Marketplace advertising teams love to call search terms “high intent” or “low intent” as if intent is printed on the keyword forever. It is not. Intent drifts. A query that looked like a clean buying term in March can become a comparison term in April, a bargain-hunting term during a promotion, a support term after a product issue, or a margin trap after competitors change price.
That drift matters once a brand spends more than roughly €5K per month across Amazon, bol and MediaMarkt. At that level, the account is no longer losing money only through obviously bad keywords. It loses money through terms that used to deserve budget and quietly stopped deserving the same treatment. The campaign still has historical sales. The bid rule still trusts the old conversion rate. The operator still sees the term in an “exact winner” campaign. Meanwhile, the customer behind the query has changed.
The named mistake I see is letting yesterday’s intent label approve tomorrow’s bid. A search term graduates from auto to exact because it produced ten orders. Six weeks later, the same term attracts shoppers comparing sizes, hunting discounts, checking accessories or looking for a newer model. The ad platform reports ACOS. The business needs to know whether the term still represents profitable demand for this SKU today.
My stance: managed marketplace advertising should run an intent drift ledger. Not a keyword dump. Not a once-per-quarter negative keyword clean-up. A weekly operating layer that reclassifies important search terms by buyer intent, SKU margin, stock cover, offer strength and channel role before bids, budgets or negatives move.
This guide is for NL and BE brands spending from about €5K per month on Amazon Ads, bol Sponsored Products and MediaMarkt retail media. FiveX helps here because the ad decision can sit next to SKU profitability, inventory insight, margin analysis, repricing context and AI recommendations. Intent is not only a word problem. It is a profit-permission problem.
What current keyword advice gets right
The existing advice is useful. BidX makes the right point that keyword research should be a recurring process, not a one-time export from a tool. Their campaign-structure guidance also explains why auto, broad, exact and product-targeting campaigns need separation: without structure, budget control and evaluation speed suffer.
Seller Sprite frames PPC keywords as “buying queries”, not just relevant queries. That distinction is important. A shopper searching “professional heavy duty garlic press” behaves differently from a shopper searching “kitchen gadgets”. Seller Labs gives beginners a practical workflow: start with the product, map how customers search, test match types and refine. SalesDuo correctly places keywords inside a wider 2026 retail media system where listings, creative, pricing, reviews, AI discovery and organic visibility all influence advertising performance. Podean’s marketplace-media positioning adds another helpful layer: non-media signals such as low stock, price changes and merchandising should inform media strategy.
The gap is not that these sources are wrong. The gap is that most keyword advice treats intent as something you identify, then optimize. In real accounts, intent is something you keep proving. The label can expire.
The angle competitors usually miss: intent has a shelf life
A search term is not a stable asset. It is a live promise about why shoppers typed those words this week. That promise can change for at least five reasons.
First, marketplace context changes. Amazon might show a new dominant competitor above you. bol might shift the koopblok because delivery promise changed. MediaMarkt shoppers might compare your accessory against a bundled retailer offer. The query text stays the same, but the shelf around it changes.
Second, price position changes. “Wireless earbuds waterproof” can behave like a buying term when your product is €39.95 and the main competitor is €49.95. It can become a comparison term when the competitor drops to €34.95. If the bid rule only sees last month’s conversion rate, it will keep buying traffic that now needs a different offer.
Third, stock changes intent quality. A term may still convert, but if the SKU has nine days of stock, the commercial role changes from scale to protect. The operator should not read a healthy CTR as permission to accelerate a stockout.
Fourth, seasonality changes the shopper’s job. “Air purifier bedroom” in January may be a health-driven buying query. In June it may be a hay-fever urgency query with a shorter decision window. In November it may be a gift-comparison query with different return risk.
Fifth, product maturity changes the term’s job. During launch, a broad term might be allowed to buy learning. After 90 days, the same term should either prove contribution margin or move into a capped learning lane. Keeping it in the same budget pool is how learning quietly becomes a hobby. A charming hobby, yes. Still expensive.
Build the intent drift ledger
The ledger does not need to be complicated. It needs to force the right conversation before money moves. For every meaningful search term or product target, capture seven fields.
