RETENTION · MARKETING · CRO · 17 SEPTEMBER 2026 · 8 MIN READ
Cyber Week is over. The retention window is open
Post-BFCM retention is decided by the parcel, the returns process and the timing of the second contact — not by the discount you send in December.
Mostly by things you build before they buy, not by what you send afterwards. A Black Friday buyer decides whether to come back based on whether the parcel arrived when you promised, whether returning something was painless, and whether there is a reason to buy again that is not a discount. Post-BFCM retention programmes that open with "welcome back, here is 20% off" convert the cheapest part of the cohort and teach everyone else to wait for the next code. Build the delivery promise, the returns experience and one well-timed non-promotional contact, and accept that a portion of this cohort — gift buyers, deal-only buyers — is not worth spending on at all.
IN SHORT
- The retention window opens the moment the first order is placed, which is why almost everything that works in it has to be built before the traffic arrives.
- Shopify defines a returning customer as one whose order history already includes at least one order, so the metric that matters is the second order, not lifetime value projections.
- A discount as the first post-purchase contact converts the people who were coming back anyway and trains everyone else to wait for the next code.
- Shopify's customer reports are based on a customer's entire order history rather than the selected timeframe, so a blended repeat-purchase rate will not tell you how the peak cohort is behaving.
- Cohort analysis groups customers by the date of their first order, which makes the peak cohort readable on its own — and it is the only reading worth doing.
- Store credit is a genuine alternative to a percentage-off code: Shopify documents accounts held per customer in a specific currency, with an optional expiry date on each credit.
- Some of this cohort should not be retained. Gift buyers bought for someone else, and a single-unit deal buyer at negative contribution margin is a cost you can choose not to repeat.
Why this is a September post about December
The retention window does not open when the campaign ends. It opens when each order is placed, and by then every lever that genuinely moves the second purchase has already been set: what you promised about delivery, whether you can keep it, what the packaging says, how hard it is to return something, and whether anybody has written the email that goes out fourteen days later.
Which means the useful version of "how do we retain Black Friday customers" is a question asked now, in the quarter before, when the answers are still changes you can make. Asked in December, it collapses into the only lever left — a discount — which is the one that does not work.
The rest of this is what to build while there is still time to build it.
The cohort is not one cohort
Peak acquires several different kinds of people at once and they are worth very different amounts. Separating them is the first piece of work, because a single retention programme aimed at all of them is aimed at nobody.
- People who wanted your product and waited for a sensible moment. The best customers you will acquire all year. They needed no discount to be interested, only permission. Treat them as ordinary new customers and the second order follows the normal path.
- Deal-only buyers. They bought a price, not a brand. Some convert into real customers; most do not, and the ones who do convert because the product was good, not because the follow-up email was clever. Spend accordingly.
- Gift buyers. They bought for somebody else. They are not a customer for that product, they will be over-represented in January returns, and a replenishment reminder to them is a small embarrassment.
- Existing customers who stocked up. Not retention at all. These are orders pulled forward from Q1, and if you count them in the peak cohort your repeat rate will look excellent in December and terrible in March.
What actually decides the second order
Three things, in this order, and none of them lives in an email tool.
Did the parcel do what the page said? A delivery date shown at the point of purchase and then met is the single strongest signal you can send that this is a store worth using again. Missing it is worse than never having shown a date, because you have converted a vague expectation into a specific broken promise. If the peak carrier situation means you cannot commit to a date in the last week of November, say a range and hit the top of it.
Was returning something easy? January returns are a retention event that most stores treat as a cost line. The person asking for a refund is in the highest-attention conversation they will ever have with you, and they are comparing you to the retailer who sent a label in the box. A store credit offered as an alternative to a refund converts some of those into a second order — Shopify documents store credit accounts held per customer in a given currency, with an optional expiry on each credit — but only where the process was painless first. Offered as friction on the way to a refund, it costs you the customer.
Is there a reason to come back that is not price? Something that arrives when the product runs out, or when the next size is needed, or when the thing they bought has a companion worth owning. This is not a campaign; it is a piece of timing derived from your own repurchase interval, and it is the only email in the sequence that reliably earns a second order at full margin.
The sequence, and what not to put in it
A workable post-peak sequence is short and mostly not promotional.
- Days 0–7: operational only. Dispatch, tracking, arrival. Nothing to sell. The job of this week is to be accurate — and a tracking email that is correct outperforms a "thanks for your order" that is not.
