Amazon agency or in-house? When your own team stops paying off

Put a marketplace specialist's salary next to an agency invoice and you are comparing two different things. On one side sits a single line in the payroll budget. On the other sits a team, a set of processes, a tool stack and experience from many accounts. In most cases an agency wins not because it is cheap, but because you are buying a capability that already exists instead of building one from scratch. An in-house team starts to pay off at the point where your scale supports not one person, but a team with real cover when somebody is away.

What follows is a framework you can run on your own numbers: every component of the full cost of an internal team, three questions that usually settle the decision faster than a spreadsheet, and a comparison of three operating models. You will not find invented "average market rates" here, because salaries in your country are something you know better than we do. What we add is the structure, and the lines that quietly go missing from these calculations.

Why the salary versus invoice comparison misleads

The conversation usually goes like this. Somebody quotes the salary a specialist expects, somebody else quotes an agency's monthly fee, and the two numbers end up side by side in a spreadsheet. The problem is that the left column holds one person's pay while the right column holds an entire function: account management, listings, campaigns, pricing, incident handling, reporting, and cover when someone is ill.

For the comparison to mean anything, both sides have to cover the same ground. On the in-house side that means adding everything needed for one person to do the work at all, and for the business not to stop when that person is unavailable. Only then are you comparing like with like against the price of a supplier who delivers the whole function.

The second mistake is subtler. Employment cost is fixed and rigid, while marketplace revenue is variable and seasonal, often with a disproportionate share landing in the Q4 peak around Black Friday. When you hire, you buy a cost that stays with you through the weak quarters too, and one you cannot reduce from one day to the next.

The full cost of an in-house team: what to actually count

Below are the lines to put in the spreadsheet. Use your own figures, because they depend on the country, the local labour market and the seniority you are hiring for. The point of the list is that no line disappears quietly.

Costs you can see straight away

• Gross salary. Your assumption, and the starting point for everything else.
• Employer social contributions. Every European country adds employer-side charges on top of gross pay, and the rates differ enough that no article can do this part for you. For scale: across the EU, non-wage costs (mainly employers' social contributions) made up 24.8% of total labour costs in 2025, 25.6% in the euro area and 32.3% in France, the highest in the Union (Eurostat, data extracted 31 March 2026). Read that carefully: it is a share of the total cost, not a markup on gross pay, so expressed as an uplift on the wage itself it is higher again. Take your actual number from payroll.
• Pension schemes and benefits. Several countries run occupational pension arrangements the employer co-funds, and on top of that come the benefits that are effectively standard in your market: private healthcare, a phone, a car. Ordinary lines, routinely left out of this particular comparison.
• Tools and subscriptions. Research and analytics, listing and price monitoring, a repricer, design, translations. You buy them for one account; an agency spreads the same licences across many clients.
• The workstation. Hardware, licences, desk space, onboarding into your internal systems.

Costs that decide whether the calculation is honest

• Recruitment. Your time and a manager's, job ads, a recruiter's fee where you use one. Plus the risk of a bad hire, which means running the whole process again.
• Ramp-up. The months in which the person costs full rate and is not yet delivering a result.
• Real availability. Start from the working days in your country's calendar, then subtract. EU law guarantees every worker at least four weeks of paid annual leave (Article 7 of Directive 2003/88/EC), national law and collective agreements often give more, and then come public holidays and sick leave, part of which the employer funds directly in most member states. What is left at the end, not twelve months, is the denominator you divide the annual cost by to get the cost of one working day.
• Management. Your time or a manager's, spent setting priorities, checking quality and developing the person. Those are real hours, and they are priced higher than the ones being managed.
• Keeping the knowledge current. Training, conferences, time spent tracking platform changes. Not a one-off, as the next section explains.
• Mistakes and downtime. A deactivated listing, a burnt advertising budget, a pricing error, a fall in account health. This is the hardest line to estimate and usually the only one capable of outweighing all the others combined.
• Single-person risk. Statutory notice periods are set nationally, are usually measured in weeks or months, and in many countries lengthen with service. In practice it means you learn that the only person who knows the account is leaving a few weeks before it happens.

Add those lines up, divide by the days genuinely available, and you get the cost of one working day from your own specialist. That is the figure worth putting next to an agency quote, rather than the gross salary on its own. If you are doing the same exercise on the product side, the rule is identical to the one that applies when you calculate margin and profitability: a calculation is worth exactly as much as the lines it does not omit.

Three questions that settle it faster than a spreadsheet

The spreadsheet is necessary, but it rarely closes the decision on its own. In practice three questions do.

1. How many channels and markets are on the plan for the next two years?

If the answer is "Amazon in one country and nothing else", one competent person may well be enough. If the plan includes Germany, a second platform and another product category, then in an in-house model every new channel means either another hire or overloading the person you already have. In an agency model it is a change of scope in a contract rather than a new position on the payroll. That is the most common reason brands end up running several platforms through one supplier: one strategy and one report, instead of a handful of independent islands.

2. How much time and how much revenue can you spend getting to competence?

We will not give you a universal number of months, because honestly we do not have one. It depends on the product, the starting point and how much of your own time goes into onboarding. That is a number you have to estimate yourself, before you sign an employment contract rather than after. If you decide to keep the work inside the company anyway, there is a middle option between a hire and full outsourcing: buying expert time by the hour, in the form of marketplace consulting.

