opens quietly down — you have usually been filtered for weeks.

This guide covers the technical setup that gets you trusted, the rules Gmail and Yahoo now enforce on bulk senders, how to warm a new sending domain, and a diagnostic sequence for when your mail is already landing in spam.

What deliverability actually measures

Two terms get used interchangeably and should not be.

Delivery rate — the share of messages a mailbox provider accepted rather than bounced. Easy to measure. Nearly always looks fine.
Deliverability — the share that reached the primary inbox rather than spam or a promotions bucket. Harder to measure and the number that determines your revenue.

A 99% delivery rate is perfectly compatible with half your mail sitting in spam. Mailbox providers accept and then quietly file. That gap is why a sudden drop in clicks with no change in bounces is nearly always a deliverability problem, not a copy problem.

Authentication: the part that is not optional

Authentication proves that mail claiming to come from your domain actually did. Without it, bulk mail is filtered or rejected outright.

SPF

A DNS TXT record listing which servers may send for your domain. Two things to watch: you may only publish one SPF record per domain (multiple records is a hard failure, and it happens constantly when a second tool is added), and SPF has a limit of ten DNS lookups — chained includes silently blow past it.

DKIM

A cryptographic signature added to every outbound message, with the public key published in DNS. Your sending provider generates the key pair and tells you which records to publish. Use a selector per provider so you can add or remove a sender without breaking the others.

DMARC

The policy layer. DMARC tells mailbox providers what to do when SPF or DKIM fails, and where to send aggregate reports. Publish it in three stages:

p=none — monitor only. Collect reports for two to four weeks and find every legitimate sender you had forgotten about (the invoicing tool, the helpdesk, the CRM).
p=quarantine — failing mail goes to spam. Roll out with a percentage tag first if your volume is meaningful.
p=reject — failing mail is refused. The end state, and the one that actually protects your domain from spoofing.

The detail people miss is alignment. DMARC does not just require SPF or DKIM to pass; it requires the passing domain to match the domain in your visible From address. Mail can pass SPF for your provider's domain and still fail DMARC. Check alignment, not just pass/fail.

BIMI

Optional, and only eligible once DMARC is at quarantine or reject. It displays your logo next to your messages in supporting clients, which helps recognition. Treat it as a finishing touch, not a fix.

The bulk sender rules you have to meet

Google, Yahoo, and Microsoft have converged on a shared baseline for high-volume senders — roughly 5,000 or more messages a day to a given provider's users. The requirements:

SPF and DKIM on every message, with at least one aligned to your From domain.
A published DMARC record on the sending domain. p=none satisfies the letter of it; enforcement is better.
One-click unsubscribe in the message headers (RFC 8058, the List-Unsubscribe-Post header), honoured within a couple of days. A visible unsubscribe link in the body is still required — this is in addition to it.
Valid forward and reverse DNS on the sending IP, and TLS on the connection.
A spam complaint rate that stays low. Google publishes 0.30% as the enforcement threshold and recommends staying under 0.10%.

Treat 0.10% as the ceiling, not the target. Complaint rate is a lagging indicator: the filtering starts before the number in your dashboard catches up.

If you send through CampGain, the provider you connect — SendGrid, Resend, or Gmail SMTP — handles the transport-level items. The DNS records and the list quality are yours.

Reputation: domain first, IP second

Sender reputation is built from how recipients treat your mail, and it attaches to two things.

Domain reputation follows your sending domain everywhere. It is the one that matters most for typical ecommerce volume, and it does not reset when you switch providers.
IP reputation attaches to the sending IP. On a shared pool you inherit the pool's behaviour; on a dedicated IP you build your own from zero.

Which means the dedicated-IP question has a clearer answer than it usually gets: a dedicated IP only pays off if you send enough, consistently enough, to establish a pattern. Below that, sparse sending from a dedicated IP looks like a dormant sender waking up every few weeks — which is exactly what spam looks like. Most stores should stay on a shared pool and put the effort into domain reputation and list quality.

