Ecommerce email marketing: the complete guide

Write down the job first, because the answer changes everything downstream. In most stores email has three jobs and they need different treatment.

The first job is turning a first order into a second one. This is where the money is for almost every store, and it is done with automated sequences that run in the background rather than with campaigns. If you only ever do one thing with email, do this.

The second job is bringing people back who did not buy. Someone browsed, or filled a cart, or signed up and never ordered. Email reaches them at a cost close to nothing, which is why it usually recovers more than any paid retargeting you could run at the same value.

The third job is telling people who already like you what is new. New products, restocks, the seasonal range, the thing you make that they have not tried. This is the campaign side, and it is the part most stores start with and most stores get wrong, because it turns into a weekly discount when nobody has planned what else to say.

The reorder cycle of your products decides the balance between the three. A store selling something consumed in six weeks lives on replenishment timing and can email often, because there is always a reason. A store selling something bought once every four years lives on referrals, accessories and care, and should email far less. Work out your typical gap between first and second order from your own order data before you set any cadence.

Also decide what email is not for. It is not the place to run your entire discount strategy, it is not a substitute for a store that answers questions, and it will not fix a product people do not want to buy again. Being clear about that keeps you from asking the channel to carry weight it cannot hold.

List quality beats list size in every measure that matters, so collect addresses from people with buying intent and stop collecting them from anyone else. A giveaway that attracts entrants brings you a list of people who wanted the prize, and they will drag down your inbox placement for years.

The biggest single source is checkout. Everyone who orders should be on the list, with proper consent captured at the point of purchase using the mechanism your platform provides. These people have paid you money. They are the most valuable addresses you will ever hold and they cost nothing to acquire.

The second source is the signup on the site. What it offers matters more than where it sits. An immediate discount works and has a cost: it trains people to expect one and it attracts a slice of people who only wanted the code. Alternatives worth testing include early access, a genuinely useful guide for your category, a fit or sizing tool, or notification when a specific product is back. Which one is right depends on whether your margin can carry a first order discount.

Timing and targeting on the form matter too. Showing it the second someone arrives is the version most likely to be dismissed. Showing it on exit, after a scroll, or on the second visit gets fewer impressions and better addresses. Ask for as little as possible. Every extra field costs signups, and you can collect preferences later from people who are already engaged.

Back in stock and waitlist signups are the most underrated source in the whole channel. Someone asking to be told when a specific product returns has declared exactly what they want, and that flow converts well because the message is genuinely wanted.

One rule with no exceptions: never buy, rent, scrape or upload a list of people who did not opt in with you. It damages deliverability for your real customers, and in most places it is against the law.

An email nobody receives cannot be measured, tested or improved, so deliverability comes before creative. Inbox providers decide placement mostly on how recipients react to your mail, with authentication as the entry ticket.

Authenticate the sending domain properly. Publish the records that let providers verify mail claiming to come from you, and use a subdomain of your real domain to send marketing mail rather than a generic address at a free provider. Sending from your own domain is what lets your reputation accrue to you.

Warm up gradually if the domain is new or has not sent in a long time. A domain with no history that suddenly sends to fifty thousand people looks exactly like a domain that has been compromised. Start with your most engaged recipients and increase volume over a few weeks.

Then protect the reputation you build. Send to people who open. Providers watch engagement, so a list where half the recipients never open teaches them your mail is unwanted, and that judgement applies to the half who would have bought. The fix is unglamorous: suppress addresses that have not engaged in a long time, define that window based on your reorder cycle rather than a generic number, and try a re-engagement sequence before you suppress.

Make unsubscribing easy and instant. Hiding the link produces spam complaints instead of unsubscribes, and a complaint costs far more than a lost address. Offer a preference option beside it so people can reduce frequency rather than leave entirely.

Watch the numbers that predict trouble rather than reacting after the fact. Bounce rate rising, complaint rate rising, open rate falling on your engaged segment, and delivery differing sharply between providers are all early signals. If placement drops, cut volume back to the most engaged part of the list and rebuild rather than sending harder.

Send different mail to different people using data you already hold, because relevance is the only thing that lets you send more often without wearing the list out. You do not need a survey to do this. Your order and browsing history contains almost everything useful.

The first split is by purchase history. Never bought, bought once, bought more than once, and bought recently at high value. These four groups need genuinely different messages, and treating them as one list means every email is slightly wrong for almost everyone.

