Can MMM predict what will happen if you increase ad spend?

Fertilize a lawn in April and it'll look great by May. That leaves you with two questions before you buy a bigger bag. How much of that growth came from the fertilizer, and how much was simply spring? And would twice as much fertilizer give you twice the lawn? Get either answer wrong and you'll pay for product the grass can't use, or hold back on something that was working.

Scaling ad spend comes down to the same two questions, and they're what you're asking a marketing mix model (MMM) to answer. So, can an MMM predict what will happen if you increase ad spend? Yes, it can, but the prediction is only reliable if the model separates ad-driven revenue from seasonal demand and doesn't assume in advance that every channel is close to maxed out. Many traditional MMMs fall short on both.

Whether you're doubling one campaign's budget or deciding how much growth money to release for Q4, the forecast behind the request decides whether that money goes to work or goes to waste.

Key takeaways

• Yes, an MMM can predict what happens when you increase ad spend, but how reliable that prediction is depends on how the model handles saturation and seasonality.
• Traditional MMMs use fixed saturation curves that build diminishing returns in by design, so they can label a channel "saturated" when it still has room to grow.
• During peak periods, traditional MMMs can mistake seasonal demand for ad efficiency, a problem called baseline leakage, and recommend overspending as a result.
• Prescient's MMM doesn't force a ceiling onto a channel or campaign, and it models how added spend builds demand that pays off over time.
• In Prescient's peak-season simulation research, following our model’s budget recommendations fell about 6% short of the best possible outcome, compared with roughly 32% and 45% for two open-source MMM approaches.
• Any MMM's predictions are firmest near spend levels you've already run, and they get less certain the further you scale beyond them.

Scaling spend is a bet on a forecast

Every spend increase rests on a prediction, whether or not anyone writes it down. Leadership's version of the question is usually blunt: "If we put $500,000 more into this channel, how much additional revenue will it generate?"

The forecasts behind that answer tend to fail in two directions:

• Sometimes a model recommends scaling a channel and efficiency collapses once the budget goes up.
• Other times a model calls a winning channel "fully saturated" when there's plenty of room left. 

Both are expensive. The first sends capital into a wall nobody could see, and the second leaves profitable revenue unclaimed because the budget was capped too early.

Why traditional MMMs get spend increases wrong

Traditional MMMs mispredict spend increases for three reasons: they build saturation into the math, they confuse seasonal demand with ad performance, and they use fixed decay rates that miss delayed effects. Each one pushes the forecast in a predictable direction.

Built-in saturation curves cap growth too early

Traditional MMMs fit spend to fixed curve shapes, such as Hill or Weibull functions, that always flatten. Diminishing returns are guaranteed by the formula, whether or not your data shows them. Where the curve flattens is inferred from where your past spend happened to stop, so the model can treat your budget ceiling as the channel's ceiling.

In Prescient's own research on campaign-level data, a straight-line fit came closer to the observed relationship between spend and revenue, with about 33% error, than a Hill curve at about 53% or a Weibull curve at about 331%. The practical result of forcing a curve where one doesn't belong is chronic underspending on things that are working.

Seasonal demand gets mistaken for ad efficiency

Brands scale spend when demand is already high, such as Black Friday or peak season. Because spend and demand rise together, a traditional model can't reliably tell how much revenue came from ads and how much came from shoppers who were coming anyway.

When that natural demand gets credited to ads, it's called baseline leakage. The channel looks more efficient than it is, the model predicts big returns if you scale further, and the brand overspends. This is a structural problem in the math, which is why running an incrementality test doesn't make it go away.

Fixed decay rates miss the delayed payoff of scaling

Ads keep working after they run, and traditional models usually capture that with one fixed rate of decay. Scaling an upper-funnel video campaign doesn't pay back that neatly. It builds demand that shows up over the following weeks, often through other channels like branded search, and a fixed decay rate under-predicts that return.

How Prescient predicts what happens when you scale

Prescient's MMM predicts scaling outcomes by modeling marketing as one connected system that changes over time, with no pre-set ceiling on any channel or campaign. Three design choices matter most for spend predictions.