The signals that move it

Complaints — the heaviest negative signal by a wide margin.
Hard bounces — invalid addresses suggest a list you did not build yourself. Keep under about 2%, and suppress permanently on the first hard bounce.
Spam trap hits — recycled or pristine addresses that exist only to catch bad list practice. Zero tolerance; a hit means your acquisition or hygiene is broken.
Positive engagement — opens, clicks, replies, and being moved out of spam all count in your favour. This is why sending only to engaged contacts during a repair actually works.

Warming a new sending domain

A new domain has no history, and mailbox providers treat "no history plus sudden volume" as the classic spam signature. Ramp deliberately:

Weeks 1–2 — send only to contacts who engaged in the last 30 days, at low daily volume. These people open, and those opens are the reputation you are buying.
Weeks 3–4 — expand to 60- then 90-day engaged contacts, increasing volume gradually. Watch bounce and complaint rates after every step and hold if either moves.
Weeks 5–6 — reach full volume, keeping unengaged contacts out until the ramp is complete.

The mistake is treating warming as a volume schedule alone. Who you send to during the ramp matters more than how many. Warming on a cold list builds a bad reputation faster than not warming at all.

List hygiene is a deliverability control

Most "our deliverability collapsed" stories are list stories.

Suppress hard bounces immediately and never re-add them, even if they come back through a form.
Set a sunset policy. Contacts with no opens, clicks, or orders in six to twelve months should stop receiving regular sends. Give them one re-permission campaign, then let them go.
Use double opt-in where acquisition is noisy — paid lead magnets, giveaways, offline capture. It costs you subscribers and saves you a sending programme.
Never buy, rent, or scrape a list. Purchased lists are how you hit spam traps, and no amount of technical setup survives that.

The counterintuitive part: cutting your list often increases total revenue, because the mail that remains reaches people who wanted it. Our customer segmentation guide covers how to build engagement tiers you can actually send against.

Content signals that still matter

Content is a smaller factor than authentication and reputation, but it is not zero:

Keep a real text-to-image balance. An email that is one giant image gives filters nothing to read and breaks entirely when images are blocked.
Send a plain-text alternative alongside the HTML. Its absence is a small negative signal and it costs nothing.
Keep your From name and address consistent. Recognition drives engagement, and engagement drives placement.
Watch link reputation. Link shorteners and domains you do not control are a genuine risk — a shortener that other people abuse drags your mail down with it.
Write like a person. Chasing a "spam word list" is mostly folklore, but all-caps subject lines, walls of exclamation marks, and hidden text are real patterns filters recognise.

Diagnosing a live problem

If your mail is already going to spam, work in this order rather than changing five things at once.

Verify authentication end to end. Send to an address you control at each major provider and inspect the headers for SPF, DKIM, and DMARC results — including alignment.
Check Google Postmaster Tools. It shows your domain reputation and spam rate for Gmail. A complaint rate spike will usually name the date the trouble started.
Line the date up against what you changed. New sending domain, new template, a big import, a list you had not mailed in a year — the cause is normally in that list.
Segment by provider. If Gmail is fine and Outlook is not, it is a provider-specific reputation issue, not a broken record.
Shrink to your engaged core. Send only to contacts who engaged in the last 30 days for two to three weeks. Rebuilding positive signal is slower than losing it, and there is no shortcut.
Fix acquisition before you scale back up. If the source of the bad contacts is still running, you will be back here in a quarter.

Where CampGain helps

CampGain does not replace your DNS setup — no platform can — but it removes the guesswork from what happened after you pressed send.

Per-contact delivery detail. Every campaign shows sent, delivered, opened, and clicked per step and per recipient, with the provider's stated reason on anything that failed. A bounce that says why is a bounce you can act on.
Test sends before launch. Send the full workflow to a handful of addresses, check rendering and headers, and only then release it to the segment.
Segments that support a sunset policy. Combine your own engagement conditions with Shopify data to build the engaged-only audiences a warm-up or a repair depends on.
Your own sending provider. You connect SendGrid, Resend, or Gmail SMTP, so the reputation you build stays yours.

