Most Amazon strategy advice is tactics wearing a strategy costume. A list of things to do is not a strategy. A strategy is knowing what to do first and what to refuse.
The 2022 playbook was straightforward: nail your keywords, run Sponsored Products, get reviews, watch ACoS. It worked because the surface area was small and the arithmetic was forgiving. Neither is true now.
Fees have compounded, a fuel surcharge has been added to fulfillment, ad costs have risen inside the third-largest digital advertising platform on earth, and the discovery layer itself changed shape in May when Amazon moved AI shopping from an optional chat drawer into the main search bar. A seller running the 2022 playbook is not doing something wrong so much as solving a problem that has changed underneath them.
This is the full picture: what the economics actually are, the five layers in build order, and how to allocate budget at each stage of growth.
Marketing stack — the ordered set of capabilities a brand builds, where each layer multiplies the effect of the ones below it. On Amazon the ordering is not preference but arithmetic: advertising multiplies conversion rate, so a strong ad programme on a weak listing produces expensive traffic and nothing else.
Why the Old Playbook Fails
Four things changed, and each one invalidates a common assumption.
- Margins compressed structurally. Total FBA selling costs now average 30 to 35% of revenue, and with advertising the effective take reaches toward 50%. A 2022 model built on 25% is not slightly wrong, it is producing decisions in the wrong direction.
- Advertising became mandatory rather than optional. Amazon's ad business hit $68.6 billion in 2025, up 22% year over year, making it the third-largest digital ad platform globally. That growth is funded by sellers who cannot get visibility without it.
- Discovery split in two. Since May 13, 2026, Alexa for Shopping sits inside the main search bar generating AI overviews and side-by-side comparisons. It runs in parallel with traditional ranking, so you now have to satisfy two systems that read your listing differently.
- The fee structure got more granular. Fuel surcharges, inbound placement fees, low-inventory-level fees, aged inventory surcharges, and new extra-large weight bands mean two products at the same price can have wildly different unit economics.
That a good product at a fair price will find its audience. It will not. Between mandatory fees, mandatory advertising and a discovery layer that surfaces a handful of results rather than a page of them, visibility in 2026 is purchased and engineered rather than earned by default.
The Economics You Actually Operate In
Every strategic decision below follows from this table, so it is worth getting exact rather than approximate.
| Cost | 2026 rate | Notes |
|---|---|---|
| Referral fee | 8–45% | Most categories 15%. Electronics and personal care 8% |
| FBA fulfillment | ~$3.22–6.90/unit | Rose ~$0.08/unit on average from Jan 15, 2026 |
| Fuel surcharge | 3.5% | Applied on top of fulfillment fees |
| Storage (standard) | $0.78/cu ft | January–September |
| Storage (Q4) | $2.40/cu ft | October–December. Triples |
| Advertising | 10–25% of revenue | Optional in theory, required in practice |
| Total mandatory | ~14–55% | Depends on category, price point, fulfillment |
| Typical all-in | 30–35% | Before advertising. Nearer 50% with it |
What this implies strategically
The spread between a 14% cost structure and a 55% one is entirely determined by category, price point and product dimensions — decisions made before any marketing happens. Which means product and packaging decisions are marketing decisions, and reducing a package by half an inch to drop a size tier can be worth more than a quarter of ad optimization.
It also means margin discipline is not a finance function here. If your contribution margin after all fees is under about 25%, no marketing strategy rescues the unit, because advertising has to come out of what remains. Our contribution margin playbook covers the arithmetic, and the 2026 FBA fee breakdown covers every line item.
Layer 1: Conversion Foundation
Everything above this multiplies whatever number this layer produces. That is not a motivational statement, it is arithmetic.
What it includes
- Title, bullets and description written for the current 75-character title constraint and for two audiences: the ranking algorithm and the AI layer reading for meaning.
- Complete structured attributes. Every empty field is a question the AI comparison cannot answer about your product.
- The image stack, which does more conversion work than copy does. Main image, feature callouts, scale reference, lifestyle, infographic.
- Listing video in the main image slot, still unused by a large share of brand-registered sellers.
- A+ content, with Premium A+ now free for eligible brands.
- Review corpus — volume, recency and semantic coverage.
- Buy Box ownership. 82% of sales go through it, so losing it is not a ranking problem, it is a revenue problem.
