The skill
Install once in Claude. It fits your saturation curve, computes marginal CAC at current spend, and finds your stop spend, or allocates a fixed budget across channels.
What's inside
Three files, one job: know when to stop scaling
Built for growth marketers, not data scientists.
Install guide
A short setup walkthrough, plus a no-install option if you cannot add skills.
Plan B prompt
A single prompt you paste into any Claude chat to run the same analysis without installing anything.
How it works
Three steps, from your spend history to a budget decision.
Step 1Bring your data
Per-channel spend and conversions with real spend variation, plus your target CAC or LTV for one channel, or a total budget to split across channels.
Step 2Claude fits the curve
It fits a concave saturation curve, computes marginal CAC at current spend against your average, and either finds the spend where marginal CAC meets your target or allocates the budget to equalize marginal returns.
Step 3Get a decision
A stop spend as a range with a scale, hold or pull-back verdict, or a per-channel split, each with honest caveats.
The skill needs real spend variation to fit a curve, and it does not connect to your ad accounts. You bring the spend-and-conversions history, it does the math. Without variation it tells you to run a burst test first.
How to calculate marginal CAC
Marginal CAC is the cost of the customers your next dollar of spend actually buys, and it is the number that decides how far a paid channel can scale. It is almost always higher than the average or blended CAC on your dashboard, and at scale the gap is what separates profitable growth from spending past the point of return. Here is how to compute it, where to stop, and how to split a budget across channels.
What is marginal CAC and why it differs from average
Marginal CAC is additional spend divided by additional customers, the cost of the customers your next dollar buys. It is almost always higher than the average or blended CAC on your dashboard, because blended folds in organic and referral customers that cost next to nothing, typically sitting 20 to 40 percent below paid-only CAC. At scale the gap widens, and marginal CAC routinely runs 1.5 to 2.5 times the blended number, which is why a healthy blended figure is not a signal that you can keep spending.
How to compute marginal CAC from a saturation curve
Advertising response is concave, so each added dollar buys fewer conversions than the last. The skill fits a saturation curve, a Hill or negative-exponential function, to your spend and conversions, then takes the first derivative to get incremental conversions per dollar at any spend level, and inverts that to marginal CAC. The half-saturation point K marks the shape, and the knee where returns fall away sits near three times K. You need real spend variation in the data for this to work, without it no curve can be fit.
When to stop scaling a paid channel
The stop line is marginal CAC equal to LTV, equivalently marginal ROAS at breakeven, not average CAC equal to your target. Average mixes your cheap responsive core with the expensive marginal customer, so it can look healthy while the next dollar loses money. The skill solves for the spend where marginal CAC meets your target or LTV and reports it as a range rather than a single number, because the optimal point carries real variance.
How to allocate budget across channels
Once you have a curve per channel, the efficient split follows the equimarginal principle, move budget from lower to higher marginal return until the last dollar in every channel earns the same. The skill runs that optimization for a fixed total budget and returns a per-channel spend, bounded to each channel's observed range so it never recommends a level you have no data for.
What the skill cannot do, and how to stay honest
It is a planning tool, not a tracker, and it works only from the data you bring. A naive spend-versus-conversion curve confuses correlation with causation, since demand and spend often rise together, so anchor the model with incrementality or geo tests and validate any large reallocation with an experiment. It also will not tell creative fatigue apart from true saturation on its own, and if a fresh creative restores performance the problem was fatigue, not a ceiling. For the full reasoning and the evidence behind marginal CAC, read Marginal CAC Decides How Far You Can Scale Paid Ads.