A marketplace ad service starts to get risky the moment everyone agrees that “the agency can just optimise the campaigns.” It sounds efficient. It is also much too vague for an account spending €5K, €10K or €25K per month across Amazon, bol and MediaMarkt.
Optimise what, exactly? Bids? Budgets? Campaign structure? Search terms? Sponsored Products on bol? MediaMarkt retail media placements? Branded defence on Amazon? A coupon-supported push? A stockout recovery? A low-margin SKU that looks fantastic in ROAS but terrible in contribution margin?
The named mistake is the everybody-can-touch-spend problem. The website agency wants traffic for a product launch. The marketplace manager wants ranking. Finance wants margin protected. Operations wants to slow down because stock cover is thin. The ad service sees a campaign with good conversion and raises budget. Nobody is trying to be reckless. The problem is that no one has written down who is allowed to move spend when these signals disagree.
My stance: every serious marketplace ad service needs a decision firewall before the first bid changes. Not a heavy legal document. A simple operating layer that says which decisions the agency may take automatically, which decisions need commercial approval, and which decisions are blocked until SKU margin, stock and marketplace readiness are known.
This matters especially for brands in the Netherlands and Belgium spending from roughly €5K per month on Amazon Ads, bol Sponsored Products and MediaMarkt retail media. At that level, the expensive mistakes are rarely caused by one bad keyword. They are caused by good specialists making narrow decisions without shared permission.
What the current agency advice gets right
The market already gives decent advice on choosing advertising partners. Flywheel’s agency directory page for Houston Web Design and Hosting is not about marketplaces, but it shows a common buying pattern: one trusted partner handles web design, hosting, basic SEO, Google Ads, Facebook, local listings, social pages and email. For many businesses, that convenience is valuable. One team. One relationship. Less coordination.
Podean’s marketplace agency overview takes the opposite end of the market. It describes large Amazon and marketplace partners with retail media, content, analytics, global expansion, marketplace operations and proprietary tooling. Useful, because it reminds brands that Amazon is not just an ad account. It is a commercial operating system.
SalesDuo’s Amazon PPC agency guidance is strong on evaluation criteria: Amazon-specific experience, Seller Central and Vendor Central knowledge, transparent reporting, raw data, TACoS, search terms, placement performance, listing quality and inventory-aware scaling. Darkroom makes a similar point with more edge: many Amazon agencies are campaign managers with a retainer, while real Amazon marketing needs advertising, catalog, content, Brand Registry, DSP and storefront work to operate together.
Reddit and YouTube discussions add the seller’s reality check. Sellers ask which PPC agencies are worth trusting, whether to outsource or stay in-house, and what questions to ask before hiring. The repeated fear is not “will someone know where the bid button is?” The fear is “will an outsider spend my money without understanding my business?”
That is the gap. Competitor guides explain capabilities, lists and evaluation questions. They rarely give the practical firewall: who may move money, under which data conditions, with which stop rules?
The decision firewall in one sentence
A marketplace ad service decision firewall is a written permission model for spend movement. It connects every bid, budget and campaign action to four checks:
- Commercial permission: does the SKU have enough contribution margin after fees, fulfilment, returns, discounts and service cost?
- Operational permission: is there enough stock, Buy Box or offer strength, delivery reliability and content quality to convert the traffic?
- Evidence permission: do search terms, conversion rate, TACoS, incrementality or channel role justify the move?
- Decision permission: is this an automatic agency action, an operator review, a client approval or a hard block?
FiveX fits naturally here because the platform already connects marketplace analytics, advertising performance, profitability dashboards, inventory signals and AI recommendations. The point is not to create more meetings. The point is to make sure the ad service sees the same SKU economics that finance and operations see before money moves.
Why “full-service” is not the same as “full permission”
Full-service marketplace advertising can be excellent. It usually means the partner manages campaign structure, keyword harvesting, bid optimisation, budget pacing, weekly reporting and strategic recommendations. Some services also support listings, promotion planning, retail media negotiations and cross-marketplace allocation.
But full-service should never mean full permission. An agency can be responsible for managing ads without being allowed to make every commercial trade-off alone.
Here is the operator distinction I like:
- Optimisation decisions improve an existing campaign within an agreed lane. Example: reduce a non-converting search term bid by 20%.
- Allocation decisions move budget between SKUs, campaigns or marketplaces. Example: shift €1,500 from Amazon generic discovery to bol branded defence.
- Commercial decisions change the profit exposure of the business. Example: scale a discounted SKU with 14 days of stock and a break-even ACOS that changed after a fee increase.
A good ad service should own many optimisation decisions. It should recommend allocation decisions with evidence. It should not make commercial decisions without a pre-agreed permission rule.
The four firewall lanes
Use four lanes. Keep them boring. Boring is good when ad spend is involved.
Lane 1: Auto-execute
The agency may act without approval when the decision is reversible, low exposure and inside pre-approved thresholds. For example: bid changes within ±15%, adding negatives for clearly irrelevant terms, reallocating up to 5% inside the same campaign role, or pausing a target that spent more than its agreed learning cap.
Lane 2: Operator review
The agency may prepare the action, but a marketplace operator reviews it because the move affects stock, channel role or learning evidence. For example: increasing spend on a SKU with only 21 days of stock cover, restarting campaigns after a stockout, or moving budget from Amazon to bol because conversion quality shifted.
Lane 3: Commercial approval
The brand must approve the move because the decision changes margin exposure. Examples include launching a promotion-supported ad push, scaling a SKU below the margin floor, funding a MediaMarkt retail media package, or changing the budget split between marketplaces for the next month.