The benchmark to beat
Amazon's average listing conversion rate runs 10 to 15%, against 2 to 3% on typical ecommerce sites. That is the bar. If your unit session percentage is below 10%, you have a conversion problem and every advertising dollar you spend is being taxed by it.
Advertising multiplies your conversion rate. Multiply a broken number by a bigger budget and you get a bigger broken number, faster.
The build detail lives in our high-converting Amazon listing guide.
Layer 2: On-Amazon Paid
The layer most sellers start with and should not.
| Format | Job | When to add it |
|---|---|---|
| Sponsored Products | Capture existing demand | First, always |
| Sponsored Brands | Category presence, brand terms | Once you have 3+ ASINs |
| Sponsored Brands Video | Win the click before the listing | Once listing video proves out |
| Sponsored Display | Retarget, defend, attack competitors | Once SP is efficient |
| Amazon DSP | Programmatic reach beyond Amazon | Meaningful volume only |
The metric that governs it
TACoS, not ACoS. ACoS tells you the efficiency of ad-attributed sales, which is a campaign metric. TACoS measures ad spend against total revenue including organic, which is a business metric. A healthy TACoS for established products is 8 to 15%. Rising TACoS with flat revenue means you are buying sales you used to get free; falling TACoS with rising revenue means advertising is successfully driving organic rank.
Average ACoS across Amazon campaigns runs 25 to 30%, but a good ACoS is entirely a function of your margin. A 30% ACoS is excellent at 60% margin and ruinous at 25%.
The discipline that separates good accounts
- Defend your brand terms. Cheap, high-converting, and a competitor will take them if you do not.
- Separate discovery from harvest. Broad and auto campaigns find terms; exact campaigns exploit them. Mixing the two makes both unreadable.
- Set a negative keyword cadence. Weekly at minimum. This is where most waste hides.
- Do not scale an unprofitable campaign into profitability. Volume does not fix a broken unit economic.
Structure detail is in the Amazon PPC strategy guide.
Layer 3: External Traffic
Where most sellers either ignore an opportunity or waste money enthusiastically, with very little in between.
Why it works
External traffic builds sales velocity, review volume and conversion signals, all of which are confirmed inputs to Amazon's performance ranking systems. The mechanism is indirect but real: you are not buying rank, you are buying the signals that produce rank.
Why it fails
Because sellers send cold traffic straight to a listing. A visitor from Meta who has never heard of you arrives at a page designed to convert someone already shopping the category, converts at a fraction of native traffic, and depresses the very conversion signal you were trying to improve.
What actually works
- Warm traffic only. Retargeting, email, creator audiences, existing customers.
- Attribution tags on everything. Untagged external traffic is unmeasurable, and unmeasurable spend becomes permanent by default.
- Landing page intermediation for cold audiences, so the qualification happens before the listing sees the session.
- Creator content amplified as paid, which arrives pre-sold rather than cold.
Amazon has not confirmed that external traffic directly improves AI recommendation ranking, and any vendor promising it is asserting something Amazon has not said. The defensible position is that external traffic builds velocity, reviews and conversion — which are confirmed factors. Build the strategy on that, not on the stronger claim.
Tracking setup is covered in our Amazon attribution playbook.
Layer 4: Owned Audience
The layer that determines whether you are building a business or renting one.
Amazon owns the customer relationship by design. You get an order, not a person. Every other layer in this stack improves your position inside a system you do not control, and a category change, a fee increase or a policy decision can compress all of it without warning.
The mechanisms available
- Amazon's Customer Engagement Tool, which lets brand-registered sellers email followers and repeat customers within Amazon's environment.
- Packaging inserts driving to warranty registration, product support or a genuine value exchange. Compliance matters here — inserts must not solicit reviews or divert from Amazon inappropriately.
- Your own storefront, where the customer relationship is yours outright.
- Brand Follow and Posts, which build an on-platform audience you can reach repeatedly.
- Creator and community audiences that exist independently of any marketplace.
The honest sequencing note
This layer is genuinely important and genuinely slow, which is why it gets deferred indefinitely. The right treatment is to start it early at low intensity rather than waiting for a moment when it feels urgent — because the moment it feels urgent is the moment you have already lost the leverage.
Our guides to the Customer Engagement Tool and owning your customers off Amazon cover both halves.
