Adstock (also called carryover) is the phenomenon where advertising effects persist and gradually decay after the ads stop running, meaning the impact of this week's ad spend extends into future weeks, diminishing over time.
The concept was formalized by Simon Broadbent in the 1970s and remains one of the foundational elements of media mix modeling.
Advertising effects don't stop when the campaign ends
Why Advertising Effects Don't Stop Immediately
When someone sees an ad, they don't always act on it immediately. They might notice the brand, think about it occasionally over the next week, and then purchase ten days later. The ad created awareness and consideration that took time to convert into action. That delayed response is what adstock captures.
Two things happen after advertising runs:
- Immediate effect: the lift in sales (or awareness, or consideration) during the period the ads ran.
- Carryover: the continuing, decaying effect in subsequent periods, even after the campaign has ended.
A television campaign that ends on a Friday doesn't have zero effect on Saturday. Its influence continues, just at a diminishing rate.
Carryover Rates by Channel
Not all channels carry over equally. The rate of decay varies significantly depending on the medium and the nature of the message.
Long carryover (4–8+ weeks): Television, out-of-home, radio, and podcast advertising tend to have long carryover effects. Brand-building messages create associations and mental availability that persist. A TV spot viewed once can influence a purchase made three weeks later.
Medium carryover (1–3 weeks): Digital display and social media advertising typically fall in this range. The impression is less persistent than broadcast but more durable than a direct-response click.
Short carryover (days to 1 week): Search advertising has very low carryover. A search ad triggers a response in the moment of intent or not at all, there is little residual effect after the campaign ends.
Near-zero carryover: Promotional discounts and flash sales generate immediate response. Once the promotion ends, the effect ends with it.
Why This Matters for Measurement
If you pause a channel and sales don't immediately drop, that's not proof the channel wasn't contributing. The carryover effect may be sustaining baseline sales for several weeks after the campaign goes dark.
This is a common and costly mistake in budget experiments. A brand pauses television advertising for a month, doesn't see an immediate sales decline, and concludes the TV wasn't working. In fact, the TV carryover was still generating lift, it just took another 4–6 weeks to fully dissipate.
The reverse is also true: when a new campaign launches, it may take several weeks before the full effect is visible in the sales data, because carryover from the prior period is mixed in.
How MMM Models Adstock
Media mix models need to account for adstock or they will misattribute the lingering effect of past advertising to whatever else was happening in that period, price changes, seasonality, or other channels.
There are two common mathematical approaches:
Geometric decay: The simplest and most widely used. Each period, the carryover effect multiplies by a retention rate (typically 0.3–0.8). A retention rate of 0.5 means each week retains 50% of the previous week's carryover. This creates a steadily declining curve.
Weibull (or Hill) transformation: More flexible. Allows for a delayed peak effect, the advertising effect actually increases for the first few weeks before declining. This better captures how some brand campaigns build awareness before driving action. Google Meridian and Meta Robyn both use variations of this approach as their default.
The choice of adstock specification, the retention rate and the decay shape, significantly affects MMM output. A model that underestimates TV carryover will underestimate TV's contribution and potentially overstate channels with more immediate response.
Adstock and Budget Decisions
Understanding adstock changes how you think about spending patterns. If you cut a high-carryover channel, you won't feel the full effect for weeks. If you significantly increase spending in a high-carryover channel, the returns won't be fully visible for weeks either.
This has practical implications for how you interpret short-term tests and how you plan spending during seasonal peaks. Advertising before a peak period, not just during it, can be more effective for high-carryover channels than last-minute spending.