A 25-person cross-border team was scanning 300–500 candidate SKUs each week across four to six platforms and multiple regional markets. The work before the work — intelligence gathering, competitor price comparison, margin estimation — consumed between 80 and 120 hours of overtime per week across the team. Product selection meetings were routinely long because the preparation was incomplete. The judgment was being crowded out by the reading.
Three invariants, one funnel
We started by watching the decisions, not the discussions. Every week the team was applying the same three tests without naming them: any item with gross margin below 22% was never advanced; any item with a logistics lead time over 14 days was held for human review rather than auto-advanced; and an item only entered sample purchasing when both review-growth and competitor price-band qualified together — a dual gate that the team had always enforced but never written down. Those three rules could be encoded. What remained — reading platform signals, weighing regional timing, deciding which borderline case deserved the sample budget — that stayed human.
We used to work late hunting for information; now we leave early because the judgment is already on the table.
The harness
We did not deliver a report. We delivered a running system: a selection funnel that ingests the candidate pool, applies the three invariants automatically, and produces an intelligence brief for each qualifying item. The brief — including margin estimate, lead-time flag, and competitor price-band — moved from a two-working-day assembly to under three hours. Over the 8-week measurement window, approximately 120 opportunity samples were processed under the new workflow. Candidate-SKU throughput rose from roughly 40 per week to 160. Per-person overtime fell from around 18 hours per week to under three. Over the same period, the company recorded roughly 20% month-over-month revenue growth for several consecutive months — a same-period business result, not a claim about the harness.
The number that matters is three: there are now three places where the team’s selection logic is written down, and those three places are the same places that enforce it.