Layer 5: AI Discovery
The newest layer and the one where the competitive gap is currently widest.
On May 13, 2026, Amazon launched Alexa for Shopping, which sits directly inside the main search bar rather than in the optional chat drawer its predecessor used. It generates AI overviews above results and runs side-by-side product comparisons. Every signed-in US shopper encounters it. The predecessor served over 300 million customers from a drawer people had to find; this reaches everyone by default.
What it reads
- Structured attributes, which feed comparisons directly.
- Bullet structure, where leading with the specific feature then the benefit parses more cleanly than benefit-first marketing language.
- Natural-language description covering use cases, materials, dimensions, compatibility.
- Reviews and questions, which supply context you did not write.
COSMO, briefly and honestly
COSMO is Amazon's commonsense knowledge layer, connecting products to situations and needs rather than to keywords. You cannot target it directly and anyone selling COSMO optimization is overselling. What you can do is supply its inputs: complete attributes, natural-language copy about contexts and uses, and content that answers real questions.
Alexa for Shopping does not replace traditional ranking. They run in parallel. A listing that ranks on keyword relevance but cannot answer a conversational question loses an increasing share of discovery, and one built purely for the AI layer will not rank at all.
The algorithm detail is in our COSMO vs A9 vs A10 breakdown.
Budget Allocation By Revenue Tier
| Annual revenue | Marketing as % of revenue | Where it goes |
|---|---|---|
| Under $250K | 20–30% | 80% ads, 20% listing assets. Everything else waits |
| $250K–1M | 15–25% | 70% ads, 20% creative, 10% external testing |
| $1M–5M | 12–20% | 60% ads, 20% creative, 15% external, 5% owned audience |
| $5M–10M | 10–18% | 50% ads, 20% creative, 20% external, 10% owned |
| $10M+ | 8–15% | 45% ads incl. DSP, 20% creative, 20% external, 15% owned |
How to read this
The percentage falls as revenue rises because organic rank compounds — a mature catalog earns traffic that a new one has to buy. If your marketing percentage is not falling as you scale, advertising is substituting for organic strength rather than building it, which is the single clearest sign that layer one was never finished.
Note also that the mix diversifies with scale, not because diversification is inherently good but because the marginal return on additional Amazon ad spend declines while the marginal return on owned audience does not.
Tier-specific spend detail is in our monthly spend by revenue tier guide.
Which layer is your bottleneck?
We will look at your conversion rate, TACoS and margin together and tell you which layer to fund next — and which to stop funding.
Book a Strategy Call →The Ecom Profit Box
Eleven playbooks on listings, conversion, images, and email. Built for operators, no fluff, no email sequence.
Grab It Free →The Sequencing Problem
The order matters more than the budget, and almost everyone gets it wrong in the same direction.
Launch, run Sponsored Products immediately, discover the listing does not convert, blame the ads, increase the budget. Expensive and extremely common.
Get unit session percentage into the 10 to 15% band on modest traffic, then scale spend against a rate that rewards it.
Cold audiences sent to a listing convert poorly and depress the exact signal the traffic was meant to improve.
Once on-Amazon ads are efficient, external traffic amplifies a machine that already works instead of exposing one that does not.
Deferred until a policy change or fee increase makes it urgent, which is precisely when you have no leverage left to build it.
Start at low intensity from the beginning. It compounds over years and cannot be bought quickly when needed.
The gate to apply
Before funding any layer, confirm the one below it is working. Concretely: do not increase ad spend while unit session percentage is under 10%, and do not buy external traffic while TACoS is above 20% on established products. Those two gates prevent most of the money that gets wasted on Amazon.
Metrics That Govern Decisions
Five numbers. Everything else is diagnostic detail.
| Metric | Healthy range | What it tells you |
|---|---|---|
| Unit session % | 10–15%+ | Whether the listing converts. Gate for ad spend |
| TACoS | 8–15% | Whether ads build organic or substitute for it |
| Contribution margin | 25%+ after all fees | Whether the unit can fund growth at all |
| Buy Box % | 95%+ | 82% of sales flow through it |
| Return rate | Category dependent | Whether your listing oversells |
The two most misread numbers
ACoS. Useful for campaign management, misleading as a business metric, because it ignores organic revenue entirely. A campaign at 40% ACoS driving organic rank can be more valuable than one at 20% that does not.
