Course contentsShow
← Data Science
Lesson 1740 of 2,14541. Marketing and Growth AnalyticsFree lesson

Saturation Curves and Diminishing Returns

Capturing nonlinear responses where additional spend yields decreasing marginal returns.

Saturation Curves and Diminishing Returns

What you'll learn: How to model the nonlinear relationship where additional marketing spend produces progressively smaller gains in response.

The Core Idea

In marketing, spending the first 10,000onadsmightgenerate1,000conversions,butthenext10,000 on ads might generate 1,000 conversions, but the next 10,000 rarely delivers another 1,000—it might yield only 600. This is diminishing returns, and we capture it mathematically with saturation curves.

Think of watering a garden: the first watering transforms dry soil into productive ground. The second helps, but less dramatically. By the tenth watering in one day, you're just creating puddles—no additional growth.

Why This Matters in MMM

In Media Mix Modeling, assuming a linear relationship (double the spend = double the results) wildly misrepresents reality. Channels saturate as you:

  • Exhaust your target audience
  • Trigger ad fatigue
  • Hit inventory limits (finite ad placements)

Saturation curves let you answer: "At what spend level do we stop getting good returns?"

Common Saturation Functions

The most popular is the Hill function (also called logistic or S-curve):

Response = (Spend^α) / (Spend^α + β^α)
  • α controls the steepness (how quickly you saturate)
  • β is the half-saturation point (spend level reaching 50% of max response)

Other options include logarithmic or exponential decay forms, all capturing the same principle: each incremental dollar works less hard than the last.

Practical Application

After fitting your MMM with saturation curves, you can identify the "sweet spot"—the spend level where marginal ROI drops below acceptable thresholds. Spend beyond that wastes budget; spend below it leaves opportunity on the table.

Key Takeaway: Saturation curves mathematically encode diminishing returns, allowing MMM to accurately reflect how channels perform at different spending levels and optimize budget allocation accordingly.