What is adstock?
An ad seen on Tuesday can still shape a purchase the following week. Adstock is the running total of that remembered advertising. Each week it gets new spend added and loses part of what was there before.
The idea comes from Simon Broadbent, a British advertising researcher who introduced it in the late 1970s. It's now a standard part of marketing mix modeling, because sales in any week reflect several weeks of advertising.
How does adstock work?
The simplest version is geometric adstock: adstock this week = spend this week + decay rate × adstock last week. With a decay rate of 0.5, half of last week's effect carries over, then a quarter the week after, and so on.
| Decay rate | Half-life | What it suggests |
|---|---|---|
| 0.3 | Under a week | Most of the effect lands right away, as is common for search |
| 0.5 | One week | A moderate carryover |
| 0.7 | About two weeks | A longer memory, often seen in video and TV |
| 0.9 | About six and a half weeks | Effects that build slowly and last, as brand campaigns aim for |
Half-life is the easiest way to read a decay rate: the number of weeks until half the effect is gone. Some models also allow a delayed peak, for channels whose effect builds before it fades. Models usually apply adstock first, then a saturation curve, so carryover and diminishing returns are estimated together.
What is an example of adstock?
A TV flight that keeps selling
A brand runs TV for four weeks, then stops. Sales stay above normal for another five weeks before settling.
A model without adstock sees spend of zero in those five weeks and credits the extra sales to whatever else was running, often search. With adstock, the model gives TV its share of those weeks, and search's credit shrinks to what it actually earned.
Adstock vs. diminishing returns
| Adstock | Diminishing returns | |
|---|---|---|
| What it describes | How long an ad's effect lasts | How much each extra dollar adds |
| Dimension | Time | Spend level |
| Key parameter | Decay rate or half-life | Saturation point and curve shape |
| Planning question | When will this spend pay back? | How much more is worth spending? |
Why does adstock matter?
Carryover decides which channels get credit. Get it wrong and a model rewards whatever ran last, which usually means lower-funnel channels, and it undervalues work that pays back slowly.
- Fairer credit. Brand channels get credit for sales that arrive weeks after they run.
- Better timing. Teams can see how long a flight keeps working and plan the next one around it.
- Honest cuts. Turning off a long-memory channel looks free for a few weeks, then sales slide. Adstock shows that lag before it happens.
Decay rates are estimated, so they carry uncertainty. A model that sets them by hand, or lets them run to extremes, can fit history well and still plan badly. Calibrating against holdout or geo-lift tests keeps them honest.
How does Alli do this?
Alli by PMG's Scenario Planner forecasts revenue with a media response model built on historical sales and media investment by channel, with diminishing-return curves that show where each channel's next dollar stops paying back. A second model adds the longer-term value of the brand from Alli Brand Equity signals.
Two-stage estimation is designed to keep forecasts steady as new data lands, and every forecast is a modeled estimate with its assumptions in view, so teams can compare options and explain the choice to finance.
Explore Alli Scenario PlannerAdstock FAQ
What is adstock in marketing?
Adstock is the carryover effect of advertising: the part of an ad's impact that lasts into later weeks. Mix models use it so sales are credited to the weeks of advertising that drove them.
What is a good adstock decay rate?
There isn't one right number. Decay rates differ by channel, brand, and market, and should be estimated from your data. Search usually decays fastest, and brand-building video and TV often decay more slowly.
How do you calculate adstock?
With geometric adstock, take this week's spend and add the decay rate multiplied by last week's adstock. Repeat for every week. The decay rate is usually estimated by the model.
What is the difference between adstock and lag?
Lag means an effect starts later. Adstock means an effect lasts longer. Some models combine the two, letting an ad's effect build to a peak and then decay.
Last updated Back to the glossary