An appliance store kept losing shopping-ad sales to silent feed disapprovals — some SKUs sat invisible for weeks. We built hourly feed monitoring that auto-repairs common errors and resubmits them. Disapproved SKUs dropped 90%, and shopping revenue rose 19%.
Typical results
78%
Errors fixed with no human
1 hr
Max time an error lives
+19%
Shopping-ads revenue
Figures reflect typical results for this kind of automation.
The challenge
Google and Meta feeds carried 4,000 appliance SKUs, and at any moment around 340 were disapproved — missing GTINs, price mismatches after ERP updates, image-policy flags. Nobody noticed until ad performance dipped; errors typically lived 72 hours before detection. The catalog team burned seven hours a week firefighting inside two different merchant consoles.
Before and after
SKUs disapproved at any time
−90%Time to detect a feed error
−99%Weekly hours on feed fixes
−86%The automation
We built a Make scenario that pulls disapprovals from Google Merchant Center and Meta every hour and classifies each error. Known issues are repaired automatically from ERP source data — the correct price pushed, a valid GTIN filled in, a compliant image swapped — and the SKU is resubmitted on the spot. Anything needing judgment becomes a ticket naming the exact fields to fix, and a live dashboard tracks feed health across both platforms.
How it works
- 01
Hourly pull
Disapprovals and warnings are pulled from Google Merchant Center and Meta every hour.
- 02
Classify
Each error is classified automatically: missing GTIN, price mismatch, image policy, broken link.
- 03
Auto-repair
Known issues are fixed from ERP source data — correct price, valid GTIN, compliant image — and resubmitted on the spot.
- 04
Ticket the rest
Errors needing judgment become tickets for the catalog team, naming the exact fields to fix.
- 05
Health dashboard
A live dashboard tracks approval rates on both platforms, so regressions surface the same day.
Starts from
- Google Merchant errors
- Meta catalog errors
Delivers
- Fixed & resubmitted SKUs
- Tickets for catalog team
- Feed health dashboard
The impact
Disapproved SKUs fell from around 340 at any time to 34 — a 90% drop — and 78% of errors now fix themselves without a human. No error lives longer than an hour, weekly feed firefighting shrank from seven hours to one, and shopping-ads revenue climbed 19%.
Questions about this automation
Which feed errors can the automation fix on its own?
Errors with a clear source of truth are fixed automatically: a price that drifted from the ERP, a missing GTIN that exists in product data, a broken product link, or an image that breaks policy when a compliant alternative is on file. The corrected SKU is resubmitted to Google and Meta in the same run.
What happens to errors that need a person to decide?
Anything without a safe automatic fix, such as a policy flag on product wording or a GTIN missing from every source, becomes a ticket for the catalog team. The ticket names the SKU, the platform, the error and the exact fields to change, so the fix takes minutes instead of a search through two consoles.
How often does the check run, and what is needed to start?
Disapprovals and warnings are pulled every hour. The build needs API access to the Google and Meta merchant accounts, read access to the ERP or product database, and a place for tickets. It is a multi-step build, usually rolled out in stages, starting with monitoring and adding automatic repairs per error type.