Marketplace ads usually get judged too early.
A shopper clicks an Amazon Sponsored Product today. bol.com attributes the sale tomorrow. MediaMarkt reports revenue in the retailer dashboard. The weekly ad report shows ACOS, ROAS, spend, clicks, conversion rate and maybe TACoS. The campaign looks clean enough to keep running. Then, two weeks later, the product comes back with “does not fit”, “arrived damaged”, “quality lower than expected”, “battery life too short” or “wrong variant”. By then the bid decision has already moved on with its life.
The named mistake I see in managed marketplace advertising is optimising before the customer has finished voting. The ad account says a keyword worked. The return reason says the wrong promise was sold. The campaign manager raises the bid because conversion looked healthy. Customer support sees five complaints about the same accessory clip snapping. Finance reserves refunds, but the ad budget still treats last week’s revenue as earned.
My stance: every €5K+ marketplace ad service needs a return-reason loop. Not a generic return-rate KPI. A weekly operating loop that connects returns, refund lag, defect notes, review movement and support tickets back to SKU-level ad permission on Amazon, bol.com and MediaMarkt.
FiveX helps here because marketplace advertising, contribution margin, product profitability, inventory insights and change logs sit in the same operating view. The point is not to punish every SKU that receives a return. Returns happen. The point is to stop buying more traffic for a promise the customer is already correcting.
What the current advertising advice gets right
The research landscape is useful, especially around ad mechanics.
BidX explains the machinery well: automation can manage keywords, bids and budgets, move search terms into manual control, add negatives and improve budget efficiency. Their dynamic bidding analysis also makes a point operators should remember: Amazon can raise or lower bids in real time based on purchase probability, but bid strategy still depends on product, margin and competitive situation.
Podean, from the agency side, is right that Amazon success is not just advertising. Their agency hiring guidance points out that Amazon specialists often sit closer to retail operations than classic media agencies do. Inventory, replenishment, product pages, chargebacks, creative and ad spend all collide. That is exactly why an ad service cannot live inside the ad console alone.
Feedvisor’s incrementality work adds another warning: the campaigns with the prettiest ROAS can be the least incremental. If attribution can overclaim revenue before returns, it can overclaim profit after returns.
AdBadger and SellerMetrics both keep ACOS honest: there is no universal “good” ACOS, and a 20% ACOS can lose money if the product only has 18% margin. Seller discussions on Reddit and Amazon forums make the same point less politely. Sellers ask how PPC can ever scale when marketplace fees, shipping, product cost and ads leave only a few euros of room. Others are told to fix reviews, price position and listing trust before feeding more clicks into a weak page.
Useful, but still missing one vote: post-purchase evidence.
What competitors usually miss: the sale is not finished at conversion
Marketplace ad reports are biased toward the moment of purchase. That makes sense technically. Ads need attribution windows. Operators need weekly decisions. Retailers need campaign dashboards. But profit does not stop at checkout.
A sale can lose commercial quality after the click because:
- the product is returned after the ad report closed;
- the return reason reveals a listing promise problem;
- a review mentions the exact feature used in ad copy;
- support tickets show confusion about compatibility, size, colour or warranty;
- refund lag hides true contribution margin for two or three weeks;
- the returned unit cannot be resold as new;
- the same issue repeats on Amazon, bol.com and MediaMarkt under different labels.
The return-rate tile is too blunt for this. “Return rate 9.8%” does not tell the ad operator whether to pause generic discovery, keep branded defence, fix the PDP or stop a MediaMarkt renewal.
A return-reason loop turns messy post-purchase evidence into advertising decisions. It asks: which demand did we buy, what promise did that demand believe, and did the product keep that promise after delivery?
The five signals in a return-reason loop
For Amazon, bol.com and MediaMarkt accounts from roughly €5K monthly spend, I would not make this complicated. Start with five signals.
1. Return reason by SKU and campaign role
Do not only review returns by SKU. Attach them to the campaign role that bought the demand: branded defence, generic harvest, competitor conquest, launch learning, promo support, clearance or retail media package traffic.
A 7% return rate on branded defence may be normal for the category. A 7% return rate on a generic launch campaign might be a warning that the ad is attracting the wrong intent. A 7% return rate on a low-margin clearance SKU may destroy the economics entirely.
2. Reason severity
Not all return reasons deserve the same ad response.
- Fit or preference: usually a measurement, size-guide or targeting issue.
- Wrong item or wrong variant: catalog, feed, fulfilment or content issue.
- Damaged on arrival: packaging, carrier or fulfilment issue.
- Quality below expectation: product promise or product quality issue.