See what the platform includes for the full picture.

The short version

Publish SPF, DKIM, and DMARC and confirm alignment. Meet the bulk sender baseline before you need to. Warm new domains on engaged contacts only. Suppress bounces, sunset the dormant, and never buy a list. Then watch complaint rate as closely as you watch revenue — it is the metric that predicts whether the rest of your email programme gets to work at all.

With that foundation in place, the campaigns in our ecommerce email marketing guide can actually reach the people you built them for.

SMS Marketing Compliance: TCPA, 10DLC and Global Rules

Dipikesh Kumar — Mon, 25 May 2026 00:00:00 GMT

SMS is the highest-attention channel in ecommerce and the one with the most ways to get it wrong. The regulations carry real penalties, the US carriers will filter unregistered traffic before a regulator ever gets involved, and both failure modes look identical from your dashboard: messages that appear to send and never arrive.

This guide covers what you need in place before your first marketing text — consent, registration, opt-out handling, timing — and what changes outside the US. It is a practical overview rather than legal advice; specifics vary by jurisdiction and by what you sell, so run your programme past counsel before you scale it.

Consent: the foundation everything else rests on

In the US, marketing texts require prior express written consent. Three things make consent valid, and all three are commonly botched.

It must be explicit and specific

The person must actively agree to receive marketing texts from you, at that number. Buried in your terms of service does not count. A pre-ticked box does not count. Consent to email does not extend to SMS, and consent for order updates does not extend to promotions.

The disclosure must be complete

At the point of sign-up, state clearly:

Who is sending (your business name).
What kind of messages (marketing, promotions, offers).
How often, at least approximately — "up to 4 messages per month".
That message and data rates may apply.
How to opt out.
A link to your privacy policy and terms.

Consent must also not be a condition of purchase. You cannot require someone to accept marketing texts in order to check out.

You must be able to prove it later

Keep, for every contact: the timestamp, the source (which form, which page), the exact disclosure language shown at the time, and the IP or session where available. If your consent language changes, keep the old version — you need to show what that person agreed to, not what your current form says.

This record is the entire defence in a complaint. Storing a phone number with no provenance is the same as having no consent.

Opt-outs: the rules tightened

Opt-out handling used to be simple: honour STOP. It is broader now.

Any reasonable method counts. A reply saying "unsubscribe", "cancel", "quit", "end", or a plainly worded "please stop texting me" is a valid revocation. So, in principle, is an opt-out sent by email or left on voicemail.
Ten business days is the outer limit. Marketing opt-outs must be honoured as soon as practicable and no later than ten business days after receipt. In practice, automate it immediately — the deadline is a ceiling, not a target.
Opt-out must propagate. If someone opts out by replying to a text, they should not receive the next campaign because a different system did not hear about it. Revocation applies across your systems, not just the one that received it.
One confirmation message is allowed. Confirming the opt-out is acceptable; using that confirmation to pitch anything is not.

The practical requirement this creates: someone or something must actually read inbound replies. A programme that only sends and never processes what comes back cannot meet the standard.

US carrier registration: 10DLC

Separate from the law, and separate from what regulators enforce. The US carriers require application-to-person traffic on standard 10-digit numbers to be registered.

Register your brand — your legal business identity, tax ID, and contact details, submitted through your messaging provider.
Register your campaign — the use case (marketing, customer care, order notifications), sample messages, and the opt-in flow. Registration is reviewed, and inaccurate descriptions get rejected.
Associate your numbers with that registration.

Registration determines your throughput and your filtering risk. Unregistered traffic on a long code is heavily filtered or blocked outright — and this failure is invisible in the worst possible way, because the message shows as accepted and simply never arrives.

