You build an ecommerce email marketing strategy with seven decisions made in the right order: goal and guiding metric, contact collection, automated flows, campaign plan, segments, deliverability, measurement. The tool matters less than it seems. Klaviyo, Mailchimp, Omnisend and Brevo do roughly the same things, and two stores on the same platform often get opposite results. For each decision you'll find what to choose, the most common mistake and an example from a made-up store. For measurement, there's a real case too.
How do you build an email marketing strategy for an ecommerce store?
By answering seven questions, in this order.
| # | Decision | The question to ask | The typical mistake |
|---|---|---|---|
| 1 | Goal and guiding metric | What should change in the store's numbers because of email? | Looking only at opens, or only at attributed revenue |
| 2 | Contact collection | Why would a visitor give me their email today? | A generic "subscribe to our newsletter" pop-up |
| 3 | Automated flows | Which moments of the customer journey are still uncovered? | Ten half-finished flows instead of four done well |
| 4 | Campaign plan | What do I send, to whom and for what purpose over the next four weeks? | Promotions only, sent whenever someone remembers |
| 5 | Segments | Who should get this email, and who shouldn't? | Everything to everyone |
| 6 | Deliverability | Do my emails land in the inbox? | Finding out when sales drop |
| 7 | Measurement | How much revenue exists because I sent those emails? | Adding up the revenue attributed to email and the revenue attributed to ads |
The order has a logic: without a goal you don't know what to measure, and without contacts your flows never start.
1. Goal and guiding metric: bring back people who have already bought
Advertising often brings in the first order. Email works mostly on what comes after: the second and third orders, which need no new ad spend. That's why it pays to read email through MER (marketing efficiency ratio): total revenue divided by total ad spend. If the share of revenue from returning customers grows, MER holds up even when ads cost more.
An example with made-up numbers. A tea store spends $8,000 a month on ads and brings in $28,000 in revenue: a MER of 3.5. In November ads cost 20% more: spend of $9,600, revenue from new customers unchanged. If repeat purchases don't grow, MER falls to 2.9. If post-purchase flows and newsletters bring in $5,600 more in repeat orders, MER stays at 3.5.
Below the goal sit the intermediate metrics:
| Metric | What it tells you |
|---|---|
| Signup rate | How many visitors leave their contact details |
| Open rate | Whether the subject line and sender catch attention, with serious caveats |
| Click rate | Whether the content brings people back to the site |
| Order rate | How many recipients buy |
| Customer lifetime value | What a customer is worth over the months |
| Attributed revenue | How much the platform credits to email (see point 7) |
To judge a single email, pick clicks. Opens have been unreliable since 2021, when Apple added a protection to Mail that hides from senders when a message is opened. A click brings the person back to your site, where the actions that trigger your automated flows happen.
The typical mistake is judging every email by the same yardstick. An email that explains how to make iced tea should be judged by clicks and product views, not by the orders placed that day.
2. Contact collection: no subscribers, no flows
Every automated flow needs an action from the person and an address to write to: cart recovery only works if you know who abandoned the cart. The main sources:
- pop-ups and forms on your site, with a concrete promise instead of "subscribe to our newsletter" (see the signup pop-up guide);
- multi-step forms: email first, then a question you'll actually use (clothing size, skin type) or the phone number, with its own separate consent;
- the checkout, with a clear marketing consent checkbox that isn't pre-ticked (if you sell in the EU, the GDPR is explicit that a pre-ticked box is not consent);
- lead ads connected to your email platform;
- dedicated signup pages for waitlists and launches.
The typical mistake is making a discount the only reason to sign up: you attract people who want the code and then disappear. A welcome code makes sense if it's valid on the first purchase only and isn't your only promise.
Example. A dog supply store with 25,000 visits a month asks for the dog's size in its pop-up, in exchange for tailored advice and 10% off the first order. It collects 400 subscribers a month and already knows who should see leashes for large dogs and who should see coats for small ones.
