The Adstock Effect: Why Last Week's Ads Are Still Making You Money

You ran a big YouTube campaign last week. This week you paused it. But conversions from YouTube-initiated demand are still trickling in. That is the adstock effect, and ignoring it leads to bad budget decisions.

Adstock (sometimes called ad carryover or the lag effect) is the idea that advertising impact does not happen instantly and does not disappear instantly. An ad you see today might influence your purchase next week or even next month.

A Simple Example

Imagine you spend $100K on TV ads in week 1. In a world with no adstock, all the impact happens in week 1 and the effect drops to zero in week 2. But that is not how people work.

With an adstock decay rate of 0.7 (which is typical for TV), the effect in week 2 is 70% of the week 1 impact. Week 3 is 49% (0.7 x 0.7). Week 4 is 34%. It takes about 6-8 weeks before the effect decays to near zero. So that $100K in week 1 is still generating results a month later.

Without modeling adstock, you would attribute all $100K of impact to week 1 and conclude that weeks 2-4 had zero TV contribution. Your budget model would say "TV only works when we are actively spending" when the reality is TV keeps working long after the ads stop.

Decay Rates by Channel

Different channels have different adstock decay rates, and the differences are large:

Notice the pattern: channels higher in the funnel (TV, YouTube) have slower decay. They build brand memory that lasts. Channels lower in the funnel (paid search) have fast decay because they capture existing demand rather than creating new demand.

Paid search has almost no adstock because the intent already exists. Someone searches "buy running shoes" and clicks your ad. That conversion is immediate. There is no carryover from last week's search ad.

Why Adstock Matters for Budget Timing

If you are pulsing your TV or YouTube budget (two weeks on, two weeks off), adstock means the "off" weeks still benefit from the "on" weeks. This can actually be an efficient strategy because you get partial impact during off weeks without paying for it. But only if you understand the decay rate.

For paid search, pulsing makes no sense. The moment you stop spending, the impact stops. Search budgets should be consistent and tied to demand patterns (seasonal, day-of-week).

For Meta and TikTok, moderate pulsing can work if you maintain enough baseline spend to keep the algorithm warm. Going from $50K/week to $0 to $50K/week hurts performance because the ad platforms re-enter learning phases. Going from $50K to $25K to $50K is fine.

How MMM Models Adstock

In a marketing mix model, adstock is modeled as a geometric decay transformation applied to the spend data before regression. For each channel, the model estimates a decay parameter (lambda) that best fits the relationship between spend and outcome.

Google Meridian (the engine behind Spendmix) uses a Hill function combined with adstock to model both saturation and carryover simultaneously. The model learns the optimal decay rate for each channel from your data, so you do not need to guess.

The output tells you two things: how much of this week's revenue is attributable to this week's spend, and how much is carryover from previous weeks. That split is critical for understanding true channel efficiency.

Without adstock modeling, you misattribute lagged conversions to whatever else happened that week (seasonality, other channels, organic traffic). The entire budget allocation shifts in the wrong direction. Adstock is not an academic concept. It is a practical measurement requirement. See how it shows up in a real model output in our demo report.