- Days 7–14: use, not buy. How to get the most out of the thing they bought. Care, setup, sizing, the answer to whatever your support inbox is asked about this product most. It reduces returns and it earns the right to send the next one.
- Days 14–45: the second-order prompt, timed to the product. Whatever the natural interval is for the item they bought, not a fortnight because a template said so. For consumables that is consumption timing; for durables it is the companion product or the next size.
- January: the returns conversation, done properly. Fast, no argument, a credit offered rather than imposed.
- Not in the sequence: a discount in week one. It converts people who had already decided, at a cost, and it tells the rest that November's price was not the real price.
Do not buy a loyalty platform in December
The December proposal is usually software: a points programme, a retention suite, a customer data platform, bought in the weeks when the peak cohort is still warm and the urgency argument writes itself. It is the wrong month and often the wrong purchase.
A points scheme is good at concentrating spending you were already partly getting, and poor at creating a relationship where none exists. Installed on top of a delivery experience that missed its dates, it adds a monthly fee to the problem. The honest sequence is to fix the reasons people did not come back, then measure whether there is a residue that a programme would address — and the measurement takes a quarter, which is exactly why nobody does it before signing.
The same applies to a CDP bought to "understand the peak cohort". Shopify's own segments are dynamic and built from behavioural attributes — order counts and dates, amount spent, products purchased — and three or four segments built from those, synced to the email platform you already pay for, answer more questions than a new system of record you will spend two quarters implementing. If you cannot describe the decision the segment changes, the platform will not supply one.
Measuring it without fooling yourself
Blended repeat-purchase rate is the wrong instrument here, and the reason is documented: Shopify's customer reports are based on a customer's entire order history rather than only the orders in the selected timeframe, so a period-level repeat rate mixes the cohort you are asking about with every cohort before it. A December figure can improve while the November cohort behaves badly, simply because your 2024 customers kept shopping.
Use the cohort view instead. Customers are grouped by the date of their first order, which makes the peak intake a population you can watch on its own. Two numbers are worth reporting: the share of that cohort placing a second order, and the margin on those second orders. The second number is the one that decides whether the acquisition was worth doing, and it is the one most post-peak reviews omit — a cohort that returns at a good rate on 30%-off codes is a cohort you are paying to keep.
Give it ninety days before drawing a conclusion, and set the window from your own repurchase interval rather than a benchmark. A coffee subscription and a mattress brand should not be reading the same clock.
One honest possibility worth holding open: the answer may be that this cohort does not repeat, and that the right response is to acquire fewer of them next November at a shallower discount. That is a better finding than a retention programme that spends the rest of the year trying to fix the acquisition.
Questions this raises
When should the first non-transactional email go out?
After the parcel has arrived, not after the order was placed. An email about getting the most out of a product that is still in a van is a small failure of attention, and it is the first thing this customer learns about how carefully you operate. Trigger on delivery, not on purchase.
Should we offer a discount for the second order?
Not as the opening move. It converts the part of the cohort that had already decided to come back and establishes that your list price is negotiable. If you need an incentive, store credit with an expiry is a better instrument than a percentage code — Shopify supports an optional expiry date on a credit — because it is bounded, tied to a person, and does not reprice your catalogue for everyone who forwards the email.
How do we tell deal-only buyers from genuine new customers?
Look at what they bought rather than the discount they used. Single-unit orders of your cheapest promoted line, no browsing depth, no other products viewed, and no engagement with anything you sent afterwards is a recognisable pattern. Build it as a segment once, attach a deliberately cheap contact strategy to it, and stop spending on individual campaigns for that group.
Is the January returns wave a retention problem or an operations problem?
Both, and the operations half comes first. A slow or argumentative returns process removes any chance the retention half had. Staff it for the volume, make the label easy, and make store credit a visible alternative rather than the default — a refund given quickly buys more second orders than a credit extracted grudgingly.
What repeat-purchase rate should we expect from the peak cohort?
There is no figure we would put on it that would be worth anything. Published benchmarks aggregate stores with different products, margins and repurchase intervals, and a cohort acquired at 40% off behaves differently from one acquired at 10%. Your own previous peak cohort, read at ninety days, is the only comparison that means anything — which is a reason to start recording it now.
Do we need a loyalty programme before Black Friday?
No. Launching one alongside peak means running an untested mechanism through your highest-volume weekend, with support and finance both learning it under load. If the programme is the right idea, it is still the right idea in February, when a mistake costs a normal week rather than the year.
NEXT STEP
Free store audit
A senior Shopify engineer reviews your storefront, theme performance and checkout, then sends a prioritised list of fixes.