Two adjacent paths: toll-free numbers have their own verification process, and short codes are a separate, more expensive, higher-trust route usually reserved for large volume. Whichever you use, the sample messages you submit should match what you actually send, because mismatches are a common cause of filtering after approval.

If you send through CampGain, SMS and RCS run on your own Twilio account, so registration is completed there and travels with your numbers.

Timing and frequency

Quiet hours. 8am to 9pm in the recipient's local time zone is the federal baseline, and several states enforce narrower windows. Schedule against the contact's time zone, not your own.
Respect the stated frequency. If sign-up promised four messages a month, six is a broken promise and an easy complaint.
Watch cumulative frequency. Your marketing campaign, your abandoned-cart flow, and your shipping notifications all arrive on the same phone. Count them together, not per programme.

Outside the US

European Union and UK

GDPR plus national ePrivacy rules govern marketing messages. Consent is the usual lawful basis, and it must be freely given, specific, informed, and unambiguous — pre-ticked boxes are explicitly invalid. Recipients have rights of access, rectification, and erasure, so you need a process for a request that arrives by email rather than as a STOP reply. You also need a data processing agreement with every vendor in the chain, including your messaging provider.

Canada

CASL requires consent and prescribes what a commercial electronic message must contain: sender identification, contact information, and a working unsubscribe mechanism that stays valid for a period after sending. Penalties are substantial and enforcement is real.

India

TRAI regulations require sender IDs and message templates to be registered on a distributed-ledger platform before they can be delivered. Unregistered content is blocked rather than merely filtered, so the registration step is a prerequisite, not a formality.

Everywhere else

Assume opt-in is required and check the specifics before your first send into a new market. The failure mode for guessing is not a warning; it is blocked traffic and a fine.

Writing texts that are compliant and still good

Constraints are not the enemy of a good SMS programme. The channel rewards brevity anyway.

Identify yourself in the first few words. An unrecognised number is deleted, not read.
One idea, one link, one action. If a message needs two calls to action, it needs to be an email.
Watch message length. A standard SMS segment is 160 characters using the GSM-7 alphabet, but a single emoji or a curly quote switches the whole message to Unicode and drops the limit to 70 — often turning one segment into three, at three times the cost. Check before you send, not on the invoice.
Use a link domain people recognise. Public shorteners are widely abused and widely filtered; a branded short domain performs better and looks less like a scam.
Include opt-out language — periodically at minimum, and always in the first message of a sequence.

Formats that consistently earn their place: back-in-stock alerts for something the recipient asked about, a genuine deadline, shipping and delivery updates, and loyalty milestones. Formats that burn a list: generic weekly promotions with no timing reason to be a text.

RCS: richer, same rules

RCS adds images, buttons, and a verified branded sender to the SMS channel. The compliance position is unchanged — consent, opt-out, and quiet hours apply exactly as they do to SMS.

What does change is the fallback. Not every recipient's device and carrier support RCS, so the message has to degrade to plain SMS. That means the SMS version needs to stand on its own, with any button destinations written into the text — a card whose only call to action is a button becomes a dead end for everyone who receives the fallback.

A pre-launch checklist

Opt-in form with complete disclosure, no pre-ticked boxes, not a condition of purchase.
Consent records stored with timestamp, source, and the disclosure text shown at the time.
10DLC brand and campaign registration approved, numbers associated (US).
STOP and equivalent keywords handled automatically, with opt-outs propagating across every system.
Inbound replies monitored by a person or a process, not ignored.
Quiet hours enforced by recipient time zone.
Frequency counted across every programme that texts the same phone.
Business name and opt-out language in the message.
Data processing agreements in place for non-US markets.
A first send to a small internal group, checked on a real phone.

How CampGain fits

CampGain sends SMS and RCS through your own Twilio connection, so your registration, your numbers, and your sender reputation stay yours.

