From 15,000 to 300,000 SKUs a month, same two people

The situation

The agency finds profitable wholesale opportunities for its clients. Finding them means checking SKUs one by one: price history, competition, fees, margin at the buy price, restrictions, velocity.

Two analysts did this by hand, roughly 15,000 SKUs a month between them. Capacity was the ceiling on the whole business. More clients meant more analysts, and quality wobbled when people got tired.

The week, before

1. Open the next SKU on the list. Look up its history.
2. Check competition, fees, and restrictions across tabs.
3. Calculate margin at the offered buy price.
4. Log the verdict and move to the next one. Repeat, all day.
5. The obvious winners get found. The long tail never gets looked at.

What we deployed

We wrote the agency’s buying rules down, all of them, and gave them to agents.

Now the agents run the scan: every SKU checked the same way, every time, against price history, competition, fees, margin, restrictions, and velocity. Up to 300,000 a month.

What survives the rules lands in a ranked opportunities queue with the reasoning attached. The two analysts spend their day judging flagged winners instead of grinding through the long tail.

SKU firehose → Agents apply buying rules → Ranked opportunities → Analysts judge winners

How it runs

Supplier catalogs and price lists flow in at whatever volume they arrive. The agents work through them around the clock, checking each SKU against the same rule set the analysts used to hold in their heads, now written down and versioned.

Most SKUs fail fast: wrong category, gated brand, no margin at the buy price. The agents discard them and keep receipts on why. The interesting ones get a full pass: price history over time, competitor depth, fee math, realistic velocity.

What emerges is a ranked queue. Each opportunity arrives with its reasoning: the margin math, the history chart, the risk flags. An analyst opens it and decides in minutes, not hours.

The buy decision never left the humans. What changed is what the humans spend their attention on: judgment on strong candidates instead of elimination of weak ones.

And the quality bar went the right way. Agents never get tired on SKU nine thousand. Every check runs, every time. The ceiling on the business is no longer how fast two people can click.

What changed

• 20x - more SKUs analyzed per month, with the same two analysts
• Every check - the full rule set runs on every SKU, no fatigue
• Ranked queue - analysts judge winners with reasoning attached

Where people stay in charge

• Buy decisions stay with the analysts, on flagged winners only.
• The buying rules belong to the agency and change on their word.