Post-purchase flow

What the second order is worth

More than almost anything else you can buy. They just paid you, the address is confirmed, attention is at its peak, and the acquisition cost was already spent on the first order. Most stores leave it uncollected in one of two ways: silence, which is a receipt and then nothing until the next blast, or greed, which is a discount code the day after somebody happily paid full price.

The one rule: one pattern, one moment, one offer. "Just bought" is not a segment. It is at least four different futures, and the order data says which one this customer is in.

How to think about it

Diagnose, then time, then offer, then write. Skipping to the writing produces the same artefact every time: a day-thirty upsell sent to everybody, which is the blast this board exists to replace.

• Four patterns, and they are computed, not assumed. Products bought together within a window. The median gap between repeat orders of the same item. Starter buyers who later bought the premium version. And the items whose buyers almost never order again.
• The one-and-done finding is a result, not a failure. Those buyers get the review ask and the referral ask; a fourth attempt at a second order the data says does not exist costs goodwill and buys nothing.
• Selling pressure grows with distance from the payment. Confirmation and delivery are service. First success is barely a sell. Replenishment and upgrade are the two moments where an ask is welcome, because by then the customer has something to say yes to.
• A discount here is usually a donation. This audience converts on relevance. The replenishment buyer was coming back at full price, so a code on that email pays you to lose margin on an order you already had.
• Credit beats percentage on an upgrade. Trade-in and credit framing reads as a next step; a percentage off reads as a correction of what they already bought.
• A holdout is what makes any of this legible. Replenishment buyers return on their own; that is what makes them a pattern. Without a holdout the flow claims their orders and everybody concludes that every send works.

Failure modes worth naming: an upsell landing between payment and delivery, which reads as "we have your money, next"; a cross-sell arriving next to an open ticket; marketing stuffed into a transactional email, which most jurisdictions constrain and every recipient punishes; and one store-wide replenishment cadence applied to a catalogue whose items run out at wildly different rates.

Numbers to hold it against

These are post-purchase numbers specifically. They look nothing like acquisition numbers, because the audience already bought.

• The first-to-second-order step is the biggest one in retention. Repeat rates for a typical store sit in the twenties to low thirties as a share of customers, and the odds of a further order rise sharply once somebody has bought twice. Hence a board aimed at the second order rather than at loyalty in general.
• Order confirmations open at 50 to 70 percent, far above any campaign the store sends. That attention is spent on trust, which is the argument against selling in it quite apart from the law's.
• A well-targeted companion cross-sell attaches in the low to mid single digits of recipients. Reaching ten percent usually means the pairing was obvious rather than the copy clever, and that is fine: obvious pairings are the ones worth a flow.
• Replenishment timing is the lever, not the copy. Send at the item's median inter-order interval minus the delivery lead time. The same email sent at an arbitrary day thirty converts a fraction as well, and the difference is entirely in the timing.
• Subscription conversion off a replenishment reminder commonly lands in the high single digits to mid teens of repeat buyers of a consumable. The small discount that funds it is the one discount on this board that buys lifetime value rather than renting an order.
• Hold out 5 to 10 percent per flow. Smaller and the comparison is noise; larger and you are declining revenue to measure it. Report lift, and expect it to be a good deal smaller than the raw numbers looked.
• Watch unsubscribes per moment, not per flow. A spike at a selling moment means the moment came too early. Move it. Softening the copy addresses the symptom and keeps the mistiming.

What moves these: whether the catalogue is consumable or durable, average order value, delivery lead time, and how well the store's own data actually links customers to orders. Anonymous guest checkouts break more post-purchase flows than bad copy ever has.

What has to be true before a card asks for approval

Pattern cards. A percentage and a base size, both computed from a real export or platform report, with the source and the window named. Dozens of occurrences, not three. "Some people buy the brush after the paint" is a story; "31 percent of paint buyers add a brush within 30 days" is a flow.

Moment cards. A trigger and a timing taken from a pattern rather than from the calendar, and a note on where it sits on the selling gradient.

Offer cards. The discount decision made explicitly, the floor named before the first send, and the person who signs off on exceptions. Post-purchase discounts leak margin invisibly precisely because the audience converts either way.

Emails. Transactional messages stay informational: order summary, honest delivery window, one useful tip, support contact. Selling messages lead with what the customer actually bought, by name and quantity. Suppressions applied: open tickets, returns in progress, delivery delays, and the exits from welcome and winback. One primary call to action, and no placeholder surviving into approval.

The result card. Lift against the holdout, margin after incentives, and time to second order, which is what separates accelerating a purchase from harvesting one.

The seed cards are the first instance, not the ceiling

Seven cards show the shape. A live post-purchase programme is denser, and the Create phase is where the thinness costs money.

A healthy Diagnose holds one card per pattern that clears the evidence bar, usually three to six including the one-and-done finding. Moments holds three to five. Offers holds one card per pattern-and-moment pair, so four to eight. Create holds one email per moment per pattern wherever they differ, six to twelve rather than two: the companion email for a paint buyer and the one for a subscription-box buyer are not the same email with a variable swapped. Measure holds the result card, re-read quarterly, plus a per-pattern card where the flows genuinely diverged.

Two email cards on this board means two of the five moments are running and the other three are ideas.

The layout is one arrangement of many

The columns run by production step: diagnose, moments, offers, create, launch, measure. That fits the first build, because the thinking has to happen in that order. Once the flows are live, a column per pattern often reads better, since each pattern is its own small programme with its own moment, offer, email and result. Arranging by moment works too, for stores where the moments matter more than the catalogue. Rearrange it; one pattern, one moment, one offer survives every layout.

When this is the wrong playbook

Below a few hundred orders, the patterns will not clear the evidence bar and the honest answer is that there is nothing to mine yet: build the welcome flow and the acquisition first. If the catalogue genuinely is one-and-done, this is a review and referral board wearing the wrong name, and running it fully will produce four phases of confirmation that nobody buys twice. If the product or the delivery is broken, post-purchase mail is an accelerant rather than a fix. And if first orders are the actual shortage, the second one is not where the quarter gets saved.