The 40,000 SKU Problem

When you run an ecommerce store with 10 products, you can check your ads every morning. You can look at the bids. You can adjust the copy. You can "feel" the market.

When you run an electrical supply business with 40,000 SKUs, that approach is suicide.

I learned this the hard way running EnergyAvenue.com. We had tens of thousands of products. Circuit breakers, connectors, conduit, lighting, controls.

The Complexity Cliff

There is a point where linear scaling of human effort hits a complexity cliff. You cannot hire enough media buyers to manually manage 40,000 keywords. Even if you could, the data volume per keyword on the long tail is so low that human intuition is useless.

If a specific 30-amp breaker gets two clicks a month, how does a human decide the right bid? They guess. And across 40,000 products, those guesses compound into millions of dollars of inefficiency.

Systems Scale Judgment

The solution is not to work harder. It is to build systems that encode judgment.

We realized that we needed infrastructure that could ingest inventory feeds, categorize products by margin and sell-through rate, and automatically generate campaigns that mapped to how buyers searched.

We stopped managing ads and started managing the logic of the ads. We built rules:

• If a product is dead stock, bid aggressively to liquidate.
• If a product is low margin, suppress bids unless conversion rate is >5%.
• If a product is a "runner" (high volume), optimize for profit volume, not ROI.

The Lesson for Stok Tap

This experience directly informs Stok Tap. We know that manufacturers cannot manually liquidate their dead stock because the operational overhead exceeds the recovery value.

But a system doesn't care if a SKU sells once a year. A system waits. A system costs zero marginal dollars to monitor that SKU. And when the demand appears, the system is there to capture it.