Building Ecommerce Email and SMS Flows That Actually Get the Second Purchase

The problem is the second order

Your store sells. What is not happening is the second purchase, so every extra dollar of ad spend buys another first-timer who disappears. The detail that fixes it is never in the trigger — it is in the exclusions, the delays, and the stopping rules.

The flows, and how they are actually built

Abandoned cart

Trigger: cart created, checkout not completed. Trivial. The exclusions decide whether it works.

• Already purchased — evaluated at send time, not entry time. Someone who abandons at 9pm and buys at 11pm must not get the 9am reminder.
• Already in another flow — set a priority order and suppress the loser, or a new subscriber gets welcome and cart mail the same morning.
• Repeat abandoner — some people use the cart as a wishlist. If an address has entered several times without converting, stop entering it.

Do not open with a discount. Assume friction, not price: a sizing doubt, a failed payment step. Message one restates the cart, answers the objections you already know from your support inbox, and gives one link back. Lead with a code and you teach the people most likely to buy anyway that abandoning is how you get one.

The delay between steps is a judgment about intent decay, not a schedule. The first is short enough that context is still in their head, catching the interrupted purchase. The second waits until a considered no-buy has had room to happen, and changes the argument rather than repeating it. Only a final step has any business carrying an incentive, and the fifth never converts someone the first four did not.

Welcome

This is where you earn the right to email later. Message one establishes three things: who you are — not the founding story, but what makes the product different from the alternative they nearly bought; what you sell, in real categories so they can self-select; and what happens next, meaning how often they hear from you and about what. Everyone skips the third, and it is the one that prevents complaints six weeks later.

Back-in-stock

The subscribe control belongs on the product page against the specific variant; someone waiting on one size does not want to hear a different size returned. The race condition is what burns you: restock three units, notify everyone, and almost all of them hit a sold-out page seconds later — teaching your highest-intent contacts that your notifications waste their time. So cap or stagger: notify a batch proportional to the units you have, wait, release the next only if inventory remains. Staggering solves scarcity; if the restock covers everyone, send it at once.

Browse abandonment and post-purchase

Browse abandonment sits below cart in intent: fewer messages, softer framing, suppressed while a cart flow runs.

Post-purchase is where repeat rate actually moves. Cart recovery rescues a sale you nearly had; this creates one you did not. Three phases:

• Transactional reassurance — confirmed, shipped, out for delivery. These get read more than anything else you send, which is why they should not carry promotions.
• Usage guidance, timed to arrival rather than purchase: how to use it, how to care for it, the mistake most people make in week one. It cuts returns and tickets, and almost nobody builds it.
• Replenishment or cross-sell, timed to the product's real consumption cycle. If a bottle lasts roughly six weeks, the reminder goes before it runs out, not at thirty days because thirty is round. Durable goods have no cycle, so the equivalent is the complement.

Once the flows are mapped, you need a platform that sees store events and reaches both channels in one place. That is where something like Omnisend fits: an email and SMS platform built around ecommerce store integrations and automation workflows, so the order data driving your conditions lives where the sends are configured.

How far to split segments

Four axes carry most of the value: purchased vs never purchased, recency, product category where categories imply different needs, and channel consent. Nothing you say to a customer should read like what you say to a stranger.

Past that it turns on you. Each send gets too few recipients to learn from — you cannot read a result off a segment where a handful of people swing the number. Variants multiply past what anyone maintains, and most soon reference a promotion that ended. Worst is overlap: narrow segments built independently intersect, and one person gets three emails the same morning. The stopping rule — if you cannot name a different action you would take for a segment, it is not a segment, it is a filter.

Sending domain and authentication

Send from a subdomain of your real domain, not a free mailbox address and not your root domain, so a bad campaign does not drag down your invoices and password resets.

Publish all three records. SPF declares which servers may send for the domain, DKIM signs messages so forgery is detectable, and DMARC ties both to the visible From address and tells receivers what to do on failure. DMARC is the one people skip, because the first two make mail flow and DMARC only makes it trusted. In 2024 Gmail and Yahoo tightened their bulk-sender requirements: authentication, working one-click unsubscribe, and low spam-complaint rates became expectations rather than best practices.

Failure here is invisible. Your own message lands in your own inbox looking perfect, because mail from your domain to your own account is treated differently. Meanwhile a growing share of the list is filtered or rejected at the gateway — sends look delivered, engagement just declines. Read per-provider complaint reporting and your DMARC reports, not your inbox.

Warm a new sending domain gradually, starting with your most engaged contacts, because a domain with no history that suddenly mails an entire list looks like a compromised account. And never load a bought or scraped list — those carry spam traps and dead addresses, and the damage attaches to the domain permanently. You do not repair it; you abandon it.

When SMS is right, and when it is a mistake

SMS earns its place when the message is time-critical and short: order shipped or out for delivery; back in stock for a specific item they explicitly asked about; a genuinely limited window with a real deadline. It is a mistake for anything long, anything they did not clearly consent to receiving by text, anything landing at 2am in their timezone, routine newsletters, and — the test that catches most bad sends — anything email would have done just as well.

Consent is strictly separate. Subscribing to your emails is not agreement to texts; SMS consent is collected on its own terms and regulated more tightly. Carriers and regulators treat abuse far more harshly, and unlike email there is no gradual reputation slide to warn you — access can simply be cut. SMS also costs real money per message, which enforces its own discipline.

What automation will not fix

A product nobody wants to buy twice. If repeat rate is low because the item did not match the page, a replenishment flow just reminds people of a disappointment on schedule. A bad post-purchase experience is the same story — slow shipping, painful returns, or support that does not answer will out-argue any email you write.

A list built from a giveaway that drew prize hunters has no purchase intent, dragging engagement down and harming deliverability for contacts who genuinely wanted to hear from you. And migration carries addresses and orders across but usually not engagement history in a form the new platform can act on, so your recency segments reset and you fly blind for a while.

Templates, sending infrastructure, and deliverability tooling are the platform's job. Offer, timing, and copy are still yours. The tool decides whether the message arrives. You decide whether it deserved to.

Where to start

Authenticate the sending domain, then build post-purchase, abandoned cart, welcome, and back-in-stock, in that order. Authentication first because everything downstream depends on arrival — and post-purchase before cart recovery, because almost every store does it the other way round, then wonders why the second purchase never comes.