We went through this client's Amazon account end to end — profit, advertising, customer behaviour, listings — on their own authorised data, including a customer-acquisition measure most sellers can't see. The fundamentals are sound and the advertising is well run. The upside sits in three levers, ordered the way they play out in practice: price, then lifetime value, then range. Anonymised; their numbers are sealed.
How to read it. Each lever's bar height is its relative upside; the track beneath is the effort to pull it. Price is a quick win you can read in a week; lifetime value is more work and a bigger prize; range is the long game with the most in it. Do them in that order. Values sealed.
Two things frame everything below. Nearly four in five of the customers this advertising brings in are new to the brand — a number Amazon doesn't report for Sponsored Products, which we measured directly. And the search-term spend is tightly controlled, with under a sliver going anywhere wasteful. They acquire new customers efficiently and run clean. The growth is in what happens after the first sale.
The account is healthy and well run — nearly four in five ad buyers are brand-new, and almost nothing is wasted. The growth is in what happens after the first sale.
2026 so far turns £▓▓k of revenue into £▓▓k of net profit — a ▓▓% margin, after real costs. Profit concentrates in baking; nuts are efficient; and their resale lines quietly outperform — higher margin at lower ACOS than own-brand. Two checks worth stating plainly, because they're clean:
Seeing new-to-brand for Sponsored Products at all takes a custom query against Amazon's advertising clean room — most sellers never see it. We ran it across the ad-driven purchases: nearly four in five are brand-new customers. The ads are an acquisition engine, and a well-run one — we stress-tested the search terms for the kind of leak that quietly drains budget and there isn't one. Nothing here needs fixing. The upside is in the three levers that follow.
Most of the range has held one price all year — the sprinkles prove testing works; it's just never been extended.
Most best-selling products held a single price all year. The exceptions are telling: they already price-test their sprinkles range competently — the 200g hero's move lifted margin without losing rank, and they actively range the 1kg with the 500g and mini moving too. So the discipline exists — they just apply it to a handful of products and leave the other four-fifths of the range completely static.
| Where the untested headroom sits | What the data shows |
|---|---|
| Nudge Vanilla Sugar | Category #2, held at one price all year, priced below the market median — a modest rise flows almost entirely to profit. |
| The static four-fifths | ~▓▓ top sellers have never moved. A large untested surface for exactly the margin gains the sprinkle tests already prove out. |
| Fix Banana Chips | ▓▓% margin — re-cost or re-price; advertising it harder won't fix the economics. |
It's the fastest lever they have, and they've already shown it works on sprinkles — it's their own price, changed in minutes and read in a week, with no new tooling or promotions to build. The move is simply to extend the same testing discipline across the range: a rolling programme of controlled steps on the four-fifths never touched, to find where the rest of the headroom is.
Great at acquiring, weak at keeping — three retention levers sit switched off.
They acquire new customers brilliantly, but most buy once. Three retention levers are open to them today, and in 180 days of transactions they've barely touched any:
| Lever | What it does |
|---|---|
| 1. S&S first-purchase discount | Turns a one-off buyer into an auto-reordering subscriber. |
| 2. Brand Tailored Promotions (never used) | Private offers to basket-abandoners, at-risk and repeat audiences. |
| 3. A higher ongoing S&S rate | Where margin allows — makes subscribing worth it on the best repeat products. |
The gold is products people demonstrably re-order but hardly ever subscribe to — a gap only visible by crossing repeat rate with subscription penetration:
| Best first-purchase targets | Repeat rate | Subscribed | Margin to fund it |
|---|---|---|---|
| Walnuts in Shells 500g | ▓▓% | ▓▓% | ▓▓% |
| Vegan Dried Fruit Mix | ▓▓% | ▓▓% | ▓▓% |
| Mix Nuts 1kg (already biggest) | ▓▓% | ▓▓% | headroom |
Three anchor subscriptions are low-stock: Mix Nuts 96 units vs 180 due, Almonds 18 vs 71, Peppercorn 39 vs 48. Missed Subscribe & Save deliveries are how Amazon starts pausing subscriptions.
Medium effort: there are promotions and offers to set up, but the levers already exist on the account and cost nothing to switch on.
Loyal but narrow: repeat households drive the revenue, yet almost none buy a second product.
We resolved their customers by household across two years of their own order data. They're loyal but narrow: repeat households drive most of the revenue — but even over two full years, at most a small share ever buy a second, different product, and barely any reach a third. They love their one thing and don't discover the rest of the shelf.