Per-contact delivery results with the provider's stated reason on failures — which is how you spot filtering that would otherwise look like silence.
Ready-made SMS and RCS presets written to stay within a single GSM-7 segment, so a template does not quietly triple your cost.
Segments and suppression built from combined contact and Shopify data, so opted-out contacts stay out of every campaign rather than only the one they replied to.
Automatic RCS-to-SMS fallback, with the card's button links carried into the fallback text.
Test sends to a handful of contacts before a campaign reaches the segment.

SMS works best as one step in a sequence rather than a channel on its own — our multi-channel campaign guide covers when a text earns its cost, and the WhatsApp Business API guide covers the channel with its own, quite different, rulebook. See how channels connect.

The short version

Get explicit written consent and keep the evidence. Register with the carriers before your first US send. Honour opt-outs immediately, however they are worded, and make sure the opt-out reaches every system. Send inside the recipient's local quiet hours. Count your total message frequency, not your per-campaign frequency.

Do that, and SMS becomes what it should be: a channel people are glad to hear from, used rarely enough that they still open it.

Email Personalization at Scale: What Actually Works

Arpit Anand — Mon, 18 May 2026 00:00:00 GMT

Personalization has a credibility problem, and it is self-inflicted. A decade of "Hi {{first_name}}" taught shoppers that a personalized email is one that got their name right and nothing else.

The version that actually moves revenue is less glamorous and more mechanical: know which product someone was looking at, know whether they have bought before, and let those two facts change what the message says. This guide covers how to build that in layers — starting with things you can ship this week — and where AI is genuinely useful rather than decorative.

Personalization is a data problem before it is a content problem

Every personalization technique reduces to the same question: what do you know about this person that would change what you say to them?

Most stores have more than they use:

Identity — name, location, language, how they signed up.
Purchase history — what, when, how often, how much, which categories.
Behaviour — products viewed, carts started and abandoned, checkouts begun.
Engagement — which channels they respond to and which they ignore.

The constraint is rarely data volume. It is data quality — specifically duplicates. If the same person exists twice under two email addresses or two phone formats, personalization actively works against you: they get two different "personal" messages, each based on half a history. Deduplicate before you personalize, or the effort compounds into embarrassment.

The three levels, cheapest first

Level 1: merge variables done properly

Inserting known values into a template. Names, order numbers, coupon codes, the specific product left in a cart, the tracking link.

This is table stakes, but two details separate the ones that work from the ones that go wrong:

Every variable needs a fallback. "Hi ," is the classic failure, and it lands specifically on the contacts you know least about. Decide what an empty field renders as — "Hi there" — before you send.
Variables in links need extra care. A merge field in a URL that does not resolve produces a broken link, which is worse than an ugly greeting. Test with a contact who has gaps, not only with your own complete record.

Level 2: segment-level relevance

The most under-used level, and the one with the best effort-to-return ratio. Instead of one campaign with dynamic blocks, send two or three campaigns to different segments.

Concretely: your new-arrivals email goes to everyone. Split it. First-time buyers get a version that leads with your bestseller and mentions your returns policy. Repeat customers get a version that leads with what is new in the category they already buy from. Same effort as a moderately complex dynamic template, and far easier to debug when something looks wrong.

Start with three splits: never purchased, bought once, bought more than once. That single distinction changes the right message more than most demographic data ever will. Our segmentation guide covers building the audiences.

Level 3: catalog-driven content

Product blocks pulled from your live catalog rather than hand-built into the template. The specific benefit is accuracy: names, prices, images, and availability reflect the state of the catalog at send time, not whenever the template was written.

Anyone who has sent a campaign featuring a product that sold out the previous day understands why this matters.

Good uses of catalog-driven blocks:

Cart contents in recovery messages — the single highest-value personalization in ecommerce.
Recently viewed products in a browse follow-up.
Category-matched new arrivals for repeat customers.
Complementary items in post-purchase, based on what they actually bought.