What is worth remembering is that marketplace competence is not a one-off cost. The platforms change the rules mid-game: Amazon FBA fees alone can change more than once a year, and new programmes, new regulatory duties and new advertising formats keep arriving. When you hire, you buy not only the learning but the obligation to keep refreshing it.

3. What happens when that one person leaves?

This question reveals the most. If the answer is "we stop", what you have is not a team but a single point of failure. What walks out with that person is the knowledge of the account, the history behind past decisions, the campaign settings and the context that exists in no document anywhere. Agencies have churn too, but on their side more than one person knows the account, and handover is a process rather than an improvisation.

In-house, agency, hybrid: a comparison

The third option, the one that usually goes missing from these comparisons, is the hybrid model: the agency does the work, and on the brand side one person owns the channel, knows the numbers and makes the commercial decisions.

A practical conclusion, and this is our recommendation rather than a fact: for most brands selling on one or two platforms and wanting to grow, the hybrid model is safer than pure in-house and more effective than pure outsourcing with nobody owning the channel internally.

When in-house genuinely wins

There are situations in which your own team is the better choice, and pretending otherwise would be dishonest. The three most common:

• Scale justifies a team, not a headcount. Once sales volume supports several people with separated roles (listings, advertising, operations) and genuine cover for each other, the single-person risk argument stops working and the unit cost of the work falls.
• The product needs deep, permanent internal knowledge. Configurable, heavily regulated or technical ranges, where a correct product page requires engineering or certification knowledge that only you hold anyway.
• You are deliberately building the capability as an asset. For example while preparing the company for a transaction, or moving to a model in which direct sales is meant to be a permanent core of the business rather than a supplementary channel. That is a strategic decision, and there the cost of learning is an investment rather than a loss.

It is only fair to say what an agency does not solve on its own either. Part of the knowledge about your account sits outside your company, you share priorities with other clients, and changing supplier always costs time. Those risks can be contained in the contract, and it is worth doing that before you sign:

• Ownership of the account and the brand. The seller account and the trademark belong to you; the agency works on access you grant.
• Access on your side. Full visibility in Seller Central, the advertising console and the reports, for the whole time you work together.
• Reporting in your metrics. Sales, margin and advertising cost, not a count of tasks completed.
• A sensible notice period and a defined handover. Agreed at the start, not at the moment of parting.

At scale the bottleneck stops being hands

There is one more reason adding headcount stops working. Above a certain number of products, channels and markets the problem is not a shortage of hands but a shortage of integration: stock levels drifting apart between channels, catalogue translations done by hand, feed mapping for every new platform, listing errors that have to be caught before the platform catches them. On Amazon the first symptom is usually financial rather than operational, showing up in storage fees and capacity limits. None of that is solved by another person. It is solved by technology.

Within our group that layer is handled by our sister technology brand Seedlight, for instance through AI automation in eCommerce operations. We point it out not to close the argument with an advert, but because it is a real alternative to the decision to "hire one more person for repetitive work".

How this looks on our side

At Amazonway we run clients' marketplace sales through the WAYPOINT framework, a route in five points. The first of them, WP 01 Direction, is a free consultation and audit: the account, the product data, the competition and the margins. It ends with a diagnosis of the blockers rather than a contract, which makes it a reasonable way to test your own cost assumptions against hard data before you commit to an operating model.

Day-to-day management (WP 04 Momentum) is billed as a retainer plus a success fee on the sales generated, so part of what we earn depends on your result. The ranges and the billing rules are published openly on our pricing page rather than quoted case by case.

The caveat you deserve: not every brand should hand this work outside, and our way of working will not suit everyone. If the audit shows that your scale justifies an internal team, we will say so. We do not promise specific sales results either, because those depend on the product, the price, the competition and the budget, not on account management alone.

Frequently asked questions

Is an agency cheaper than hiring a specialist?

It depends what you include on the employment side. Compare a gross salary with a monthly retainer and the agency usually looks more expensive. Add employer contributions, pension and benefit costs, tools, recruitment, ramp-up, paid leave and sick days, and the proportions often reverse, particularly where one person is expected to cover several channels. The sensible way to judge a quote is against your own spreadsheet, once both sides cover the same scope.

How many people does it take to run an Amazon account in-house?

There is no single number, because the scope depends on how many products and markets you have and whether you run advertising. The more useful question is the reverse one: how many people have to know the account for a holiday or a resignation not to stop sales? If the answer is "more than one" and the budget stretches to one position, the in-house model is underinsured.

Who owns the account and the data when an agency does the work?

You should. The standard worth insisting on: the seller account and the trademark registered to your company, the agency working on access you grant, and permanent visibility for you in Seller Central and in the reports. A supplier proposing anything else is a warning sign regardless of price.

Can you start with an agency and move in-house later?

Yes, and it is a common path. In that case it is worth working in the hybrid model from the start: one person on your side owning the channel, learning on an account that already works rather than on their own mistakes. The move then becomes a handover instead of a start from zero.

Does the agency cost fall when sales fall?

Partly, if the billing model contains a component tied to results. In a retainer plus success fee model the fixed part stays and the variable part falls with turnover. Employment cost in the same situation does not move at all.

Test your assumptions against real data

Before you decide on an operating model, price both sides over the same scope: the full cost of an internal team using the list above, and the cost of an agency for the same body of work. If you want real numbers from your own account on the other side of that calculation, start with an audit.

Book a free consultation and account audit →