Cohort analysis means grouping customers by a shared starting point, usually the month of their first purchase, and watching how each group behaves afterwards. On Amazon this is harder than on your own website, because Amazon does not hand sellers a customer list. You can still get a useful answer to the question cohorts are meant to settle: do the customers you pay to acquire come back, and is that worth the money? This guide shows what you can measure, what you cannot, and how to make decisions from it.
TL;DR
- A cohort is a group of buyers who share a starting point, such as the month of their first order. Tracking each group over time shows whether retention is improving or slipping.
- Amazon does not give sellers customer-level data. Brand-registered sellers get aggregated repeat purchase figures in Brand Analytics, which is the most reliable native starting point.
- Sellers have reported that Amazon’s data changes in 2026 made order-level customer matching less dependable in third-party tools, so treat any customer-level cohort figure with caution.
- You can build a product-level and time-based “cohort-style” view that still supports real decisions on ads, pricing and stock.
- Retention only matters if the product earns a profit. Check unit profit before you chase repeat buyers.
Table of Contents
- What cohort analysis actually tells you
- Why Amazon makes this harder than a Shopify store
- The native option: Brand Analytics
- Build a cohort-style view yourself
- Which products suit cohort thinking
- How to act on what the cohorts show
- Protect the repeat purchase
- Our perspective
- Get the profit figures your cohorts depend on
- A simple quarterly routine
- FAQ
- Sources
What cohort analysis actually tells you
Imagine two months of sales that look identical on a dashboard. In March you sold 400 units, and in April you sold 400 units. A blended report says nothing changed. A cohort view might show that March buyers came back at a much higher rate than April buyers, perhaps because April’s sales were driven by a deep discount that attracted one-off bargain hunters.
That is the value. Cohorts separate new customer quality from sales volume. Each cohort row answers three questions:
- How many customers did you acquire in that period?
- How many of them bought again in the following months?
- What was each group worth in revenue and, more importantly, profit?
For an Amazon seller these answers feed straight into decisions: how much you can afford to bid to win a first order, whether a promotion buys loyal customers or one-off ones, and whether a consumable product deserves a different ad strategy from a one-time purchase.
Why Amazon makes this harder than a Shopify store
On your own store you have an email address, a customer ID and an order history. On Amazon you have none of that for the buyer. Amazon owns the customer relationship, and sellers see orders, products and fulfilment data, not a profile of who bought.
That is why traditional cohort tools, built around a customer ID, struggle on marketplaces. Third-party analytics providers have also reported that changes to Amazon’s Selling Partner API in early 2026 reduced their ability to tell whether two orders came from the same person. Because the detail of that change is best checked against Amazon’s own developer documentation, check what any tool you use says about how it matches customers before you trust its repeat-rate figures.
The practical conclusion is simple. Do not buy or build a cohort report that promises customer-level precision on Amazon without asking how it identifies a customer. Where the answer is vague, treat the numbers as an estimate.
The native option: Brand Analytics
If you are enrolled in Amazon Brand Registry, you already have a legitimate source of repeat-purchase data. Brand Analytics is a set of dashboards in Seller Central that shows aggregated customer data, including customer loyalty reporting. The Repeat Purchase Behavior report within it gives you the number of repeat customers, repeat purchase orders and sales from repeat purchases for your brand.
To find it, go to Seller Central, then Brands, then Brand Analytics. Amazon staff have confirmed in a Seller Central forum post that Brand Registry enrolment is required to see these reports.
What Brand Analytics does well
- It uses Amazon’s own full customer data, so it is not an estimate built from fragments.
- It breaks repeat behaviour down by period, so you can compare one month with the next.
- It is free to brand-registered sellers.
Where it falls short
- It is aggregated. You cannot follow a named group of customers from their first purchase onward, so it is not a true cohort table.
- It is only available to brand-registered sellers. If you resell other brands’ products, you will not see it.
- It says nothing about cost. A repeat purchase that comes from a sponsored ad you paid for is worth less than an organic one, and the report does not separate them.
Use it as a monthly health check, not as the complete answer.
Build a cohort-style view yourself
You do not need customer IDs to learn from cohorts. You can group by acquisition period and product instead of by person. Here is a method that works with a spreadsheet and a few Seller Central exports.
Step 1: Define the cohort
Choose the month as your unit. For each month, record the number of orders for one product or one product family. This is your starting group, even though you cannot name the individuals in it.
Step 2: Record the later behaviour
From Brand Analytics, note the repeat purchase figures for the same product in the following three, six and twelve months. Place them in a table with the acquisition month down the side and months since acquisition across the top.
| Acquisition month | Orders in month | Repeat orders, month +1 | Month +2 | Month +3 |
|---|---|---|---|---|
| January | Your figure | Your figure | Your figure | Your figure |
| February | Your figure | Your figure | Your figure | Your figure |
| March | Your figure | Your figure | Your figure | Your figure |
As figures arrive, express repeat orders as a percentage of the starting group. Reading across a row shows how one month’s buyers behave. Reading down a column compares months at the same age.
Step 3: Tag what drove the acquisition
Note anything unusual that month: a Lightning Deal, a coupon, a change in advertising spend, a stock-out, or a price change. This is the step most sellers skip, and it is what turns a table into an explanation. If the cohorts with a coupon repeat less, you have learned something about what that coupon buys.
