System 02 — automation
Ads that answer tomargin, not ACOS.
Advertising Intelligence sets every bid against contribution margin per SKU — with stock position as a hard gate. 2,400 bid decisions a day, 12 rules, 1 named owner. TACOS vs target: −1.2 pts, trailing 90 days.
09:14:07BID ↓ $0.12AQ-FILTER-20 · MARGIN 31.2% · RULE 04
09:14:31NEG ADDED"aquarium decor" · CVR 0.3% · RULE 09
09:15:02BUDGET HOLDAQ-HEATER-50W · COVER 19D ≤ 28D GATE
09:15:40BID ↑ $0.08AQ-PUMP-300 · MARGIN 18.1% → CEILING $1.14
Demonstration log: the bid engine lowers and raises bids by rule, adds negative keywords, and holds budgets when the stock gate from the demand system reports 19 days of cover.
What it replaces
ACOS is a ratio. Margin is money.
An ACOS target treats a 40%-margin SKU and a 12%-margin SKU as the same customer. They aren't. One of them is buying rank with money it doesn't have.
One target, every SKU. Low-margin SKUs bleed quietly at "target." Ads keep running into stockouts, buying rank the inventory can't serve.
A ceiling per SKU, a gate per unit of stock. Rules execute; changes outside the rules go to a human, with the reason attached.
The mechanism
2,400 decisions a day, 1 owner.
The ceiling is computed from contribution margin, not revenue — a bid that wins the click and loses money never clears it.
The stock gate is a signal from System 01. Spend throttles at ≤28 days of cover; rank is not bought for shelves that will be empty.
Anything outside the 12 rules stops and asks. The review queue averages 6 items a day — with the rule that would have caught each one proposed alongside.
Automation with judgment
What the machine runs. What it never will.
- Bid changes within rules 2.4K/DAY
- Search-term harvest and negatives DAILY
- Budget pacing HOURLY
- Stock-gate throttling SIGNAL: SYSTEM 01
- Rule and threshold changes LOGGED
- Campaign architecture
- Brand-defense strategy
- Review queue ~6 ITEMS/DAY
Typical results — 2025 cohort, installed ≥2 quarters
TACOS down. Contribution up. Stated basis.
TACOS trend, median client: 14.1 percent falling to 8.6 percent over 10 months, against a 9.0 percent target. The final two months are forecast, drawn dashed.
Categories with under 60 days of data run conservative defaults. The ramp takes a quarter. Accounts are told this before signing, and the first-quarter scorecard shows it.
Find out what your ads cost you in margin.
The audit rebuilds your TACOS against contribution, not revenue. Median finding: 23% of spend serves SKUs that lose money on the click.