Why German Amazon campaigns underperform UK ones — and what the data shows.

Ad Badger's 2026 international report carries one number that does not get the airtime it deserves. Sellers who copy their UK Amazon campaigns to Germany without per-marketplace localisation see CTR drop by 42 percent and ACoS rise by 28 percent. Same products, same keywords (translated), same bid logic. A different country, a different result.

That gap is not a translation problem. It is a bidding architecture problem. Most PPC tools were built in the United States, for the United States Amazon marketplace, and then exported. The European marketplaces were treated as instances — flags on the same engine — instead of as the genuinely different demand environments they are. This post walks through what changes between DE and UK at the bid level, why a single optimisation engine cannot handle both well, and what a per-marketplace architecture actually looks like.

What is actually different between DE and UK on Amazon

Three things that move the bid math, in order of impact.

1. Conversion behaviour is not a translation of UK behaviour

German shoppers spend more time on product detail pages and consult more reviews before purchasing. UK shoppers convert faster on average and respond more elastically to price. A bidding engine that learns from blended Bayesian conversion rates across both markets gets the priors wrong in both. The disciplined fix is per-marketplace Bayesian priors — one BetaBinomial per category per marketplace — so that the first 100 impressions in Germany do not get rated against UK click-to-convert curves.

2. Keyword semantics differ even when the translation is correct

The German shopper searches with longer, more compound queries ("wasserdichte herren wanderschuhe gr 44") than the UK shopper ("men's waterproof hiking boots"). A symmetric semantic model that treats query and product the same way under-matches German long-tail queries against product detail pages. Asymmetric embeddings — RETRIEVAL_QUERY for search terms, RETRIEVAL_DOCUMENT for products — are roughly 19 percent more accurate on the German side of the catalogue in our internal benchmarks.

3. Inventory dynamics and FBA fees differ per marketplace

Days-of-cover is the single most under-used input in Amazon PPC. The same DOC threshold that protects a UK ASIN can starve a German ASIN whose lead time is longer because of cross-border shipping windows. A bidding system that does not read SP-API inventory per marketplace, and that does not throttle bids proportionally as DOC thins, will overspend into a stockout — and a stockout in Germany costs the seller their ranking long after the inventory arrives.

Why a US-first PPC tool struggles to fix this

The architecture problem is not that US tools cannot be configured for Germany. It is that they are configured by surface — currency, language, marketplace ID — not by demand. The bid math underneath is the same model trained on blended data. Per-marketplace timezone-aware scheduling, per-marketplace spend anomaly thresholds, per-marketplace circuit breakers (Japan kill switch at 6x baseline, Brazil at 3x, default at 4x) — these are not flags. They are decisions the model has to make differently in each market.

We built Mirox EU-native because the alternative — bolting Europe onto a US engine — is exactly the pattern that produces the +28 percent ACoS that Ad Badger keeps measuring.

What a per-marketplace bid actually looks like

Here is the kind of decision trace we surface for a single bid on a German campaign. Every line is CSV-exportable; the dashboard renders it inline.

Bid · DE · ASIN B07XXX · keyword "wasserdichte wanderschuhe" · proposed €1.84
— pCVR (German prior, 14-day rolling): 8.3% · alternatives ruled out: €1.62 (under-bid for first page), €2.10 (M_supply gate tripped at 0.6 DOC)
— relevance score (asymmetric semantic): 0.81 · safety gates: ROAS CB pass · spend anomaly pass · SP-API freshness 12min · M_supply 0.71 · marketplace gate pass
— agent: Tactician · proposer: Strategist · sentinel: 6.2 days cover

That trace is the thing that lets a marketing director defend the bid to a CFO who asks why the campaign is paying €1.84 instead of €1.50. It is also the thing that lets the marketing director catch a mistake — a stale inventory feed, a wrong category prior — before three months go by.

What to do if your tool cannot tell you why

Three concrete moves, in order of cost.

1. Pull last week's bid changes in your current tool and ask it, for one keyword, why the bid changed. If the answer is "the algorithm decided," you are not in a position to defend the spend.
2. Audit your German campaigns separately from your UK campaigns. Track ACoS, CTR, and TACoS independently. If you have been blending them in a single dashboard, the +28 percent gap may be invisible to you right now.
3. Watch a per-marketplace AI on your account, in shadow — read-only, free, indefinitely, before a cent moves. You see what the AI would have done on each campaign, with the trace. If the AI does not surface decisions your current setup missed, the per-marketplace argument was wrong. If it does, you have your answer.

The point of the third move is not that we want you to buy something. It is that the only way to know whether the +28 percent gap is reachable on your account is to measure it. Read what a full decision trace contains, or see how the sixteen agents handle per-marketplace context.