Response isn't forced to flatten

Our model doesn't push spend through one universal saturation curve. If returns hold steady across the range you've spent, the model lets the response stay close to a straight line, and it shows saturation only when your data supports it. Saturation also isn't always one smooth bend. A campaign can dip in efficiency and then reach another efficient range at higher spend, and a flexible model can show you that instead of telling you to back off.

Added spend feeds a system instead of a static equation

In Prescient’s MMM, more spend doesn't simply add sales to a formula. It builds awareness and demand, which then affect how well your other campaigns convert. That's how Prescient measures halo effects—the revenue a campaign drives indirectly through another channel—alongside the direct response. Any campaign can produce them, though upper-funnel campaigns tend to produce more, and they're a large part of what a scaling decision returns.

Ad-driven revenue is separated from seasonal demand

Prescient models seasonal demand and marketing together instead of treating them as independent pieces that can be cleanly split after the fact. In Prescient's research, that approach sharply reduced baseline leakage during peak periods, so efficiency estimates held up when spend and demand spiked at the same time.

What flawed scaling predictions cost at peak season

In Prescient's simulation research, following a traditional MMM's budget recommendations at peak season left roughly a third to nearly half of the achievable outcome unrealized. The study used a simulated marketing environment where the best possible budget is known, so each model's advice could be scored against it.

The two traps from earlier both show up here. The model that overspent had credited ads with holiday demand, and the model that underspent had underestimated how far marketing could scale. Total budget is only part of the picture, too. How the budget is split across channels matters just as much, which is why the model that was closer on total spend still had the larger shortfall.

Where spend predictions are strongest, and where they get less certain

An MMM's spend predictions are strongest close to spend levels you've already run, and they get less certain the further you go beyond them. Three factors—spend range, funnel position, and test data—shape how much weight a prediction can carry.

On that last point, Prescient can incorporate an incrementality test into the model as a prior. It shows you how the model performs with and without the test included, and you make the final call.

What reliable spend predictions change for the business

Reliable spend predictions turn a budget increase from a judgment call into a decision you can defend. That shows up in three ways.

• Forecasts finance can pressure-test. A revenue projection with a stated range gives a CFO something concrete to challenge, which is a much stronger position than a single hopeful number.
• What-if planning before the money moves. Growth teams can compare spend increases across channels and campaigns first, and commit budget second.
• Scaling without the two traps. Brands can push harder on what's working without being stopped early by a forced saturation curve or misled by seasonal demand.

Where Prescient comes in

Prescient's MMM measures omnichannel brands across the places they sell, including their own site, Amazon, and retail partners, with models that update daily. Media Forecaster builds on that measurement to show what's likely to happen when you scale a channel's spend or shift budget into it from other channels.

Each forecast comes with a confidence band, and a separate model confidence rating shows how firm the read is, so you can match the size of a move to your comfort with risk. If you'd like to see what scaling would look like for your own channels and campaigns using the Prescient platform, book a demo.

FAQs

Can a marketing mix model predict the impact of increasing ad spend?

Yes. An MMM can predict what happens when you increase ad spend by estimating how revenue responds at different spend levels. The prediction is only as good as the model's assumptions, though. Models that force every channel onto a fixed saturation curve, or that can't separate ad-driven revenue from seasonal demand, tend to get spend increases wrong.

Why would an MMM say a channel is saturated when it isn't?

An MMM can show false saturation because traditional models use curve shapes that always flatten, regardless of what the data shows. The point where the curve flattens is inferred from the highest spend the model has seen. If you've never spent beyond a certain level, the model can read your budget limit as the channel's limit.

What is baseline leakage in MMM?

Baseline leakage is when an MMM credits ads with revenue that came from natural demand, such as seasonality or holidays. It happens most during peak periods, when brands raise spend at the same time shoppers are already buying more. The result is inflated efficiency estimates and recommendations to overspend.

How far beyond current spend can an MMM predict?

An MMM predicts most reliably near spend levels you've already run. The further a planned budget sits from anything in your history, the more the model is extrapolating and the wider the range of likely outcomes. Scaling in steps gives the model new evidence at each level and keeps predictions grounded.

Do I need incrementality tests to trust an MMM's spend predictions?

No. An incrementality test can be a useful input, but it reads one place at one moment, so it doesn't prove what will happen at a different time or a larger budget. Prescient can incorporate a test as a prior and shows how the model performs with and without it, leaving the decision to you.