Step 4: Add profit per order
A cohort that repeats well but earns almost nothing per order is not a good cohort. Add a column for profit per unit in each month, after Amazon fees, advertising, returns and cost of goods. Our guide to SKU profitability on Amazon walks through how to work out what each product really earns, which is the number your cohort table needs.
Which products suit cohort thinking
Not every product has repeat buyers. Be honest about which of yours do.
- Consumables and refills. Supplements, pet food, cleaning products and filters are bought on a cycle. Cohorts are very informative here.
- Products with accessories or upgrades. A first buyer may return for a complementary item. Repeat behaviour may show at brand level rather than on a single listing.
- One-off durable goods. A kettle or a tent is rarely bought twice in a year. Cohort analysis tells you little, so focus on conversion, reviews and profit per sale instead.
If most of your catalogue falls in the third group, do not force a retention model on it. A seller of one-time purchases gains more from tightening ad waste and getting reviews than from studying repeat rates.
How to act on what the cohorts show
If repeat rates are falling
Check the obvious causes first. Did you change the product, the packaging or the supplier? Did reviews slip? Did a competitor launch a cheaper alternative? Then look at the acquisition mix. A rising share of orders from discounted promotions often drags repeat rates down.
If one cohort outperforms the rest
Find out what was different that month. If the answer is a specific keyword, a bundle or a price point, repeat it deliberately. Analysing the best cohort is usually more profitable than fixing the worst.
If repeat orders are strong but profit is thin
This is common and often missed. Loyal customers are worth little if fees, returns and ad costs eat the margin on each order. Review your cost structure, including FBA dimensional weight charges and storage, before you spend more on acquisition.
If advertising is buying the first order
The argument for accepting a break-even first sale is that a share of buyers return. That only holds if repeat purchases are real and measurable. Set a ceiling on what you will pay for a first order using your observed repeat rate, not a hoped-for one, and revisit it quarterly. It is also worth checking whether your ads are simply capturing sales you would have made anyway.
Protect the repeat purchase
Repeat buyers only return if the product is in stock and the experience was good. Two operational habits support any cohort strategy:
- Keep stock continuous. A stock-out breaks the buying habit and can lose ranking. See our guides on automating replenishment without letting FBA fees erode margin and on calculating FBA headroom.
- Stop paying for clicks on products you cannot ship. Here is how to pause ads when a product is out of stock.
Reviews matter too, because they influence whether a first-time buyer chooses you and whether they trust you enough to buy again. If you want more of them, read how to automate review requests on Amazon UK safely within Amazon’s rules.
Our perspective
We build tools for Amazon sellers, and we will be direct: cohort analysis is a useful lens, but most sellers we talk to are not held back by a lack of cohort tables. They are held back by not knowing, order by order and SKU by SKU, what they actually keep after fees and advertising.
That is why we would sequence the work in this order. First, get an accurate picture of profit per product. Second, get advertising under control so you are not paying twice for the same sale. Third, once those numbers are trustworthy, layer in retention analysis for the products where repeat buying is realistic.
We would also be cautious about any claim of perfect customer-level tracking on Amazon. Amazon keeps buyer identity to itself, and we would rather you work with aggregated, honest numbers than precise-looking ones that rest on guesswork. Our own tools focus on what Amazon’s data can support reliably: profit tracking, PPC optimisation and automation, listing improvements, inventory, review requests and reimbursement tracking. We do not present them as a replacement for Brand Analytics, and you should use both.
— Harry
Get the profit figures your cohorts depend on
A cohort table is only as good as the profit number beside it. Osellpa connects to Amazon’s API to track profit and loss by product and to manage PPC campaigns automatically, so the unit economics are in one place rather than scattered across spreadsheets. Plans are Launch at £10 per month (£96 per year), Scale at £20 per month (£192 per year) and Advanced at £40 per month (£384 per year). You can compare them on the Osellpa pricing page.
A simple quarterly routine
You do not need a complicated system. Set aside an hour each quarter:
- Export orders and profit per unit for your top products by month.
- Pull the Repeat Purchase Behavior figures from Brand Analytics for the same months.
- Update your table and calculate repeat percentages.
- Mark each month with its promotions, price changes and stock-outs.
- Pick one change to test next quarter, such as a different launch discount or a bundle, and note what you expect to happen.
Writing down the expectation matters. Without it, it is easy to explain any result after the fact.
FAQ
What is an Amazon cohort?
An Amazon cohort is a group of customers who share a starting point, most often the month they first bought from you. You follow that group over time to see how many return and how much they spend. Because Amazon does not share customer identities, sellers usually work with aggregated or product-level approximations.
Can I run cohort analysis without Brand Registry?
Yes, but with less data. Without Brand Analytics you can still group your orders by month and product, record promotions and costs, and compare periods. You will not see Amazon’s repeat purchase figures, so treat conclusions about loyalty as tentative.
How many months of data do I need?
For consumables, six to twelve months gives you enough to see a repeat cycle. For products bought less often you may need longer, and the exercise may not be worth it. Start with your two or three best-selling repeat-friendly products rather than the whole catalogue.
Is cohort analysis worth it for a small or part-time seller?
Only if you sell something customers buy repeatedly. If you do, even a simple monthly table is valuable. If you do not, your time is better spent on accurate profit tracking, advertising waste and stock availability.
