Four: a marketplace-only managed service, specialists hired per function, an internal marketplace lead, or a consultancy kept purely for strategy while execution sits elsewhere. Run a five-stage selection with a gate at each stage, and disqualify anyone whose product research amounts to a review count and a sales estimate.
The short version
- Generalist ecommerce and Amazon are different trades. Site conversion work does not transfer to a channel where you control neither the template nor the checkout.
- Run the selection as gates, not as meetings. Nobody advances until the previous gate is passed on evidence.
- Research depth is the cleanest disqualifier. Ask what they analyze besides review count and estimated monthly sales.
- Execution capacity is a separate question from strategy. Ask which verbs are in the scope of work.
- The free audit is your best comparison instrument. Two written diagnostics on the same account tell you more than four calls.
If you are here, you probably already have a deck
The usual situation looks like this. You engaged a broad ecommerce consultancy, the discovery work was thorough, the strategy document was intelligent, and six months later your listings look much the same. The recommendations were about brand positioning, channel mix, and lifetime value, none of which is wrong, and none of which changes your primary image or your backend keywords.
This happens because the two disciplines optimize different things. On your own site, the levers are template, funnel, email, and creative testing. On Amazon, the template is fixed, the checkout is not yours, the traffic is bought inside a closed auction, and the ranking system responds to velocity and conversion on specific search terms. A consultancy strong in the first world is not automatically weak, it is pointed somewhere else.
The five-stage selection, with a gate at each
- Stage one: define the constraint. Write one sentence naming what is actually limiting you: traffic, conversion, margin, supply, or catalog breadth. Gate: you can state it in a sentence without using the word growth.
- Stage two: request a written audit from each candidate. Give the same access to each. Gate: at least two documents come back that name specific ASINs and specific fixes rather than generic best practice.
- Stage three: interrogate the research. Ask what data they used to reach their conclusions. Gate: they cite something beyond review count and an estimated sales figure. Details below.
- Stage four: read the scope of work for verbs. Advise and recommend are strategy. Write, build, file, and ship are execution. Gate: the verbs match what you thought you were buying.
- Stage five: read the exit before the entry. Notice period, ownership of deliverables, account access, handover. Gate: you could leave in 30 days and still own everything you paid for.
Skipping a gate is the failure I see most often. Sellers meet three firms, like one, and sign. The gates exist so that the decision survives an impressive personality.
Stage three deserves its own section
Most weak recommendations trace back to shallow research. The common method is to look at review counts and an estimated monthly sales number from a browser plug-in, then declare a niche either crowded or open. Both conclusions are frequently wrong, because neither number describes profitability.
We run a category through more than 90 data points before anyone commits inventory. Market size and growth trajectory, so we know the ceiling. Return rate, because a category with structural returns eats margin invisibly. Segment dynamics, so we can tell a real subsegment from a keyword. The rating gap, which is where differentiation actually comes from: read the negative reviews of the top ten results, count what customers repeatedly complain about, and build to that instead of inventing a feature nobody asked for.
You do not need our specific list. You need to ask each candidate what theirs contains and to hear more than two items. Ask what they analyze besides reviews and volume, and listen for return rate and margin structure in the answer.
| Research input | Why it matters | Weak substitute you will hear |
|---|---|---|
| Category size and growth | Sets the ceiling on what you can earn | This niche looks hot |
| Return rate by segment | Silently determines net margin | Returns are manageable |
| Rating gap in the top ten | Tells you what to build and what to say | We will differentiate on quality |
| Segment dynamics | Separates a real subsegment from a keyword | There is low competition |
| Landed cost sensitivity | Decides whether advertising can ever pay | We will optimize the ads |
What most agencies will not tell you
Discovery is profitable and low risk to sell, which is why so many engagements are heavy at the front and thin afterwards. A four-week discovery phase produces a document, an invoice, and no change to your account. Ask what will be different inside Seller Central in week three, and get the answer in writing.
The second thing: some of the work you are being sold as strategy is a two-hour task. Choosing between two price points, rewriting five bullets, or fixing a variation family does not need a workshop. If your constraint is executional, buy execution and keep your strategy budget for the decision that actually needs it.
Related answers
- How to choose an Amazon FBA marketing partner
- Alternatives to big-name Amazon aggregators’ advisors
- What does a good Amazon account audit include
- Amazon agency vs in-house team pros and cons
- Done-for-you Amazon management: the complete guide
Bring the same account to two audits, including ours at Flapen, and compare the documents.