The data even shows why that's a missed opportunity. The products that acquire the most new customers are one-purchase items — peppercorn and paprika bring the most brand-new buyers — brilliant front doors that currently lead nowhere. Meanwhile the ad mix is almost all Sponsored Products and barely any Sponsored Display, the tool built for exactly this.
We also checked what their multi-product customers actually combine, in Amazon's clean room. The result is stark: the only pairing bought by more than 100 customers is two sizes of their rainbow sprinkles — the 200g and the 500g. There's no cross-category habit anywhere near that threshold. So cross-sell here isn't harvesting existing behaviour, it's building it — and the one place range-buying already happens (within a family, by size) is the template to copy into nuts and seeds first.
1. Consolidate the range into variation families. Almost every nut, seed and dried fruit sits as its own standalone listing — the sprinkles are the one family grouped by size, and they're the only co-purchase. Group the singles under one parent. Start with seeds: parenting Chia, Sunflower and Pumpkin onto the established Omega Seed Mix lets new launches inherit its reviews, ranking and variation picker — rescuing the launches and cross-exposing the range in one move. The rule that keeps search real estate intact: parent the weak or new SKUs onto a strong anchor, and never merge two strong listings that rank on the same term. It's reversible, so run one family and watch total sessions before rolling wider.
2. Then cross-sell with Sponsored Display — idle today — remarketing the range to people who've already bought, and product-targeting across their own catalogue. The two together turn a loyal single-product customer into a two- or three-product one.
| Pros — why do it | Cons & watch-outs |
|---|---|
| New / weak SKUs inherit the family's reviews, rating and rank — launches start with credibility, not zero. | Merging two strong same-term listings can cost a search slot — parent weak onto strong instead. |
| The picker cross-sells the range on one page — the mechanic customers already respond to. | Amazon can split an improper family — usually a painless reversal. |
| Pooled sales velocity lifts organic rank. | A poor variant drags the family's rating — keep families coherent. |
A note on method & accuracy. We grouped orders by delivery postcode as a stand-in for the customer. At their volumes that's a fair proxy — so these figures are upper bounds: grouping by postcode can only ever over-state how much a single customer repeats or ranges, never understate it — so the real picture is, if anything, narrower still. A precise, customer-level view requires a compliant reveal of customer identity (a paid layer), which would sharpen every number here.
Nothing's broken — a few fill-in-the-box structured-data gaps, and recent launches already reading demand right.
Read the way Amazon's search and its AI assistant read them, the listings are in good shape — backend keywords are well populated on live SKUs. The one consistent gap is structured data, and it's fill-in-the-box, not a rebuild:
| Structured field | Missing on | Why it matters |
|---|---|---|
| Allergen information | ▓▓% of SKUs | A food listing without it is invisible to dietary filters. |
| Diet type (vegan / dairy-free) | ▓▓% | The core claim lives only in prose; filters and AI read the fields first. |
| Serving size | every SKU | A field the AI assistant leans on directly. |
These are likely why "kosher" shoppers click but don't convert. On range, they're already ahead of us — Chia, Sunflower, a Super Seed Mix and a Super Berries blend have all launched recently, reading the sustained seed demand correctly. Too new to judge yet; the priority is supporting those launches rather than adding more.
One proxy left — the customer. A precise identity layer sharpens all three levers.
Everything above we measured directly. The one place we work from a proxy is the customer — postcode gets us a directional read, but an upper bound. A compliant, customer-level identity layer turns every retention and range decision from "directional" to "precise", and opens the questions the account can't answer alone:
Resolve who the repeat customers actually are and what they're worth over time — so retention spend chases the customers who come back, and cross-sell targets the products buyers genuinely move to next, not a postcode approximation of them.
A clean-room query for which products customers actually buy together — the exact target list for the cross-sell advertising in lever 3.
Replicate the Subscribe & Save annuity in lifecycle email — higher margin, and the reorder timing is theirs to control, not Amazon's.
Measure whether TikTok activity lifts Amazon search demand — the cross-channel halo most brands spend on blind — and time it against the baking season.
Each is the same method: their own authorised data, measured directly. Start with the easiest lever tomorrow; the precise view is the door to the rest.
A full end-to-end audit on your authorised data — profit, ads, customers, listings — with the levers ordered easiest first.