This chapter is about measuring the Amazon business at the level of the customer rather than the order: how to classify a product by its repeat-purchase rate, where in Seller Central the repeat and loyalty data actually lives, and how to approximate customer acquisition cost and lifetime ACoS (LACoS) from reports Amazon never designed to be joined. It then covers what to do with those numbers — reading the loyalty segments and their forecasts, finding the first-purchase ASIN that acquires the most valuable customers, and pulling the levers (Subscribe & Save, virtual bundles, variation consolidation) that raise lifetime value once you can see it. Throughout, the honest caveat is that none of this reconciles to an exact figure: Amazon's own definitions of "new" and "repeat" disagree with each other.
Every metric Amazon puts in front of a seller by default — ACoS, TACoS, unit margin, the profitability math from Product Research & Validation — treats a sale as a one-off event. The customer arrives, converts, and disappears from the scoreboard. That framing is what makes so many accounts look marginal on paper while the brands next to them, selling similar products at similar prices, are comfortably profitable.
The claim behind this whole chapter, per Customer Lifetime Value (LTV) as the Primary Amazon Margin Lever, is blunt: lifetime value — how much a customer spends with the brand across repeat purchases — has a larger impact on margin than nearly any other lever available, including supplier cost negotiation, packaging optimization, and shipping cost negotiation combined. Moving a customer's average purchase frequency from once a year to twice a year outweighs the gains from all of that operational grinding. And Amazon does not surface it anywhere obvious. LTV has to be dug out of Brand Analytics and back-calculated from percentages, which is why sellers who never do the exercise are optimizing against an incomplete picture of their own economics.
Before doing the work, though, find out whether it matters for your catalog. Repeat-Purchase Rate Product Classification (Consumable / Semi-Consumable / Non-Consumable) uses the share of sales coming from repeat purchasers as the classifier: 20–50% is typical for a consumable, as low as ~8% marks a non-consumable in the example case, and semi-consumables fall between. That number decides how much of the "shift budget toward new-to-brand acquisition" argument later in this chapter applies to you — it's strongest for consumable and semi-consumable brands, where a new customer reliably becomes a repeat buyer.
One caveat on reading that classification mechanically. Study Key's flashcards are non-consumable — nothing runs out and needs replacing — yet the brand saw a high repeat-purchase rate anyway. For a product with no replenishment cycle, repeat purchases can't be a reorder signal, so a high rate instead reads as customer satisfaction and brand loyalty: repeat buyers, gifters, or customers expanding into adjacent product lines. Same number, different meaning depending on what you sell.
On a DTC storefront like Shopify or WooCommerce the seller owns the full funnel and customer-level metrics fall out of one system. On Amazon the underlying data exists but is scattered across three, none of which was built to be joined to the others.
Alongside these sits Customer Loyalty Analytics (Brand Analytics Dashboard) (Seller Central hamburger menu → Brands → Brand Analytics → Customer Loyalty Analytics), which segments customers by recency, frequency, and spend and has both a Brand aggregate view and a Segment drill-down. It also surfaces the average repeat purchase interval — the number that tells you how long a lookback window on retargeting ads should actually be. And Brand Analytics Demographics Tool (Customer Avatar) (Brands tab → Brand Analytics → Demographics) profiles the customer base by age, income, education, gender, and marital status.
Now the caveat that limits every calculation in the next two sections. Brand Analytics counts a purchase as "repeat" only if the same customer bought the same product within the prior 90 days. Amazon Advertising's own "new-to-brand" definition treats anyone who hasn't purchased in 365 days as new. A customer that one system calls new, the other may already call repeat. So figures pulled from the repeat-purchase report cannot be reconciled precisely with new-to-brand orders from the ad console — not because you pulled them wrong, but because the two systems are answering different questions.
With those reports in hand, Amazon Customer Acquisition Cost (CAC) Calculation Method stitches together an approximate customer acquisition cost. Base all of it on Units Ordered, not Total Order Items.
The headline formula is blended CAC:
Blended CAC = Ad Spend / (Total Orders − Repeat Unit Orders)
Total Orders and Repeat Units both come from the Repeat Purchase Behavior report (Consumer Behavior Analytics (Repeat Purchase Behavior Report)). Netting repeat units out of total orders approximates the count of new customers; dividing all ad spend by that count gives a blended figure that captures customers won organically as well as through ads. It's the same lens TACoS applies to advertising efficiency — total cost against total result — pushed down to the customer level. In one seller's worked example this came out to roughly $4.46.
There are three routes to an ads-only figure, all imperfect:
Treat all three as a cross-check on each other rather than as proof. Because of the 90-day versus 365-day definition mismatch described above, none of them reconcile to a single precise number. What they give you is an order of magnitude and a trend line, which is enough to decide whether an acquisition price is sane — and that's the honest ceiling on what this data supports.