The second split is by category. Someone who has only ever bought one line from you should be hearing about that line, its accessories and its replacements. If your catalogue spans several categories that do not overlap in audience, sending everything to everybody is the fastest way to train people to ignore you.

The third split is by engagement. Recent openers can take more frequency. People who have not opened in months should get fewer messages, not more, and eventually a re-engagement attempt. Most stores do this backwards and increase pressure on the least interested people.

The fourth split is timing based, and it is the most valuable for consumables. Calculate your typical gap between orders per product category, then build segments around where each customer sits in that cycle. Someone at week five of a six week cycle is a different audience from someone at week one.

Keep the number of segments small enough to maintain. A dozen well defined segments used consistently beats fifty that nobody remembers the definitions for. Write the definition of each one down somewhere the whole team can read, including exactly which customers fall in and out, and review them when the catalogue changes.

Automated flows earn most of the revenue per message; campaigns earn most of the total volume. Build the flows first, because once they are live they keep working whether or not anyone touches them that month.

A flow is triggered by something a person did, so it arrives at the moment it is relevant. Someone signed up, abandoned a cart, placed an order, received a delivery, or stopped buying. Because the timing is tied to their behaviour rather than your calendar, the message is wanted, and that shows in every metric.

The core set for a store is small. A welcome sequence for new subscribers. A cart and checkout recovery sequence. A browse follow up for people who looked at a product and left. A post purchase sequence covering the period from order to arrival and beyond. A replenishment reminder for anything consumable. A winback for customers who have gone quiet. And a back in stock notification if you ever sell out of things people want.

Campaigns are the scheduled sends: the new arrival, the seasonal edit, the story about how something is made, the guide that helps people use what they bought. They exist to keep the relationship warm between purchases and to give people a reason to come back that is not a discount.

The common mistake is to run this the other way round. A store with no flows sends a weekly campaign to everybody, sees mediocre results, and sends harder. Building the flows first usually raises revenue from the same list without increasing how often anyone hears from you, which is the cheapest improvement available in this channel.

Once the flows are live, treat them as products rather than projects. They need reviewing when the catalogue changes, when shipping times change and when the offer changes, or they quietly start telling customers things that are no longer true.

Decide what you are sending for a whole quarter in one sitting, because campaigns written the day before are the ones that default to a discount. A calendar removes the Friday afternoon question of what to send, which is where most bad email comes from.

Start with the fixed points. Product launches, restocks, seasonal ranges, the periods when your category genuinely gets busy, and any dates that matter to your customers rather than to marketers generally. Put those on the calendar first, then fill the space between them.

Give the filler sends a purpose. A rotation of a few recurring formats works well: how to use or care for what you sell, a story about how something is made or sourced, a customer question answered properly, a set of products chosen around a situation rather than a category, and the occasional plain note from the founder when there is something worth saying. Name each format and reuse it, since a format you can repeat is a format you will actually send.

Balance the calendar so promotional sends are the minority. If you count your quarter and two thirds of it is discount led, the list will learn to wait, and your full price sales will fall to pay for it.

Decide the audience for each campaign at planning time, not at send time. Most campaigns should not go to everybody. Writing the segment into the calendar entry forces the question of who this is genuinely for, and it prevents the habit of sending everything to the whole list because it is easier.

Leave slack. A calendar with every week booked has no room for the restock you did not expect or the thing that suddenly sells out, and those unplanned sends are often the best ones. Plan around three quarters of the slots and keep the rest free.

One email, one job. The message that tries to announce a launch, clear old stock, ask for a review and share a story will do none of them, because a reader scanning on a phone gets one decision from you and picks the first thing they see or nothing at all.

The subject line and the preview text work as a pair, so write them together. The subject says what this is, the preview adds the reason to care, and neither should try to trick anyone into opening. Curiosity gaps get opened once and resented afterwards, and resentment shows up in your complaint rate. Being specific about what is inside outperforms being clever in nearly every store, because your list opens mail from you for a reason.

Write the first line as though the reader is standing in front of you. The opening should land the point immediately, since a phone shows very little before the fold. Then one image or a small set, one paragraph of context, one button. Anything else goes below, where the people who want more will find it.

Use their words, not your internal ones. Product names customers do not use, seasonal collection names that mean nothing outside the office, and internal shorthand all cost clicks. If a customer would describe the product differently, describe it their way and put your name for it second.