1. Current intent class
Use simple classes that operators can apply consistently: protect, harvest, compare, learn, support and misfit. Protect terms defend branded or hero demand. Harvest terms have proven profitable buying intent. Compare terms attract shoppers weighing alternatives. Learn terms buy evidence. Support terms indicate post-purchase or troubleshooting behaviour. Misfit terms do not belong with this SKU.
2. Previous intent class
This is where drift becomes visible. If a term moved from harvest to compare, the action is not automatically “lower bid”. It may be “check price position”, “improve image clarity”, “move to a comparison campaign”, or “protect only branded variants”. FiveX dashboards are useful here because the operator can see whether the drift appeared alongside margin, stock, price or conversion changes instead of blaming the keyword alone.
3. Drift trigger
Name what changed. CPC up 28%. Conversion down 19%. Return rate up 7 points. Stock cover down to 12 days. Main competitor undercut price by €6. Review rating fell from 4.5 to 4.1. bol delivery promise moved from tomorrow to three days. MediaMarkt bundle went live. A ledger without a trigger becomes a diary. Lovely for feelings, less useful for spend control.
4. SKU profit permission
Connect the term to contribution margin. A €2.10 CPC can be fine for a €129 electronics accessory with €38 contribution margin and 7% conversion. It is not fine for a €24.95 consumable with €5.20 contribution margin and 4% conversion. FiveX product profitability and margin analysis turn this into a real permission check: what CPC, ACOS or spend cap can this SKU absorb after fees, returns, discounts and service costs?
5. Stock and offer permission
Search intent without availability is a trap. If the offer is weak, the ad should not work harder to compensate. For bol, check LVB or own-warehouse delivery promise and koopblok stability. For Amazon, check Buy Box, Prime/FBA status, review quality and stock cover. For MediaMarkt, check retailer approval, delivery promise, bundle exposure and stock allocation. FiveX inventory insights make the difference between “this term converts” and “this term is allowed to scale”.
6. Action label
Use labels that prevent overreaction: scale, hold, cap, isolate, repair, negative or watch. A term that drifts from harvest to compare may need isolation, not deletion. A support-intent term may need a negative in Sponsored Products but a content fix on the product page. A learn term with high CPC may need a capped budget, not a dramatic farewell ceremony.
7. Review date
Intent labels expire. Set a review date based on spend velocity and business risk. High-spend exact terms might be reviewed weekly. Low-spend discovery terms can wait two or four weeks. Terms affected by price changes, stockouts or promotions should be reviewed when the commercial event ends.
Example 1: NorthPeak Nutrition and the “protein powder vanilla” trap
NorthPeak Nutrition sells a €34.95 vanilla protein powder on Amazon.nl. The SKU has €11.40 contribution margin before ads, a normal return rate below 3% and healthy FBA stock. In May, the term “protein powder vanilla” produces 74 ad-attributed orders at 23% ACOS, so the agency moves it into an exact harvest campaign with a €0.82 bid.
In June, two things change. A larger competitor runs a €5 coupon and Amazon starts showing more multi-pack offers above NorthPeak. CPC rises from €0.76 to €1.03. Conversion falls from 12.8% to 8.1%. The term still produces sales, but the break-even CPC based on current contribution margin and conversion is now roughly €0.92. If the operator only sees historical winner status, the bid stays too high.
The intent drift ledger reclassifies the term from harvest to compare. Action: cap bid at €0.88, split multi-pack search terms into a separate watchlist, and ask whether repricing or bundle strategy should respond. FiveX helps by putting Amazon Ads performance next to SKU margin and repricing context, so the service does not treat a price-position problem as a pure PPC problem.
Example 2: BrightNest Home on bol and the delivery-promise drift
BrightNest Home sells a €59.95 cordless desk lamp on bol.com. The search term “draadloze bureaulamp” looks like a clean buying query: 9.4% conversion, €0.41 average CPC and 18% ACOS over four weeks. The account spends €6,800 per month across bol and Amazon, so this is meaningful but not unlimited money.
Then the Dutch warehouse misses a replenishment handover. Delivery promise slips from “morgen in huis” to three days for five business days. The term keeps getting clicks, but conversion falls to 5.6% and bol spend reaches the daily budget earlier because the algorithm still trusts recent history. The keyword did not become irrelevant. The offer became weaker for the same intent.