- Compatibility confusion: targeting and content issue, especially in electronics and accessories.
FiveX product profitability views are useful here because a return reason only matters commercially when you see the refund, handling cost, resale value and ad cost beside the SKU’s contribution margin.
3. Refund lag
If the category has a 14- to 30-day return window, last week’s ACOS is not final. That does not mean you freeze everything. It means you assign evidence maturity.
I like four labels:
- Fresh: 0-7 days after sale, too early to trust return-adjusted profit.
- Settling: 8-21 days, early returns and complaints are visible.
- Mature: 22-35 days, most return lag has spoken.
- Closed: after the category’s normal return curve plus finance adjustments.
Automation should treat Fresh evidence differently from Mature evidence. A Fresh 19% ACOS on a SKU with late-return history should not scale like a Mature 19% ACOS with low refunds.
4. Review and support-ticket echo
Returns become much more serious when they match public reviews or support tickets. One “too small” return can be noise. Three returns, two tickets and a new one-star review mentioning the same sizing issue is an ad veto waiting to happen.
This is where an operator earns their fee: spotting the repeated customer sentence before the average becomes obvious.
5. Cross-marketplace repeat
The same issue often wears different clothes by marketplace. Amazon might show “not as described”. bol.com might show a buyer message about missing accessories. MediaMarkt might report lower conversion on the same bundle because the compatibility promise is unclear.
FiveX helps by pulling marketplace, advertising and operational signals into one view. If the same SKU family is creating post-purchase friction across channels, the decision is not “lower Amazon bids by 8%”. The decision is “stop buying generic traffic until the offer promise is fixed everywhere”.
The operating rule: returns do not pause ads, patterns do
A return-reason loop should not make the team nervous about every refund. That would be silly, and also impossible. The rule is simple:
Individual returns create notes. Repeated return reasons create bid permission changes.
I use four decision labels.
Keep
The return reason is within category norm, margin still survives, reviews are stable, and the campaign role is valid. No ad change needed. Keep watching.
Shrink
The pattern is real but not yet fatal. Lower bids, cap budgets, narrow targeting or move the SKU from SCALE to PROTECT. Keep branded defence if customers are already looking for the product, but stop paying aggressively for cold generic demand.
Quarantine
The ad evidence is no longer clean. Do not scale from it. Hold new budget, separate the SKU from broad campaigns, and require a content, fulfilment or product fix before spend can grow again.
Stop
The campaign is buying demand the product cannot profitably satisfy. Pause the role that creates the bad demand. This is especially important for competitor conquest, generic broad match, Sponsored Brands Video and fixed retail media packages where spend can keep moving even while the SKU’s promise is breaking.
Notice the trade-off. You may give up short-term revenue. Good. Revenue that returns with a complaint is not growth; it is expensive customer research.
Named example 1: NovaBrew’s bol.com milk frother
NovaBrew, a fictional coffee-accessory brand, spends €6,200 per month across Amazon.nl, bol.com and MediaMarkt. Its bol.com Sponsored Products campaign for a compact milk frother looks promising: €1,180 spend, €5,900 attributed revenue, 20% ACOS and 68 units sold in two weeks.
The old decision would be easy: raise budget from €55 to €85 per day and add more generic terms like “melkopschuimer elektrisch”.
The return-reason loop says wait.
After 18 days, 9 of the 68 units have return or complaint signals. Four mention “too weak for oat milk”, three mention “battery cover loose”, and two support tickets ask whether the product is rechargeable even though the listing says battery-operated. The SKU sells for €34.95. After bol fees, fulfilment, product cost and expected handling, NovaBrew has €9.40 contribution margin before ads. Each attributed ad sale paid about €17.35 in ad cost. The reported campaign was already relying on repeat and organic halo to justify spend. Returns now remove the remaining comfort.
The verdict: Quarantine generic traffic. Keep branded defence at €12 per day because shoppers searching “NovaBrew frother” are close to purchase. Cut generic bids by 35%, pause “oat milk frother” until the content is honest about use cases, and create a product ticket for the battery cover. FiveX’s change log records the bid cut, the return reasons and the reopen condition: fewer than 6% issue-related returns over 35 mature orders after content correction.
Named example 2: ArcLight’s Amazon USB-C hub
ArcLight sells electronics accessories and spends €9,500 per month on Amazon.de, bol.com and MediaMarkt. An Amazon Sponsored Products exact campaign for “usb c hub hdmi macbook” reports €2,400 spend, €11,760 attributed sales and 20.4% ACOS. On paper, it deserves more budget.
