How to prioritize which PDPs to fix first.

The slow way: the spreadsheet triage meeting

The manual version is a Monday meeting with a 400-row spreadsheet. Someone sorts by audit score, someone else objects that the bottom rows are all discontinued colorways, and a third person pulls last quarter's revenue export to cross-reference. An hour in, you have a hand-merged sheet with three sort orders and no agreed formula. The ten PDPs that get fixed are the ones somebody advocated for, which biases the quarter toward whoever attends the meeting.

The honest manual fix is to build the formula yourself: export per-SKU revenue from Shopify, join it to your audit scores in a spreadsheet, add a column for how many engines skip each SKU, and multiply. That produces a defensible queue and takes an afternoon — once. The decay problem is the same as every manual process: revenue ranks shift, fixes ship, new regressions appear, and the sheet is stale in two weeks. Prioritization is not a one-time sort; it is a standing computation.

The eCommerce Insights way

1. Score everything first. Prioritization needs the full picture: every SKU's citation score and agent-readability score from the catalog scan, plus the named gaps behind each. Partial audits produce confident wrong answers.
2. Weight by revenue. With Shopify connected, revenue rank joins automatically. A 40-point gap on a top-20 revenue SKU outranks a 60-point gap on a clearance item — every time, by construction.
3. Factor engine coverage. A SKU invisible on all six engines has more recoverable upside than one missing only on Gemini. Coverage breadth multiplies opportunity in the ranking.
4. Take the top 20, split by fix type. Schema gaps batch to the developer queue (often thirty SKUs in one approval pass), copy rewrites go to the writers via the rewrite workflow, review-signal gaps go to whoever owns the review app. Parallel tracks, one ranked queue.
5. Let the queue refresh. On each refresh — weekly on Starter, daily on Growth — fixed SKUs drop out, regressions enter, and the top 20 stays current. The Monday meeting reviews the queue instead of building it.

The queue's expected-lift numbers also give you the leadership story for free: "we fixed the twenty highest-opportunity PDPs this quarter" is a defensible sentence in a way "we fixed four hundred things" never is. That thread continues in prove AI search ROI to leadership.

What "good" looks like

Cadence figures from patterns across active eCommerce Insights catalogs, illustrative. The test of a good queue is what it excludes: if your writers spent zero hours this month on long-tail PDPs that nobody buys, the prioritization is working.

Related jobs