The ledger marks the drift trigger as delivery promise changed and moves the term from harvest to protect. Action: reduce budget by 35%, keep exact coverage for bottom-of-funnel shoppers, pause broad variants, and restore scale only after two days of normal delivery promise and conversion above 8%. FiveX inventory and marketplace-operation data make that decision faster because the ad service can see the fulfilment issue before simply lowering bids and calling it optimization.
Example 3: VoltEdge Accessories and MediaMarkt comparison demand
VoltEdge Accessories promotes a €79.99 USB-C docking station through MediaMarkt retail media. The product has €21.50 contribution margin after marketplace costs and a planned monthly media budget of €3,000. The term cluster around “usb c docking station laptop” initially performs well: €0.64 CPC, 6.8% conversion and a comfortable contribution-per-click.
During back-to-office week, MediaMarkt features a competing dock in a bundle with an HDMI cable. The search term now attracts comparison shoppers. CTR remains strong because the category is hot, but conversion drops to 4.1% and assisted sales shift toward cheaper accessories. A normal ad dashboard might say “demand is strong; hold visibility”. The margin view says the term no longer has the same job.
The ledger reclassifies the cluster from harvest to compare-learn. Action: isolate the cluster with a €45 daily cap, route budget toward exact model-compatible terms, and use FiveX AI recommendations to flag whether the product page needs a compatibility table rather than another bid increase. The operator does not punish the term. They change its permission.
How to review drift every week without drowning in search terms
You do not need to classify every tiny query. Start with the terms that can actually hurt or help the account.
- Top spenders: the 20% of terms or targets that drive most spend.
- Recent graduates: terms moved from auto, broad or discovery into exact campaigns in the last 30 days.
- Margin-sensitive SKUs: products with low contribution margin, high returns or active discounts.
- Operationally fragile SKUs: products with low stock cover, changing delivery promise, Buy Box instability or retailer approval risk.
- Cross-marketplace overlap: terms where Amazon, bol and MediaMarkt all compete for the same demand pool.
The weekly meeting should ask four questions. Which winner lost its old intent? Which loser is actually a shelf problem? Which learning term has earned harvest permission? Which term should stop spending because its commercial event ended?
This is where managed service becomes more valuable than task execution. Anyone can download a search term report. A good operator can explain why a term changed role, what evidence would change the decision back, and which business owner needs to fix the non-media issue.
The FiveX way: connect intent to profit permission
In FiveX, the useful version of keyword management is not “add more keywords”. It is “decide what each search term is allowed to do with money”. That decision improves when advertising data sits next to product profitability, stock cover, returns, price changes and marketplace performance.
For an Advertentie Service account, the workflow looks like this:
- Pull Amazon, bol and MediaMarkt ad performance into one operating view.
- Attach SKU margin, fees, returns and discount assumptions to each advertised product.
- Flag search terms whose CPC, conversion, spend share or order mix changed materially.
- Use inventory insights and offer checks before scaling terms that still look efficient.
- Let AI recommendations suggest drift candidates, but require an operator decision for budget increases, negatives and cross-marketplace shifts.
The important part is restraint. Automation can spot drift. It should not blindly decide that every conversion drop deserves a bid cut or every efficient term deserves more budget. Some terms need a margin fix. Some need content. Some need repricing. Some need patience because they are still buying evidence. The operator’s job is to classify the commercial reality behind the query.
Final thought: keywords do not spend money, assumptions do
When a marketplace ad account is small, keyword work can be mostly research and hygiene. Find relevant terms. Add negatives. Move winners into exact. Keep ACOS under control. That is a perfectly respectable starting point.
At €5K+ monthly spend, the job changes. The expensive mistakes come from stale assumptions: this query still means buying intent, this SKU can still afford the click, this marketplace still has the best offer, this stock position can still handle demand, this campaign still has the same role.
An intent drift ledger makes those assumptions visible. It gives Amazon, bol and MediaMarkt operators a practical way to say: this term was allowed to scale last month, but today it needs a cap; this query looked wasteful, but the shelf is broken; this discovery term has earned harvest status; this branded term is stealing credit and should defend only.
That is the difference between managing keywords and managing profit. Keywords are the surface. Intent is the customer’s job. Profit permission is the business decision. Keep all three connected, and the next euro of marketplace ad spend has a much better chance of coming back with friends.