Before
- Missing product images sat in a backlog, attacked ad hoc
- Suspicious offers were checked manually, each price judged in isolation
- The quality scorecard was maintained by the team
- Editors spent their time hunting and fixing defects
Heureka repaired hundreds of broken listings across three markets in the first weeks.
95%+
of identified missing-image products got a screened replacement
99%+
of runs complete without intervention
€15–50K
click revenue recovered per year, modelled
Operating problem
On a price-comparison marketplace, a product without an image and an offer with a wrong price are the same thing: clicks that do not happen and shopper trust that erodes.
A listing without a proper image converts fewer clicks, an offer whose merchant silently changed its price or product misleads shoppers, and a quality scorecard nobody has time to maintain means nobody manages any of it.
For a small content team covering three markets, the backlog of such defects was structurally larger than the team.
After Duvo
- Missing images worked off continuously: a screened replacement proposed for 95%+ of identified products, the rest documented
- Suspicious offers cleaned daily across three markets, every price market-verified
- The quality scorecard updates itself daily
- Editors approve prepared fixes
Duvo has been an amazing partner in this journey so far. We are glad to partner with Duvo on our AI transformation and it's really exciting to see how quickly it spreads across the company and adds value.
Jan Knettig
Product Leader, Heureka
The workflow from signal to outcome.
01
Finds replacement images
For products missing images, Duvo finds candidate images on the open web and screens each one against hard quality criteria: exact product match, no watermarks, sufficient resolution.
02
Hands editors a ready proposal
The best image goes to an editor as a ready-to-approve proposal. Products where nothing qualifies are closed with the reason recorded, so nothing is silently skipped.
03
Verifies suspicious offers daily
Duvo checks the flagged merchants daily across all three markets, verifies every price against what the rest of the market charges for the same product, and corrects, removes, or clears the offer. Anything ambiguous goes to a human.
04
Keeps the scorecard current
The daily quality scorecard for all three markets fills itself, without anyone touching it.
Evidence status
Measured result with modelled impact
The hundreds of listings repaired in the first weeks, the 95%+ of identified missing-image products that received a screened replacement, the daily market-verified price checks across three markets, and the 99%+ of runs completing without intervention are customer-approved operating figures. The €15–50K/yr of click revenue recovered and the €15–25K/yr of editor capacity returned are modelled, with repaired listings earning €5–20 a month each once visible again.
Systems involved
Heureka product catalogue, Open-web image sources, Merchant offer feeds, Daily quality scorecard
Control points
- Every candidate image is screened against hard quality criteria (exact product match, no watermarks, sufficient resolution) and an editor approves the proposal before it is applied; products where nothing qualifies are closed with the reason recorded.
- No flagged price is judged in isolation: every one is verified against what the rest of the market charges for the same product, and anything ambiguous is escalated to a human.
The €15–50K/yr of click revenue recovered and the €15–25K/yr of editor capacity returned are modelled from repaired listings earning €5–20 a month each once visible again, growing with continued repair volume.