Write the call to action as an action rather than a category. See the sizes, check what fits, read how it is made. It sets the expectation of what happens next, and matching expectation to the landing page is most of what makes a click convert.

Finally, send it to yourself and read it on a phone before it goes. Almost every fixable problem in an email is visible in that one check, and almost nobody does it.

Most of your list reads on a phone, often on a poor connection, and increasingly in dark mode, so design for that first and check the desktop version afterwards. Emails designed on a large screen tend to fail in all three conditions at once.

Use a single column. Multi column layouts either collapse unpredictably or force people to pinch and scroll sideways, and either way the second column is rarely read. One column, generous spacing, and a clear order from top to bottom will serve every device you care about.

Make the text real text rather than part of an image. Images can fail to load, get blocked, or take too long on a weak connection, and an email whose entire message is inside one picture becomes a blank rectangle. Real text also lets people increase the size, which matters more than most teams assume.

Size the tap targets for a thumb. Buttons need to be large enough to hit without care, with space around them so a mis tap does not open the wrong thing. Small text links stacked closely together are the most common cause of accidental unsubscribes.

Check dark mode explicitly. Backgrounds invert, logos with white backgrounds turn into bright rectangles, and dark text on a dark background disappears entirely. Test with the actual email, not a preview render, on at least one phone in each mode.

Keep the whole thing light. Heavy images and long templates get clipped by some clients, which hides the unsubscribe link and the footer, and a clipped email looks broken. Compress images, keep the template lean, and put the important content near the top where clipping cannot reach it.

Accessibility is the same work by another name. Alt text on images, sufficient contrast, a readable text size and a sensible reading order help everyone reading in a hurry.

Every discounted send teaches your list something, so decide what you want it to learn. A store that sends a code every week has taught its best customers to never pay full price, and it will pay for that lesson for years.

Use discounts where they change a decision. A first order incentive can be worth it when your repeat rate is strong enough to earn the margin back. A clearance offer on genuinely old stock is honest work. A recovery message where price is the actual objection is reasonable. A discount attached to a routine campaign because nobody had a better idea is not.

Before discounting anything, work out what it costs. Take the margin on a typical order, subtract the discount, and compare that with the margin on the orders that would have happened without it. The uncomfortable part of that sum is that most redemptions come from people who were going to buy anyway, which means the discount is a cost applied to existing revenue.

When you do discount, put a boundary on it. A specific product range, a specific segment, a real end date, a minimum order, or a reason attached to it. Boundaries make the offer feel like an event rather than the normal price, and they let you stop without it feeling like a withdrawal.

Build the alternatives, because they are what let you stop relying on price. Free shipping over a threshold set from your own order data. A gift with purchase that costs less than the equivalent discount. Early access for existing customers, which costs nothing. A bundle that raises the order value while giving the customer a reason to feel well treated.

If your list only responds to discounts, that is information rather than a strategy. It usually means the list was built with a discount and has never been given another reason to open, and the fix is a few months of email worth reading rather than a bigger code.

Open rate is the least reliable number in the channel and the one most often reported, so demote it. Privacy features cause opens to be recorded for messages nobody read, which inflates the figure unevenly depending on which providers your list uses. Watch it as a trend on a stable segment, never as a target.

The number that answers the real question is revenue per recipient: total revenue attributed to a send divided by how many people it went to. It compares a small targeted send with a large one fairly, and it stops the reflex of judging success by total revenue, which always rewards sending to everybody.

Alongside it, track click to open rate as your content quality signal, unsubscribe and complaint rate as your fatigue signals, and list growth net of removals as your health signal. That last one is more useful than raw growth, since a list adding people at the front and losing them at the back is standing still.

Measure flows separately from campaigns and each flow on its own. Flows are compared against themselves over time, not against campaigns. A welcome sequence and a winback sequence have completely different jobs and completely different numbers, and averaging them tells you nothing.

Be careful with attribution. Email platforms typically credit any order from someone who opened or clicked within a set window, which will overlap with what your ads platforms and analytics also claim. Pick one window, keep it, and look at the trend rather than arguing about the absolute number. For a truer read, compare total store revenue in periods with different email activity, or hold back a portion of a segment from a send and compare.

Finally, tie the channel back to the business measure it exists to move: the share of customers who buy again and how long they take. If flows are working, that gap shortens and that share rises, and those two numbers are harder to flatter than anything in a campaign report.