Scaling rules that actually work (and the dumb ones that don't)

The classic scaling rule looks like this: if ROAS > 3 for 3 days, bump budget 20%. Pause if ROAS drops below 2.

It's a rule based on one variable (ROAS) measured against a fixed number (3.0) with zero context. It doesn't know your margin. It doesn't know your return rate. It doesn't know that ROAS 3.0 on a bestseller is great and ROAS 3.0 on a 15%-margin SKU is a loss.

Account-wide ROAS rules flatten every product into a single metric. That's why every brand using them hits a ceiling, usually somewhere between 2x and 3x, and can't push through.

A real scaling decision looks at: current ROAS, margin tier, 30-day velocity, seasonality, day-of-week pattern, and pacing against the monthly budget. It weighs those together before moving a bid or a budget.

Concretely: if a campaign is running 3.2x ROAS on a hero-tier product (high margin, high velocity), push budget up. If it's running 3.2x ROAS on a low-margin filler SKU during a normally-slow weekday, hold. Same ROAS, opposite move, because the context is different.

A mid-seven-figure apparel brand had been stuck at 2.1x ROAS for six months. Their agency was running exactly the kind of naive rules described above.

The changes that moved the number: labeling the full 1,800-SKU catalog by margin and velocity, splitting Shopping into tier-specific campaigns, and turning on context-aware scaling with a ROAS floor of 3.0 for hero SKUs, 4.0 for workhorses, and a hard pause below 2.0 for bleed.

Inside 14 days, the account was at 3.1x. By day 60, it was sustaining 4.2x with 18% more spend. The rules weren't smarter in the abstract. They were smarter *per tier*.

Seasonal pacing rules: auto-pull back budget on historically slow days, push on historically strong ones. Most brands have weekly and monthly patterns they never touch.

Early-kill rules: pause any new ad group or creative after 3x target CPA with zero conversions. Saves budget before the spend report does.

Anomaly rules: flag, not pause, when ROAS drops >40% vs the 30-day baseline. You want a human looking before the automation pauses a seasonal swing.

Any rule that uses account-wide ROAS as its only trigger. Delete it.

Any rule that uses fixed dollar budgets instead of percentages. Deletes itself when spend scales.

Any rule that auto-pauses on a 7-day ROAS dip below a fixed number. Ignores seasonality and will kill your December.

Any rule that sends a Slack alert but doesn't change anything. Noise.

If you're building rules manually: label your catalog by margin tier first, then write per-tier scaling rules. You'll get most of the lift from that one change.

If you want the context-aware version without building it yourself: that's what Scaley's scaling rules engine does, 24/7, with every change logged and one-sentence-justified. Either path works. The naive-rule path does not.