How to choose an Amazon agency: a seller’s checklist

Before you talk to anyone: what are you actually buying?

Most bad agency engagements start before the first call, because the seller has not decided what problem they are paying to solve. "Grow our Amazon business" is not a brief, and an agency will happily sell you a retainer against it.

There are really only four problems, and they need different suppliers.

• Nobody finds us. A visibility problem. You need keyword and listing work, then advertising.
• People find us and do not buy. A conversion problem. You need listing, imagery and pricing work, and advertising will make it worse until that is fixed.
• We sell fine but make nothing. A margin problem. You need ad efficiency, fee recovery and often a hard look at which ASINs deserve to exist.
• Something is broken. Suppression, suspension, a hijacked listing, a stuck case. You need a specialist for a defined piece of work, not a monthly retainer.

Write down which of those you have, in one sentence, before the first call. If an agency cannot tell you which one they think you have after twenty minutes of looking, they are selling capacity rather than judgement.

Two things worth settling at the same time, because they change who you should even be talking to. First, what would success look like in ninety days, expressed as a number you would recognise — organic rank on six terms, a total advertising cost of sale below a figure, a conversion rate on your top three ASINs. Vague goals produce vague retainers. Second, what you are not willing to do: some brands will not discount, some will not run coupons, some have a price floor set by their wholesale channel. Say so early, because half of the standard Amazon playbook assumes otherwise, and an agency that only discovers your constraints in month two will have spent month one planning around them.

The twelve questions, and what a bad answer sounds like

In rough priority order. If you only get through the first five, you will still learn most of what you need.

1. How many accounts does the person doing my work carry?

The single most predictive question in the list, and the one most likely to be deflected. You want a number and a name. Somewhere between four and eight accounts is a person who can think about yours; fifteen is a person processing a queue.

A bad answer: "We work as a pod" or "it varies by client". Both are true statements that avoid the number. Ask again.

2. Who will actually be in the account, and are they employed by you?

Plenty of agencies sell a senior strategist and staff the work elsewhere. That is not automatically wrong — it is wrong when it is concealed. Ask where the team sits, what hours they work relative to yours, and whether you will speak to them or only to an account manager who relays.

A bad answer: any answer that describes the team as "our global delivery network" without naming a country or an employment relationship.

3. Show me a month where it went badly.

Every account has one. An agency that cannot produce one is either new, or curating. What you are listening for is whether they noticed it before the client did, what they changed, and whether they say what it cost. This question is worth more than any case study.

A bad answer: a reframed win. "We had a slow month while we rebuilt the campaigns, and then sales tripled."

4. What happens if we miss the agreed benchmark?

Not "do you guarantee results" — nobody credible does, because the platform is outside their control. The question is whether missing a target triggers anything at all: a review, a fee adjustment, an exit without penalty. If the answer is nothing, the benchmark is decoration.

A bad answer: "We're confident that won't happen."

5. Who owns the listings, creative and ad account if I leave?

This has to be in writing. Some agencies build A+ content and imagery in their own accounts and take it with them. Some run advertising through their own ad console rather than yours. Both are recoverable and both are miserable to discover in month thirteen.

A bad answer: "That's never been an issue."

6. What is your notice period, and does the contract auto-renew?

An initial term is reasonable — Amazon work takes a quarter to show. A twelve-month auto-renewing term with ninety days' notice is a different proposition, and it is usually buried rather than pitched.

7. What is the pricing model, and what changes it?

Flat retainer, percentage of ad spend, percentage of revenue, or a hybrid. Percentage of ad spend rewards an agency for spending more of your money, which is not a reason to reject it but is a reason to cap it. Percentage of revenue can be very expensive on a brand that was already growing.

A bad answer: a price that appears only after a discovery call. You are allowed to know the range before you invest an hour.

8. What do you do that risks my account?

Ask it exactly that way and watch the pause. You are checking for incentivised reviews, review gating, giveaway-driven rank manipulation, and any "guaranteed ranking" mechanism. All of these work briefly and all of them put the account at risk, and the person who loses the account is you, not them.

A bad answer: "We use proprietary ranking methods."

9. What are you going to do in the first thirty days?

A good answer is boring and specific: audit, catalogue clean-up, campaign restructure, listing priority order, baseline metrics agreed. A vague answer here usually means the first month will be spent deciding what to do, at your expense.

10. How do you report, and will I understand it?

Ask for a real, redacted report from an existing client. If it is forty slides of platform screenshots with no argument in it, that is what you will get monthly. The useful report is short, says what changed and why, and names what did not work.

11. What do you need from me?

Agencies that have done this before will have a clear answer — margin data, stock visibility, a named decision-maker, turnaround on creative approvals. An agency that says "nothing, we handle everything" is either overselling or about to make pricing decisions without knowing your cost of goods.

12. Why would you turn this account down?

The best closing question. Anyone with real experience has a list: no stock, no margin, no brand registry, a category they do not know, a founder who wants daily calls. An agency that cannot name a single reason it would decline work is telling you how it selects clients.

How to read the evidence they give you

Every agency arrives with proof. Most of it is constructed to be unfalsifiable, and knowing how tells you what to ask next.

