Advertising to earn stock, not just sales

This business does not buy as much stock as it wants. Each quarter its suppliers decide how many units it receives, and that decision is made on one number: how fast the previous allocation sold. Sell through quickly and the next allocation grows; leave units sitting and it shrinks, permanently, because the volume goes to a competitor who moved faster.

The account was being run as though none of that were true. Campaigns were managed to a target ACoS, which meant bids were pulled back on exactly the lines that needed to clear fastest, and slow SKUs were quietly de-prioritized rather than pushed. The result was a pattern the team had misread for years: hot lines sold out in week five and sat dead for the rest of the quarter, while slow lines dragged the average velocity down and cost the business its next allocation. Advertising efficiency was excellent and the business was shrinking its own ceiling every ninety days.

We changed what the advertising was being optimized for.

• Sell-through velocity as the primary objective — each SKU given a weekly units target derived from its allocation and the days left in the quarter, with ACoS as a constraint rather than a goal.
• Spend directed at the laggards, inverting the usual instinct: the units that threaten next quarter's allocation are the ones that need the budget, not the ones already selling.
• A pacing review every week, because a quarterly metric can only be corrected while the quarter is still running.
• Allocation requests backed by evidence — velocity and rank data presented to suppliers, which turned the quarterly conversation from a negotiation into a demonstration.

Quarterly allocation rose by 46%, and with sell-through up by half again revenue followed it from $520k to $1.19m a month — this business grows exactly as fast as its suppliers let it.

Weekly sell-through went from 58% to 91%, and the account no longer spends the back half of each quarter out of stock on its best lines and buried under its worst. Blended ACoS rose about two points in the process, and it was the right trade: an efficiency number measured against a supply ceiling is measuring the wrong thing.