Timing is personalization too

Everyone treats personalization as a content question. Half of it is a timing question, and timing is often easier to get right.

Trigger-based timing — a message an hour after a cart is abandoned is personalized by definition. It arrives because of something that person did.
Replenishment timing — for consumables, work out the median gap between repeat orders per category and send a few days before it. This routinely outperforms much more sophisticated content personalization.
Lifecycle timing — a message on day 30 after a first order is a different message from day 30 after the fifth.
Time zone — sending at the recipient's local time rather than yours. Necessary on SMS, where it is also a compliance requirement.

Where AI genuinely helps

Be specific about this, because the category is full of claims that do not survive contact with a real account.

It is good at production

The bottleneck in segment-level personalization is writing four versions of an email instead of one. That is exactly what AI drafting is good for: give it your brand voice, the offer, the audience, and the goal, and get a first draft per segment in minutes. You still edit — but editing four drafts is a different job from writing four emails.

It is similarly good at rewriting one section in a different tone, generating subject line variants to test, and producing the short-form versions a message needs for SMS or WhatsApp.

It is not good at deciding what to personalize

The judgement calls — which segments deserve their own message, what objection is blocking a purchase, whether this audience should get a discount — come from knowing your customers and reading your data. A model with no access to your margins and no memory of last quarter's campaign cannot make them.

Two rules for AI-written marketing

Everything gets a human read before it sends. The failure mode is not gibberish; it is plausible copy that misstates your returns policy or promises a discount you did not authorise.
Give it your brand context, not just a prompt. Tone, product vocabulary, what you never say. Generic input produces generic output, and generic output is exactly what personalization is meant to avoid.

Personalization that backfires

Knowing too much, too visibly. "We noticed you looked at this three times" is accurate and unsettling. "Still thinking about this one?" says the same thing without the surveillance framing.
Broken variables. A visible {{placeholder}} destroys more trust than personalization built.
Personalizing on stale data. Recommending something they bought a month ago reads as not paying attention.
Personalizing the wrapper, not the content. The name is right and the products are the same ones everyone got. Shoppers notice.
Ignoring per-channel consent. A personalized WhatsApp message to someone who only ever agreed to email is a compliance problem regardless of how relevant it is.

Privacy-first is the only durable version

Third-party tracking has been steadily restricted, and the personalization that survives is built on data customers gave you directly — orders, on-site behaviour on your own store, and stated preferences.

That is a better foundation anyway. First-party data is more accurate, it does not disappear when a browser policy changes, and it is easier to explain to a customer who asks. Two practical habits: collect preferences explicitly where it helps (which categories, how often, which channel), and keep contact records clean and deduplicated so what you know about someone is in one place.

What CampGain provides

Merge variables resolved per recipient across email, SMS, RCS, and WhatsApp — contact fields, campaign coupon values, and product details, substituted individually for every message rather than at the template level.
A synced Shopify catalog. Products assigned to a campaign step resolve to current names, prices, and links at send time.
Segments that combine sources — your own conditions plus Shopify data in one audience, with duplicate people collapsed on email or phone so personalization draws on one complete record.
An AI email writer. Store your brand kit once, then generate a draft from the goal, tone, occasion, and length, review it section by section, rewrite individual sections, and save the approved result as a reusable template.
Preset template libraries for email, WhatsApp, and SMS/RCS — the WhatsApp presets written against Meta's category rules, and the SMS presets kept to a single segment so a personalized message does not silently become three.
Test sends to a handful of contacts, which is how you catch an unresolved variable before the whole segment does.

See how templates and campaigns work together.

Where to start

Pick your highest-volume campaign. Split it three ways — never purchased, bought once, bought more than once — and write a genuinely different opening paragraph and product selection for each. Add fallbacks to every merge variable. Send a test to a contact with an empty first name.

Compare revenue per recipient against the unsegmented version. That result, from your own list, is worth more than any benchmark — and it is the argument you will use for everything else in this guide.