Adstock and Carryover Effects
What you'll learn: How to model the delayed and decaying impact of advertising that persists long after an ad is shown.
The Problem with Immediate-Only Effects
When you run an ad campaign, its influence doesn't vanish the moment it stops airing. A TV commercial you saw yesterday might still influence your purchase decision today or next week. Similarly, brand awareness builds gradually over time. If you treat advertising as having only an immediate, one-time effect in your Media Mix Model, you'll dramatically underestimate its true value.
What is Adstock?
Adstock (also called "advertising stock") is a transformation that captures two key phenomena:
- Carryover effect: Advertising impact persists and decays over time, like a drug slowly leaving your bloodstream
- Delayed response: Customers may not act immediately; there's a lag between exposure and conversion
Think of it like watering a plant. The water you give today doesn't just help today—it carries over into tomorrow, but with diminishing effect as the soil dries out.
The Mathematical Transformation
Adstock transforms your raw advertising spend into an "effective exposure" metric:
Adstock_t = Spend_t + λ × Adstock_(t-1)
Where λ (lambda, 0 to 1) is the retention rate—the fraction of yesterday's effect that carries into today. A λ of 0.7 means 70% of yesterday's advertising impact is still active today, 49% from two days ago (0.7²), and so on.
Why It Matters in MMM
Without adstock transformations, your regression model sees only same-day correlations between ad spend and sales. With adstock, you correctly attribute results to campaigns that build momentum over weeks, capturing the true ROI of your marketing investments.
Key Takeaway: Adstock transformations model advertising's realistic decay pattern over time, ensuring your MMM doesn't penalize channels with delayed or sustained effects.