Revenue growth. Growing revenue while contribution margin falls is not growth, it is buying market share with your own money. Track them together or you will discover the problem two quarters late.
The reporting to actually use
Search Query Performance in Brand Analytics gives observed impression, click and purchase data per query rather than a third-party estimate. It is free with Brand Registry and a large share of sellers have never opened it. Our guide to Search Query Performance covers the reporting properly.
The 12-Month Calendar
Amazon is a seasonal business whether or not your product is, because the fee structure and the competitive intensity both move.
| Period | Focus | Why then |
|---|---|---|
| Jan–Feb | Fee review, listing overhaul, Q5 tail | New fee rates land January 15; competition is quiet |
| Mar–Apr | Creative production, new ASIN launches | Cheap CPCs, time to build before summer |
| May–Jun | Prime Day prep, inventory in | Deal submissions and stock deadlines fall here |
| Jul | Prime Day execution | Highest-velocity window before Q4 |
| Aug–Sep | Q4 readiness, creative shot, inventory shipped | Storage triples in October; act before it does |
| Oct–Dec | Execute, do not experiment | Q4 storage at $2.40/cu ft; freeze changes |
| Dec 26–Jan 15 | Q5 — cheap traffic window | Gift card redemption, collapsed competition |
The two dates that matter most
The October storage cliff. Standard storage goes from $0.78 to $2.40 per cubic foot in Q4. Inventory sitting through that window costs more than three times as much to hold, which makes September the deadline for both shipping in and clearing slow movers.
The Q4 change freeze. Pick a date in late October after which you stop changing listings, restructuring campaigns and testing creative. Amazon's systems take time to re-evaluate a listing, and the cost of a mistake during peak is far higher than the upside of an optimization.
Inventory timing detail is in our inventory reorder guide.
Common Strategy Failures
- Scaling spend on a listing that does not convert. The most expensive error available, and the most common. Fix the denominator before increasing the numerator.
- Optimizing ACoS instead of TACoS. Produces a beautifully efficient ad account attached to a shrinking business.
- Treating Amazon as the whole business. Every layer except owned audience improves your position inside a system you do not control.
- Ignoring product and packaging as marketing decisions. Half an inch off a package can beat a quarter of ad optimization.
- Launching too many SKUs. Five ASINs done properly outperform twenty done adequately, and a scattered catalog also weakens your brand's category signal to the AI layer.
- Discounting to fix a conversion problem. It works, temporarily, and it resets your price anchor permanently.
- Experimenting in Q4. The one quarter where the cost of being wrong exceeds the value of being right.
- Building for one algorithm. Traditional ranking and the AI layer run in parallel and read your listing differently. Optimizing for either alone forfeits the other.
The strategic question worth asking annually
If Amazon changed a policy tomorrow that cut our margin by ten points, what would we still have? The answer is your owned audience, your brand recognition and your off-Amazon channels — and if the honest answer is "nothing," that tells you which layer has been underfunded regardless of what your revenue looks like.
For the channel diversification view, see our omnichannel strategy guide and the broader 2026 ecommerce marketing strategies overview.
The Short Version
- Total FBA selling costs average 30 to 35% of revenue before advertising and approach 50% with it. Referral fees run 8 to 45% by category and a 3.5% fuel surcharge now applies to fulfillment.
- Five layers built in order: conversion foundation, on-Amazon paid, external traffic, owned audience, AI discovery. Each multiplies the one below it.
- Do not increase ad spend while unit session percentage is under 10%. Amazon's average listing converts at 10 to 15%, which is the bar.
- Manage to TACoS, not ACoS. Healthy is 8 to 15% for established products. ACoS ignores organic revenue and can look excellent while the business shrinks.
- Marketing as a percentage of revenue should fall as you scale. If it is not falling, advertising is substituting for organic strength rather than building it.
- Since May 13, 2026, Alexa for Shopping sits in the main search bar and runs in parallel with traditional ranking. You have to satisfy both systems.
- Q4 storage triples from $0.78 to $2.40 per cubic foot in October, which makes September the real deadline for shipping in and clearing slow movers.
- Product and packaging decisions are marketing decisions. Dropping a size tier can be worth more than a quarter of ad optimization.