Case studies

Look for three things: a named client or a named category, a date range, and a starting point. "Grew revenue 340%" from a base of $4,000 a month is a different achievement to the same percentage from $400,000, and case studies that omit the base are omitting it deliberately. A dated engagement you can place in time also tells you whether the tactics still apply — a 2023 result predates several material Amazon changes, including the July 2026 title limit.

The most useful follow-up is: "what was the second-best month on that account, and what happened in it?" Anyone can produce one extraordinary month. The shape of the ordinary ones is the actual service.

Review scores

Clutch, DesignRush, Trustpilot and G2 all carry agency profiles, and the number beside the star rating matters more than the rating. A 4.9 from four reviews and a 4.9 from forty are not comparable claims. Check the service mix on the profile too: several agencies rank well on directories under categories that have little to do with Amazon, so the reviews you are reading may be for different work entirely.

Partner badges

Amazon publishes only two advertising partner statuses, verified and advanced, awarded on product adoption, certification and partner-led investment. Amazon does not publish a spend or client-count threshold for either. Anything described as "certified", "elite", "top 5%" or "premier" is either a different programme or the agency’s own framing — worth asking which, without treating it as a disqualification.

Client lists

A logo wall proves an account existed, not that it still does or that the agency did the work that made it notable. Ask when each engagement started and whether it is current. Two current clients you can call beats twenty logos.

The contract clauses nobody reads

The pitch is the marketing. The contract is the product. Five clauses do most of the damage.

None of these need a lawyer to spot. They need you to ask for the contract before the call where you are expected to sign, and to read the five clauses above first.

What should this cost?

Published pricing in this industry is rare, which is itself informative. Of the agencies we checked while researching this, only two publish figures at all on their own sites: eStore Factory runs a full rate card with PPC packages from $499 to $1,833 a month on three-month terms, and SellerPlex publishes "starting at $2,500 a month plus 5% of incremental sales above baseline". Everyone else quotes after a call.

That makes benchmarking hard, so use structure rather than a number. Ask which of these you are being sold:

• Flat retainer. Predictable, easiest to budget, and the model where you most need to check what hours or deliverables it actually buys.
• Percentage of ad spend. Typically the cheapest headline and the one with the worst incentive alignment. Cap it, or agree the fee against a spend band rather than a raw percentage.
• Percentage of revenue. Aligns interests in principle. In practice it charges you for growth that would have happened anyway, so agree a baseline and pay on the increment.
• Project fee. The right model for a fix — a listing rebuild, a reinstatement, a catalogue migration. If your problem has an end state, do not buy a retainer.

Our own pricing is published, tier by tier, for the same reason we are telling you to ask for it: an agency that will not name a range before a call has made the first hour of your evaluation expensive on purpose.

Red flags

Ranked by how much damage they do rather than how bad they sound.

Guaranteed rankings or guaranteed sales. Nobody controls Amazon’s results. A guarantee means either a meaningless definition of the promise, or a mechanism that puts your account at risk. Ask which.

Reviews as a service. Any offer to supply reviews, source reviewers, or incentivise positive feedback. Amazon suspends sellers for this, not the agency, and in the US the Federal Trade Commission’s rule on consumer reviews makes buying and selling reviews a civil-penalty matter. This is the fastest way to lose an account you spent years building.

No named person on the account. If nobody’s name is attached, nobody is accountable, and your work will be whatever is left after the accounts with named owners are served.

Reporting that only goes up. Every account has bad months. A supplier whose reporting never shows one is either selecting metrics or not looking.

Pressure to sign inside the call. Discounts that expire at the end of a meeting are a sales technique, not a commercial reality. A good agency is comfortable with you reading the contract.

They never say no. If every idea you float is a good idea, you are buying agreement. You can get that cheaper.

Should you hire an agency at all?

Three alternatives, and the honest case for each.

A freelancer costs less and can be excellent, particularly for one discipline — a PPC specialist, a copywriter, a designer. The risks are bus factor and breadth: one person on holiday is your whole Amazon function on holiday, and few freelancers cover advertising, listings, catalogue and account health at once. Good fit under roughly $50k a month of revenue, or for a single well-defined discipline.

In-house is the right end state for most brands that get big enough. A competent Amazon manager costs more in salary than most retainers, plus tooling, plus the months before they are effective, plus the risk that they leave and take the institutional knowledge with them. The usual sequence is agency first, in-house once the playbook is stable, agency retained for specialist work.

Nothing is a real option more often than agencies admit. If your listings have not been rewritten since the 75-character title change in July 2026, if your images are phone photographs, or if you have no Brand Registry, a month of your own attention will beat a month of anyone’s retainer. Fix the free things first.

If you do go to market, brief three agencies on the same one-sentence problem and compare the answers, not the decks. The one that tells you something uncomfortable about your account in the first call is usually the one that will be useful in month six. Ours is a written audit back within 48 hours, and it is deliberately the same document whether or not you hire us — partly because it is the fairest way to be compared, and partly because it is the only way to answer question three in this checklist about ourselves.

Questions people also ask