The SKU economics are tighter than the dashboard suggests. The hub sells for €49.90. Referral fee, FBA, product cost, warranty reserve and expected returns leave €13.20 contribution margin before ads. Average CPC is €0.92 and conversion rate is 8.1%, so the ad cost per order is roughly €11.36. That leaves only €1.84 before delayed refunds. Not a disaster, but not a victory parade either.
The return loop adds the missing evidence. Over the next three weeks, 14 of 122 ad-attributed orders are returned. Seven mention “does not work with two monitors”, three mention “gets hot”, and four are generic “not compatible”. Customer support has six tickets with the same dual-monitor question. The product detail page states “4K HDMI” but does not make the single-display limitation obvious above the fold.
The verdict: Shrink and fix. The campaign is not paused because the intent is commercially valuable, but budget drops from €120 to €70 per day, the phrase “dual monitor” is added as a negative, and the product page gets a compatibility block before the first image carousel. FiveX inventory insights also show 19 days of stock, so scaling would have created a stockout risk even if returns were clean. Reopen condition: mature return rate below 8%, support tickets under two per 50 orders, stock above 28 days and contribution after ad cost above €4 per order.
Named example 3: StudioNest’s MediaMarkt speaker package
StudioNest runs a €4,000 MediaMarkt retail media package to support a compact Bluetooth speaker. The retailer report looks attractive: 420,000 impressions, 3,900 clicks, €0.78 CPC, €18,600 reported sales and a 16.4% media cost of sales. The package manager suggests renewing for the next promo window.
The return-reason loop changes the meeting.
The speaker sells for €79.99 and carries €21.50 contribution margin before ads and returns. The package’s reported sales include bundles and delayed attribution, so FiveX caps decision revenue at 70% until sell-out and return lag mature. That turns €18,600 reported revenue into €13,020 decision revenue. Against €4,000 spend, the decision media cost is 30.7% before returns.
Then support data arrives. MediaMarkt messages and marketplace reviews mention “bass weaker than expected” and “not loud enough for garden party”. The ad creative showed the speaker outdoors next to a barbecue. The product is actually strongest for kitchen, desk and small-room use. The campaign bought a use case the product did not want to own.
The verdict: Stop the package renewal, absorb the learning. StudioNest does not renew the fixed package. It moves €900 into always-on search for “compact bluetooth speaker kitchen” and “small room speaker”, lowers the max CPC to €0.54, and updates creative away from outdoor-party positioning. FiveX’s advertising automation keeps the new budget capped until return reasons from the repositioned traffic mature.
How to build the return-reason loop in practice
You need a repeatable weekly table.
| Field | Why it matters | Example |
|---|---|---|
| SKU / ASIN / EAN | Connects the complaint to the advertised product | USB-C hub 7-in-1 |
| Marketplace | Shows whether the issue is local or cross-channel | Amazon.de, bol.com, MediaMarkt |
| Campaign role | Decides whether to protect, shrink or stop traffic | Generic harvest |
| Return reason group | Turns messy labels into usable categories | Compatibility confusion |
| Evidence count | Separates noise from pattern | 7 returns + 6 tickets |
| Margin after ad cost | Shows whether the SKU can absorb the issue | €1.84 before delayed refunds |
| Decision label | Forces action | Shrink and fix |
| Reopen condition | Prevents old vetoes from becoming permanent | <8% mature return rate |
The reopen condition matters. Without it, teams either ignore the loop because it feels negative, or freeze spend forever after one bad week. A good service model lets constraints expire when better evidence appears.
Where FiveX fits naturally
This is exactly the kind of operating layer marketplace teams struggle to build in spreadsheets.
FiveX connects ad performance to SKU profitability, stock, inventory pressure, marketplace fees and operational signals. That means an account manager can see that a 22% ACOS is acceptable for one SKU and reckless for another. It means AI recommendations can be filtered through return maturity instead of blindly raising bids on a campaign with fresh, unadjusted revenue. It means the change log can show why a MediaMarkt package was not renewed even though the retailer dashboard looked positive.
For Advertentie Service clients, the hook is simple: Amazon, bol.com and MediaMarkt ads are not isolated dashboards. They are profit decisions with operational evidence attached.
The weekly meeting question
If you want to start tomorrow, add one question to the weekly ad meeting:
“Which campaigns looked good before returns, complaints and reviews had finished speaking?”
That question changes the room. It stops the team from treating ad-attributed sales as final and gives customer support and finance evidence a seat in the budget decision.
The best marketplace ad services do not just buy traffic efficiently. They learn when traffic reveals a promise problem. Then they stop paying to repeat it.