3. Automated flows: which to build first
Flows and campaigns are two different jobs. A flow responds to something the person did and runs on its own, day and night. A campaign is something you decide: date, audience, message. In Klaviyo's 2026 benchmarks, flows generate nearly 41% of attributed email revenue with just 5.3% of sends, and they have an average click rate of 5.58% against 1.69% for campaigns.
To decide which flows you need, line up the moments of the customer journey:
| Moment | Flow | What triggers it | Goal |
|---|---|---|---|
| Signs up | Welcome | New signup | First order, getting to know the store |
| Views a product and leaves | Browse abandonment | Product viewed with no cart | Bring them back to the product or similar ones |
| Leaves the cart | Abandoned cart | Product added with no order | Complete the order |
| Stops at payment | Abandoned checkout | Checkout started with no order | Remove the doubt that stopped them |
| Buys | Post-purchase | Order placed | Product use, second order |
| Receives the package | Review request | Order delivered | Collect social proof |
| Disappears for months | Win-back | No order for a long period | Bring them back, or stop emailing them |
The recommended order: welcome and cart and checkout recovery first, because they catch people who have just shown interest. According to the average of 50 studies compiled by the Baymard Institute, 70.22% of online shopping carts are abandoned. For recovery, use the timing from the abandoned cart recovery guide: one hour, 24 hours, then between 48 and 72 hours (the reasoning is in when to send abandoned cart emails). The discount never goes in the first email, and codes are valid on the first purchase only, as explained in discounts in abandoned cart emails. After that come post-purchase, browse abandonment, reviews and win-back.
The typical mistake is switching on every template the platform offers and never looking at them again: ten generic flows written in an afternoon earn less than four written for your customers and reviewed every month.
Example. A handmade ceramics store builds the welcome series and the abandoned cart flow in its first month. In the second it adds post-purchase: how to care for the pieces as soon as the package arrives and, after thirty days, a piece that goes with the one they bought.
Once the basic flows work, you grow in two directions: new moments (back-in-stock alerts, reorder reminders) and extra emails or branches where clicks stay high.
4. Campaign plan: a calendar, not Monday-morning inspiration
Campaigns work when they start from a retail calendar (sales, Black Friday, collection launches, key dates in your industry) turned into an editorial plan. For every send, the plan states:
- who it goes to (the segment, never "everyone");
- what kind of email it is: promotional, educational, informational or personal;
- what goal it has: a click, a product view, an order;
- two subject line versions to test against each other.
Frequency depends on your audience and on how much you have to say: raise it step by step, and only for people who respond. For peak periods, there's the four-week Black Friday email plan.
The typical mistake is sending nothing but promotions. Your list learns to wait for the discount, margins shrink, and anyone who doesn't want that offer stops opening.
Example. A hiking store plans October: the winter boot guide for people who viewed footwear, the jacket launch for engaged subscribers, early access for customers with at least two orders.
5. Segments: deciding who doesn't get the email too
A list is static: it holds whoever signed up. A segment is dynamic: it automatically gathers everyone who meets certain rules, such as "bought in the last 60 days" or "clicked on a category". Five segments are enough to start:
- recently engaged subscribers and customers (clicks or orders in recent months);
- customers with two or more orders;
- customers with a single order;
- subscribers who have never bought;
- contacts inactive for many months, to exclude from most sends.
Interest-based segments come next, by category bought or viewed.
The typical mistake, the opposite of "everything to everyone", is over-segmenting a small list. Groups of eighty people don't produce readable data, and they multiply the work.
Example. The tea store sends its new Japanese green teas to people who have already bought green tea, and its evening herbal blends to people who only buy herbal tea.
A segment, though, treats everyone who meets the same rule in the same way. The next step is in the section on artificial intelligence.
6. Deliverability: your emails have to reach the inbox
Everything above is worth nothing if your emails end up in spam. Google asks senders to keep their spam complaint rate below 0.10% and never to reach 0.30%. The basics:
- a domain authenticated with SPF, DKIM and DMARC;
- one-click unsubscribe, visible and with no hoops to jump through (in the US, CAN-SPAM also requires you to honor opt-outs within 10 business days);
- inactive contacts excluded from regular sends;
- volumes that grow gradually, not overnight;
- subject lines that deliver what they promise.