Worth flagging plainly: the material here gives one blended worked example and a handful of illustrative inputs, not a full worked P&L across a catalog. If you want a per-product true margin and CAC computed automatically, My Amazon Guy has in-house software that pulls all of these reports and does it, aimed at repeat-purchasable CPG brands — but it isn't sold publicly, so it's a note about what's possible rather than a tool you can go buy.
A customer's literal lifetime is 80–90 years, which makes it useless for payback math. LACoS (Lifetime ACOS) and Lifetime TACOS — Sophie Society's methodology — solves this by bounding the calculation to a fixed window, typically one year, and asking what percentage of that window's revenue from a customer was consumed by ad spend. That's Lifetime ACoS (LACoS); the same treatment applied to total sales gives Lifetime TACoS.
The calculation runs through LTV Calculation Spreadsheet (Back-Calculating LACoS), a proprietary Sophie Society spreadsheet built specifically to reconstruct what Brand Analytics refuses to show. Its inputs are the percentages the repeat-purchase report does expose — repeat customers, repeat customer %, repeat ordered units, repeat ordered units % — plus product price and current ACoS. Its outputs are total customers, total units, average units per customer, and lifetime ACoS. That first output is the missing link: Amazon never displays a total customer count, so everything has to be inferred from ratios. The spreadsheet isn't a public self-serve tool; per the source, it's obtained by commenting on the video post or emailing Sophie Society directly.
What you buy with that number is permission to spend. Once LACoS is known, a brand can justify a headline ACoS that would normally trigger an emergency review — even 90–100% — because the first sale is deliberately bought at breakeven or a loss, and the repeat-purchase rate means a chunk of future revenue from the same customer arrives without further ad cost. The practical rule: before cutting a high-ACoS campaign, check whether the product's repeat-purchase rate justifies it. Don't manage to the headline number alone.
This has a direct consequence for budget allocation, which is where Ad Console New-to-Brand Reporting (LACoS-Justified NTB Spend) comes in. The Ad Console has a report showing ROAS/ACoS specifically for new-to-brand Sponsored Ads campaigns, and those campaigns reliably run at lower ROAS and higher ACoS than repeat-customer campaigns — unsurprisingly, since they target the subset of shoppers who have never bought from you. Confirm the gap in that report, then use LACoS rather than headline ACoS to argue for continuing or increasing that spend. For a consumable or semi-consumable brand, that argument says to tilt budget toward new-to-brand acquisition, because those customers become repeat buyers with good long-run economics. For a brand sitting at ~8% repeat sales, it says much less.
Re-running the full calculation constantly isn't necessary. The ongoing monitor is the repeat purchase rate itself, tracked over time in Brand Analytics: a rising repeat rate signals improving lifetime economics without redoing the model.
Aggregate LTV tells you the average customer's worth. Customer Loyalty Segments (Top Tier / Promising / At Risk / Hibernating) tells you which customers those are. The Customer Loyalty Analytics dashboard buckets the base into four segments by recency, frequency, and spend:
Each bucket has its own move. Top Tier and Promising respond to brand-tailored promotions: they're already warm and cheap to convert, so promotions aimed at them hold a good ACoS — buying back a repeat sale from a known customer is far more efficient than acquiring an equivalent new one. Usefully, Amazon labels these two segments internally as "high spend customers" and "promising customers," and Amazon-native promotions and coupons can be targeted directly at those labels. At Risk customers are addressed with sponsored display purchase retargeting on a shorter lookback window; Hibernating customers with the same mechanic on a longer one — and the dashboard's average repeat purchase interval is the number that tells you what those windows should be. The campaign mechanics themselves belong to PPC Optimization, Analytics & Advanced Targeting.
The part that's easy to miss is the forecast. Drilling into any segment inside Customer Loyalty Analytics (Brand Analytics Dashboard) opens a view plotting last year's actual sales for that segment against Amazon's predicted sales for the coming year, labeled as growth or decline. Check the Top Tier line specifically. A Top Tier segment forecast to decline means the base looks healthy today but is quietly eroding underneath the top-line number — treat that as a retention emergency and point purchase retargeting and segment-targeted promotions at it before the decline shows up in revenue. Set a longer time range and start from the segment pie chart to see the distribution.
Running alongside the segments is Brand Analytics Demographics Tool (Customer Avatar), which composes a picture of who is actually buying — age, income, education, gender, marital status. That avatar feeds three downstream systems rather than sitting in a report: Sponsored Ads audience targeting in the Ad Console, DSP targeting (particularly valuable for consumables brands, where repeat-purchase economics reward hitting the right audience early), and listing copy written to that customer's language and priorities rather than to a generic buyer — see Listing Content & Conversion Design for the copy and imagery side. One presenter treats the demographics data with explicit skepticism about its accuracy and source and uses it only directionally, for instance to bias model casting in product images toward the customer's likely gender and age skew, while still varying age, gender, ethnicity, and body type across the image set unless the product genuinely targets one group exclusively. Directional input, not demographic fact.
Everything above measures customers in aggregate. The next question is sharper: across a multi-SKU catalog, which product is bringing in the customers who go on to spend the most?
