What Managing $5M+ in Amazon Ad Spend Taught Me
Supporting more than $5 million in annual Amazon advertising spend only works when efficiency, net sales, and the work required to optimize are part of the same decision.
Zachary O'Donnell · · 2 min read
Supporting strategy for more than $5 million in annual Amazon advertising, inside a business above $50 million, changed how I read a campaign report.
Platform metrics are a starting point. ACOS, spend, clicks, and attributed sales describe activity. They do not, by themselves, say whether the business made a better trade.
Efficiency has to sit next to net sales
The account moved from an ACOS above 16% to 14.5%, and net sales increased. I keep those two facts together on purpose. A lower ACOS that comes from starving demand is not a win. A higher ACOS that nobody can connect to profitable sales is not a growth strategy.
I did not control the entire advertising budget. I analyzed it, recommended where the logic should change, and built the workflow that made the review repeatable.
Placement is where the average lies
Top of search, rest of search, and product pages do not earn the same return. A campaign-level ACOS can look calm while one placement absorbs money. The useful cut was the one that changed a bid or a modifier, not the one that made the slide symmetrical.
Scale is an operating problem
Manual review works until it does not. Past several thousand campaigns, consistency disappears. Two analysts can apply two unofficial playbooks and both feel rigorous.
The response was twofold. Consolidate structure — the account moved from more than 5,000 campaigns to fewer than 1,000 — and encode the first pass of analysis so the same rules hit every row. That system reduced manual optimization work by about 60–80%. Judgment stayed on the exceptions.
Profit is the question, with honest limits
I wanted advertising decisions tied to margin, not only to attributed sales. Cost and product context made some of that possible. Clean incrementality did not fall out of the ad console. Where I could not measure it, I did not invent a lift number. The recommendation stayed inside what the data could support: efficiency, mix, and visibility.
The lasting lesson is that the metric has to match the decision in the room. If leadership is trying to grow profitably, the review has to be able to say whether spend is helping that, and what should change before next week.