The detail is in the guide on why emails go to spam.
The typical mistake is emailing the whole list to "hit the numbers". People who aren't interested mark you as spam, and your domain reputation drops for the people who would have bought too.
Example. A store with 30,000 contacts finds 11,000 who haven't clicked or bought in over a year. It removes them from newsletters, sends them three win-back emails over a month and keeps those who respond.
7. Measurement: attributed revenue is not caused revenue
The platform credits an email with every order placed within a set window after an open or a click. Some of those orders would have come in anyway: someone who buys the same coffee every month would have bought it without a reminder. Your ad platform can claim the same order too: added together, attributed revenue often exceeds your real revenue.
Three habits for measuring honestly:
- Look at the total. Overall revenue, share from returning customers, customer lifetime value, month over month.
- Follow the email's path. Sent, clicked, ordered: if clicks are good and orders are few, the problem comes after the click (page, price, offer); if clicks are few, it's in the content or the audience.
- Keep a control group. A small random group that doesn't receive a flow for a few weeks, compared with the people who do.
Example. The dog supply store randomly excludes one subscriber in ten from the welcome flow for six weeks. Among people who get the welcome series, 9% buy within 30 days; among those excluded, 6%. Those 3 points are the work of the flow. The rest would have come in anyway, even though the platform credits it to email.
The real case: the same months of the previous year
When you change a flow, the most common comparison is with the year before. A footwear store we work with replaced its original cart recovery emails with emails written for the individual customer, starting in May 2026. Revenue from the recovery flows, from the store's Klaviyo account (months rounded, totals exact):
| Month | 2025, original emails | 2026, new emails |
|---|---|---|
| May | €7,300 | €11,500 |
| June | €9,800 | €12,800 |
| July | €7,800 | €9,200 |
| August | €7,400 | €12,500 |
| September | €11,800 | €17,600 |
| Total | €44,092 | €63,555 |
The total grows by 44%, but it says less than it seems. Comparing the same months removes the seasonal effect, not the effect of traffic, prices and promotions: with more visitors, recovery would have brought in more even with the old emails.
The sturdier figure is the share. The recovery flows go from 11.7% to 14.7% of the revenue the platform attributes to the store's emails. The rest of the store's emails had the same traffic and the same prices: three extra points mean recovery grew faster than the rest of the channel.
These are still attributed numbers, and they include sales that would have happened anyway. Real proof takes a test over the same period: carts split at random into two groups, one with the original emails and one with the new ones, and orders per cart compared. Same weeks, same conditions, one difference.
Where artificial intelligence fits in (and where it doesn't)
Artificial intelligence doesn't change the seven decisions. It removes manual work in some places and leaves the store the ones that need judgment.
| AI can help | Stays with the store |
|---|---|
| Writing copy for the individual person, based on what they bought, viewed or left in the cart | The strategy: goal, guiding metric, which flows and which campaigns |
| Reading which angles drive orders (novelty, social proof, discount) and how much data backs them | The offer: how big a discount, to whom, when |
| Turning those readings into a brief for ads and newsletters | Consent: what you ask for, how you collect it, what you promise |
| Measuring honestly: tests over the same period, small samples flagged | Deliverability: authenticated domain, clean list, frequency |
The most concrete point is the first. Where a segment sends the same email to everyone, a language model can write a different one for each customer inside the same flow. Some cart recovery tools do this, AchillesAI among them; what works and what doesn't is in what AI really does in email marketing.
The right-hand column can't be delegated. Good copy with the wrong discount teaches people to wait for the code, a pop-up that promises one thing and delivers another collects contacts who don't buy, and no copy arrives if the domain isn't authenticated or the list is full of inactive contacts (why emails go to spam).
Where to start this week
- Write down in one line the goal of your emails for the next six months and the metric that measures it.
- Check that welcome, abandoned cart and abandoned checkout are live and written for your customers.
- Prepare next month's campaign plan, with a segment and a goal for every send.
- Remove from regular sends anyone who hasn't clicked or bought in a year.
You can switch tools in a month. These decisions, made well, stay with you.