Amazon answers this internally with DSI — downstream impact (DSI (Downstream Impact) Framework): for a given customer action, how much is it worth in future value? Amazon reportedly used DSI to decide which seller categories to prioritize recruiting, valuing a first wireless purchase at roughly $1,200/year, a PC at ~$3,000, and a camera at ~$2,000 — which is why it courted PC sellers over camera sellers. The interesting part is that a seller can reverse-engineer the same logic without any access to Amazon's internal dashboards:
The ASIN that comes out on top is the gateway drug ASIN (Gateway Drug ASIN Targeting): the first-purchase product generating the highest average downstream repeat value, the one most likely to turn a first-time buyer into a high-LTV repeat customer. Once you've identified it, the strategy is concentration rather than even spread — push PPC spend specifically at driving first-time purchases of it, run price promotions to lower the acquisition barrier, and invest disproportionately in its content and listing quality to maximize conversion.
Notice what that commits you to. The acquisition cost on the gateway ASIN is justified by downstream value that lands elsewhere in the catalog, not by that ASIN's own unit economics. Judged alone it may look like a mediocre product to be advertising this hard. That's the same reasoning that governs line extensions and brand-family expansion in Scaling, Omnichannel & Brand Growth, and it only holds if you actually have the repeat data to back it — otherwise it's just overspending with a story attached.
Measurement is only useful if something changes afterward. Four levers show up in this material, and all of them raise revenue per customer without competing for ad budget.
Subscribe & Save. Subscribe & Save trades a discount for scheduled recurring deliveries, and it's the most direct LTV instrument Amazon offers for consumables. The case figure that anchors the argument comes from the In Motion Hemp audit (Helium 10 Scale Stories): subscribers spend roughly $300 over their lifetime on average versus $49 for non-subscribers — about a 6x gap. But that brand, a ~$140K/year hemp pain-relief cream seller, had set its S&S price nearly identical to its one-time price, silently erasing any financial reason to subscribe. The program was technically enabled and economically inert. The prescribed fix: discount at least 5% (10% pending a full P&L review), stack a first-time S&S coupon redeemable only on a customer's first subscription order as a "double hook," and target roughly 30–35% of daily orders coming through Subscribe & Save. Before concluding that a subscription program isn't converting, audit the S&S price against the one-time price — that's the first check, not the last.
Note the strategic split here, because the sources genuinely differ. The mentors used the $49-vs-$300 multiple, not the discount percentage, as the business case for pushing S&S hard — accepting thinner margin now to buy lifetime value. But one guest's brand, Jungle Powders at ~$3–4M/year, layers S&S coupons on purely to lift conversion rate while the business still earns full margin every month, explicitly not as discount-for-LTV growth. Both are defensible; which one you run depends on whether monthly profitability or scale-for-LTV is the stated goal.
Auto Buy is a structural threat to all of this. Amazon Auto Buy (Price-Trigger One-Time Purchase Feature) lets a shopper set a maximum price and a watching window on a product page; Amazon then automatically completes a one-unit, one-time purchase the moment the listing price drops to or below that threshold. It's FBA-only and excludes any listing running a coupon or promotion. Because it has no promo exception, any discount you run to move inventory or win a Buy Box fight instead clears the Auto Buy queue at the lowest available price — converting a would-be subscriber into a one-off sale. It also distorts demand signals: a shopper who sets a trigger and walks away stops comparison-shopping, so search data goes quiet on customers who are still in-market rather than genuinely churned. The countermeasure is to make Subscribe & Save the easier choice before anyone reaches for Auto Buy: a subscription discount that beats waiting, a stable list price that keeps the 90-day-low trigger out of reach, and reliable FBA stock.
Virtual bundles. Amazon Virtual Bundles (Buy Box Cross-Merchandising) pair two complementary owned ASINs — a roller with its companion lotion, say — into a purchasable bundle Amazon surfaces in the ad slot directly under the buy box, with no PPC spend. That does two jobs at once: it cross-educates customers into the rest of the range, and it occupies real estate a competitor's sponsored ad would otherwise fill. Since bundles cost nothing to set up, that's a free, permanent placement denied to competitors. In Motion Hemp's audit recommended exactly this after diagnosing its single-SKU listing as a growth ceiling.
Variation consolidation. Single-SKU-to-Variation Consolidation (Merging Standalone Listings for AOV/LTV) addresses a common structural mistake: selling a single unit and a multipack as separate standalone listings. That fragments reviews, traffic, and ranking signal across ASINs and forces the customer to compare pages instead of picking an option. Merging them into one parent-child family concentrates demand and lets Amazon's variation UI upsell the multipack from a listing the customer already trusts. Case data shows the merge lifting single-item conversion rate, average order value, lifetime value, Subscribe & Save uptake, and daily order volume, while cutting the repeat ad spend otherwise needed to resell the same customer a different pack size. This is a retroactive merge of already-live SKUs for AOV/LTV economics — a different move from using variations to transfer reviews and rank to a new launch, which belongs to launch-time ranking work.