Skip to main content
The AdAdvisor scale skill prevents the most expensive mistake in Meta (Facebook) ad management: scaling fast, resetting Meta’s learning phase, watching CPA spike, panicking, killing the winner. Scaling is the most-talked-about and least-understood part of Meta ads. The skill enforces eligibility checks and respects the cadence senior buyers use.

​
When should I use the scale skill?

Triggers automatically on:
  • “Scale”
  • “Scale this campaign”
  • “Scale this winner”
  • “Increase budget”
  • “Give it more budget”
  • “Ramp up”
  • “Grow spend”
  • “Expand”
  • “Raise budget”
  • “Duplicate to new audience”

​
How does the AdAdvisor scale eligibility check work?

Before suggesting any budget change, the skill verifies:
1

3+ stable days above target

ROAS ≥ target × 1.2 (or CPL ≤ target × 0.9 for lead-gen) for at least three consecutive days. Pulled from adadvisor:get_timeseries.
2

Past Meta's learning phase

50+ conversions in the trailing 7 days. Scaling in learning resets the clock and costs you the established CPA.
3

Stable new-customer rate

A “scaling winner” that is actually serving repeat customers will not survive vertical scale. The skill flags this when surfaceable.
4

Healthy frequency

Prospecting under 2.5, retargeting under 6. Frequency creep above these means audience saturation, not headroom.
5

Pixel firing recently

Last fired within 12 hours. If signal is stale, any scaling decision is based on guesswork.
If any check fails, the skill stops, explains why, and routes to diagnose for a 48-hour hold and re-check.

​
What are the four scaling axes for Meta ads?

The skill picks the right axis based on your current spend, CPA stability, and audience availability:

​
What is the 20% rule for scaling Meta ads?

The consensus default: raise daily budget by no more than 20% every 2 to 3 days. Larger jumps reset the learning phase. Math example: a 100adaywinnerscaledby20100 a day winner scaled by 20% every 2 days reaches 173 a day in a week, $430 a day in a month, with CPA in range the whole way. Aggressive scaling (50% or 100% jumps) can work on confirmed winners with deep conversion volume, but only with bid caps and explicit acceptance of a learning reset. The MCP server enforces a 2x / 0.5x guardrail on change_entity_budget. The skill respects it by staging changes (100→100 → 120 → 144→144 → 173) rather than passing force=True to override.

​
How do I scale Meta ads from 1kadayto1k a day to 10k a day?

Three stages, each 2 to 6 weeks depending on creative production cadence:
Vertical 20% on confirmed winners. Add 1 to 2 ad sets with broader audience. Lock in 3 to 5 winning creatives. Track CPA inflation. If it crosses 1.5x target, hold.
Horizontal scaling. Duplicate winners into new audiences (LAL expansions, geo). Introduce ASC alongside manual (with exclusions to prevent cannibalization). Creative production becomes weekly. Start tracking MER as the company-level number.
Bid-cap diversification (Tichenor). Target ROAS campaigns (Faris). Layer brand/awareness at 10% to 15% of budget. Creative production cadence: 4+ new concepts per week. Most operators stall here because they protect a misleading ROAS instead of optimizing for total profit dollars.
At every stage transition, the skill makes you confirm: CPA will inflate 10% to 20% during the stage. That is expected. The job is to keep marginal revenue positive vs marginal spend.

​
What anti-patterns does the scale skill prevent?

Resets learning, CPA spikes, you panic-kill. The skill enforces 3-day stability and a 20% cadence.
Scaling typically INFLATES CPA short-term as you exit the optimal audience. The skill sets expectations explicitly: CPA will creep 10% to 20% in the first 3 days. Kill if it crosses target × 1.5.
Wasted budget. The skill switches axes (horizontal or bid-cap) when daily delivery flattens.
The guardrail exists for a reason. The skill stages changes (1.5x, wait, 1.5x, wait) rather than 5x in one call.
Does not survive new-customer exposure. The skill checks new-customer rate before recommending scale.

​
How do I install the scale skill?

Claude Code

Ships with the full plugin install.

Claude.ai & Desktop

Download adadvisor-scale.zip and upload via Settings, Capabilities, Skills.

​
Frequently asked questions

Apply the 20% rule. Raise daily budget by no more than 20% every 2 to 3 days. Bigger jumps reset Meta’s learning phase, which costs you the established CPA. The AdAdvisor scale skill enforces this cadence and stages larger increases automatically.
When vertical scaling causes CPA to inflate over 10% per 20% budget increment. That signals the original audience is saturated and adding more budget just buys lower-quality impressions. Horizontal scaling (duplicating the winner to new audiences, geos, or LAL tiers) creates a new cost basis. The skill watches for this signal in the timeseries data.
Charley Tichenor’s scaling technique. Once Lowest Cost campaigns hit a spend ceiling and 50+ conversions per week, duplicate the winning ad set with a bid cap at target_CPA × 1.20. Build a portfolio of bid-capped duplicates that each find their own price point. The skill recommends bid caps when vertical and horizontal scaling have both flattened.
Andrew Faris’s discipline: 70% of mature DTC spend belongs on Target ROAS campaigns (LOWEST_COST_WITH_MIN_ROAS bid strategy with optimization_goal VALUE). After a winner is stable for 7+ days, duplicate it into a Target ROAS variant for stability floor. The skill recommends this once you have confirmed, post-learning winners.
No. The skill enforces an eligibility check (3+ stable days, past learning, healthy frequency, stable new-customer rate, pixel firing) and a 20% cadence, but every actual budget change requires your explicit confirmation. The MCP server also enforces a 2x guardrail on change_entity_budget to prevent runaway scaling.

Diagnose

Confirms scale eligibility before any budget change.

Targeting

Horizontal-scale audience picks (LAL tiers, geo expansion).

Playbooks

Full stage plan with operational checkpoints.
Last modified on May 13, 2026