Finding the saturation point: when to stop scaling Meta

"It's working - spend more" is the most expensive sentence in growth marketing. Channels do not scale linearly. They follow a saturation curve: steep and generous at low spend, flattening as you exhaust the cheap audience, until the marginal return dips below what the same pound earns elsewhere.

Average ROAS hides the cliff

Suppose Meta returns a blended 3.2x at 400k/month. Great number. But blended ROAS is the average over every pound - including the early, efficient ones. The question that should set the budget is marginal: what does pound 400,001 return?

On a saturating curve those diverge sharply:

• Average ROAS at 400k: 3.2x
• Marginal ROAS at 400k: maybe 1.4x
• Marginal ROAS at 550k: possibly below 1x - you are now paying the platform for reach that buys nothing

The dashboard shows the first number. Your P&L feels the last one.

How the curve is estimated

An MMM fits a saturation function (we use Hill curves) to each channel from your own spend history. The periods where you pushed spend up and down - even accidentally, even during a botched Black Friday - are what teach the model where the bend is. Variation in spend is data. Flat, steady budgets are the enemy of measurement.

The sweet spot is a moving target

Saturation is not a fixed property of a channel. Creative refreshes, new formats, seasonality, and competitor activity all move the curve. This is why we re-fit models continuously rather than delivering a quarterly PDF: the answer to "should we scale Meta?" changes, and the model should notice before your CFO does.

The practical rule: scale a channel until its marginal return matches the best alternative use of the pound - another channel, or margin. Then stop. The curve tells you where that is; guesswork tells you three months later.