A workable in-house stack is four layers: Seller Central plus a market and keyword data tool, an advertising console or bid tool, a creative and listing production pipeline, and a reporting layer someone reads every week. Subscriptions cost hundreds per month. The operator running them costs far more.
The short version
- Four layers, not twenty subscriptions. Data, advertising, creative production, reporting. Everything else stays optional until those four have a named owner.
- Software is the cheap half of the invoice. Licenses are predictable. Operator hours are not, and they dominate the budget from month one.
- Capacity decides this, not features. 50 operators here cover about 70 brands, close to 1.4 brands each, and that is with dedicated sourcing and creative studios doing production behind them.
- Buy the data layer before the automation layer. Automated bidding on a badly chosen catalog loses money faster.
- Every report needs a named reader and a fixed day. Otherwise you are paying a subscription for charts nobody opens.
The four layers and what each has to earn
| Layer | What it must do | What actually drives your cost | What fails without it |
|---|---|---|---|
| Market and keyword data | Size a category, map demand, read the rating gap, watch share | Seats, number of marketplaces, refresh depth | You launch into categories too small to repay their own acquisition cost |
| Advertising | Build campaigns, set bids, read placement and search term reports | Whether a human sets targets or a static rule set does | Spend drifts upward while targets never change |
| Creative and listing production | Images, A+ modules, copy, video, translations | Whether design and copy sit inside the team or get bought per asset | Traffic lands on a page that does not convert |
| Reporting | One weekly written view of profit, ACoS, conversion, rating, inventory | Who compiles it and how long that takes them | Decisions get made from memory and last week's mood |
Notice that three of the four cost columns describe people, not software. That is the whole finding. A tool license is a rounding error against a salary, so the interesting question is never which platform to buy. It is how many hours per brand per week the work takes, and whether you have those hours.
The line item nobody budgets
Brands hand us their account and get about two hours of their own time back per month once onboarding finishes. During an active product launch, that rises to four to six hours a week. Running the same work in-house does not remove those hours. It moves them onto your payroll, and it adds the hours the agency was absorbing quietly: search term harvesting, creative revisions, case work with Seller Support, translation checks, inventory forecasting.
Our internal ratio is the cleanest capacity signal I can give you. About 1.4 brands per operator, and that operator is supported by an in-house sourcing studio in Guangzhou, an in-house creative studio in Dubai, and a technology team building our advertising, marketing, and brand valuation tools. Strip that support away and a single unsupported hire carrying one brand is at a reasonable load, not an easy one. If your plan says one person will handle three brands with off-the-shelf software, the plan is a hiring forecast dressed as a budget.
Running the arithmetic
Work through these five numbers in order before you buy anything.
- Advertising floor. There is no hard minimum spend, but below about $1,000 a month there is not enough data for meaningful optimization, so a bid tool has nothing to learn from.
- Launch capital. A single product launch typically needs $8,000 to $15,000 in total upfront capital. A five-product brand runs $25,000 to $50,000. Tooling competes with inventory for the same cash.
- Salary, loaded. Take your operator's full cost, add recruitment and the ramp period, then divide by the number of brands or product lines they will carry.
- Production. Every image set, A+ module, and translation is either a salary line, a freelancer invoice, or an agency inclusion. Price it as one of the three, never as zero.
- The managed comparison. Our flat fee runs $800 a month for one product to $2,400 for five, with all 50+ services at every tier, no commission and no revenue share. Whatever your in-house total comes to, compare it against a real quote rather than an assumed one.
Most in-house budgets I have reviewed get numbers one, two, and five right and skip three and four entirely. That is why the build case looks cheap on the spreadsheet and expensive by month four.
The build case that actually works
In-house wins when your catalog is deep enough to keep a specialist busy every day, when Amazon is your primary channel rather than one of five, and when someone senior enjoys the work enough to stay. It also wins when your product knowledge is the differentiator, because no external team will learn a technical catalog faster than the person who built it.
It loses when the role is a fraction of somebody's job. A marketing generalist given Amazon as a Friday task will keep the account alive and will not grow it, and you will not find out for two quarters.
What a software vendor will not tell you
Platforms are sold on the promise that the tool does the work. It does not. It shortens the work. Bid automation still needs a human deciding what the target should be this month for this product. Keyword tooling still needs somebody choosing which terms deserve budget and which are vanity. Listing graders still need a writer.
The second thing rarely said out loud: seat pricing quietly encourages you to give one person too many brands, because adding brands to an existing seat is cheap and adding a person is not. That is exactly backwards from how the work scales. Decide headcount from workload first, then buy the seats that headcount needs.
Related answers
- Amazon agency vs in-house team pros and cons
- Which roles to hire first for an in-house Amazon team
- Cost comparison: Amazon freelancer, agency, software
- Top ranked Amazon agencies vs DIY software stacks
- Build vs buy for your Amazon channel: the complete guide
If you want the managed side priced against your own stack, every tier is published at Flapen.