Lane 4: Hard block
No one moves spend until the missing condition is fixed. Typical hard blocks: unknown SKU contribution margin, stock cover below the agreed minimum, Buy Box or offer instability, missing product data, active listing suppression, unresolved fee change, or a return rate that makes the reported ROAS untrustworthy.
Named example 1: NorthPeak Home and the attractive Amazon campaign
NorthPeak Home spends €8,000 per month across Amazon.nl and bol. A Sponsored Products campaign for a compact air purifier shows a 19% ACOS and a 10.8% conversion rate. The agency wants to raise daily budget from €180 to €320 because the campaign keeps running out by 15:00.
Without a firewall, that looks like a sensible optimisation. With a firewall, the action moves to operator review. FiveX’s profitability dashboard shows the SKU has only 17% contribution margin after referral fees, FBA fulfilment, expected returns and a planned €6 coupon. Inventory insights show 11 days of FBA stock and a replenishment ETA in 16 days.
The decision changes. Amazon budget is lifted only from €180 to €220 for branded and high-converting non-branded terms. The remaining €2,400 monthly headroom is moved to two bol SKUs with 31% and 34% contribution margin and 42 days of stock. The agency still optimises. The firewall stops the account from buying demand it cannot serve profitably.
Named example 2: LumaPet and the “quick” bol launch push
LumaPet has €6,500 monthly marketplace spend and a web agency preparing a new category landing page. The agency asks the ad service to “quickly support the launch” with bol Sponsored Products and a small MediaMarkT placement for pet grooming accessories. Sensible request. Dangerous wording.
The decision firewall separates launch support into three SKUs. The grooming bundle has 42% contribution margin, 45 days of stock and a clear search-term match. It enters auto-execute for a €750 bol test. Replacement blades have 28% margin but only 18 days of stock, so they go to operator review with a €300 cap. Training mats have 12% margin, high return risk and only eight days of stock, so they are hard-blocked.
That one split prevents a common launch mistake: treating a category as one commercial unit. FiveX’s marketplace analytics make the SKU differences visible in the same view as ad spend and campaign role. The launch still happens, but €1,100 that would have gone to the weak SKU stays out of the auction.
Named example 3: VoltEdge and the MediaMarkt retail media package
VoltEdge sells electronics accessories and spends €15,000 per month across Amazon, bol and MediaMarkt. A MediaMarkt retail media package is offered for €5,000. The placement is attractive: high visibility around laptop accessories, good seasonal timing, and a product that performed well last quarter.
Then finance updates landed cost. The hero USB-C hub is now €4.20 more expensive because inbound freight and warranty reserves changed. Break-even ACOS drops from 18% to 12%. At the same time, Amazon branded defence is running at 9% ACOS and the bol accessory bundle has 38 days of stock with a 29% contribution margin.
The firewall sends the MediaMarkt package to commercial approval, not agency discretion. The decision is to fund only €2,000 of the package, demand a SKU-level placement split, and keep €3,000 available for Amazon and bol lanes with stronger profit permission. FiveX AI recommendations flag the cost change and attach it to the budget decision, so the discussion is not “do we like MediaMarkt?” It is “does this SKU still deserve that much paid visibility?”
The weekly firewall board
The decision firewall works best as a 30-minute weekly board. Not a long status call. A board.
- New spend moves: every planned bid, budget or marketplace allocation change above the auto-execute threshold.
- Blocked decisions: SKUs waiting for margin, stock, content, offer or finance evidence.
- Exception decisions: cost changes, return spikes, stockouts, Buy Box loss, fulfilment delays or CPC shocks.
- Released decisions: actions approved because evidence improved.
- Expired permissions: campaigns that need a fresh margin or stock check before continuing.
This is where an Advertentie Service becomes more valuable than a bid manager. The agency brings campaign evidence. The brand brings commercial constraints. FiveX brings the connected view: spend, revenue, margin, inventory, repricing context and recommendation history in one place.
What to put in the service agreement
You do not need a 40-page scope document. You need seven fields that remove ambiguity:
- Auto-execute limits: bid percentage, budget amount and learning-loss cap.
- Margin floors: break-even ACOS and minimum contribution margin by SKU or SKU group.
- Stock rules: minimum days of stock before scaling, plus restart rules after stockouts.
- Marketplace roles: which products are defend, grow, launch, learn, harvest or wait on Amazon, bol and MediaMarkt.
- Approval owners: who approves margin exceptions, cross-marketplace reallocations and retail media packages.
- Evidence windows: how long a test runs before the result is trusted.
- Stop-loss rules: when spend is reduced, isolated or frozen automatically.
If these fields are missing, the ad service will still make decisions. They will just be made through Slack pressure, dashboard anxiety and whoever happens to be most confident on the call. Funny in hindsight. Expensive in real life.
The better question before you hire
Do not only ask a marketplace advertising service: “Can you improve our ROAS?” Ask: “Which decisions are you allowed to make without us, and what data do you need before you make them?”
The right partner will welcome that question. It protects them too. A decision firewall gives the agency room to act quickly where speed helps, and it prevents them from becoming the accidental owner of margin, stock and channel strategy decisions that belong with the business.
That is the practical standard for €5K+ marketplace spend. Not more dashboards. Not more meetings. Clear permission. Better evidence. Faster decisions when the answer is obvious, and a firm stop when the account is about to buy growth the P&L cannot keep.