Can MMM distinguish organic demand from paid advertising's impact?

A farmer who wants to know whether her new irrigation system is worth the money can't just weigh this year's harvest. Some of that crop came from rain. Some came from soil she's been building up with compost for years. Some came from the irrigation. In a wet year, the system might look like a miracle when the rain did most of the work. In a dry year, it might be the only reason anything grew at all, and the harvest still looks disappointing. The harvest total is real, but it doesn't tell her which part of it the irrigation actually earned.

Marketers face the same question every time they look at revenue. Some sales would've happened without any ads running this week, thanks to brand awareness, loyal customers, and marketing you did months ago. Some sales happened because of the campaigns you're paying for right now. If your measurement can't tell those apart, you'll either keep funding campaigns that are taking credit for the rain or cut the ones keeping your revenue alive in a dry spell. Getting this split right is what separates a budget that grows revenue from one that just moves credit around.

Key takeaways

• Organic demand is the revenue you'd earn without current advertising, built from brand awareness, word of mouth, repeat customers, and the lasting effects of past marketing.
• Paid and organic performance are tightly connected, since ads can create later organic searches, direct visits, and marketplace purchases while branded ads often capture demand that already existed.
• Baseline leakage happens when a model gives advertising credit for demand that would've existed anyway, or files real marketing impact under baseline demand, and either mistake can swing ROAS dramatically.
• A marketing mix model estimates baseline and marketing contributions by modeling expected demand and accounting for marketing inputs, seasonality, delayed effects, and uncertainty.
• An "organic" sale only describes how a shopper arrived, not why they wanted to buy, so organic revenue can still include the effects of advertising.
• Prescient treats branded search and Amazon advertising the same way, separating the revenue those ads contributed from other revenue and then estimating how much outside marketing drove the rest.
• Before trusting an MMM's baseline, ask how it models baseline demand, how it handles branded and marketplace ads, and how it separates long-term marketing effects from baseline.

What marketers mean by organic demand

"Organic" gets used loosely, so it helps to pin down what's actually inside it. When marketers talk about organic demand, they usually mean revenue that shows up without a paid ad click attached. That revenue comes from several very different sources.

That last row is the one that matters most for budget decisions. Baseline demand, in the strict sense, is the revenue you'd keep if you turned off your current advertising. Everything above it in the table can feed that baseline, but some of it is still the lingering effect of marketing you paid for, and it'll fade if you stop investing.

Why paid and organic performance are interconnected

It'd be convenient if paid and organic revenue lived in separate boxes. In practice, they're constantly feeding each other, and that's exactly why they're hard to separate. A few patterns show up for almost every brand.

Branded search advertising captures existing intent

When someone searches your brand name on Google, they've already decided they're interested in you. Your branded search ad sits at the top of the page, they click it, and the ad gets credit for the sale. Many of those shoppers would've clicked your organic listing right below it. Branded search can still have a role, like keeping competitors from taking that top spot, but much of its reported revenue is demand that existed before the ad appeared.

Paid social can create later organic searches

The reverse happens too. A shopper scrolls past your Meta or TikTok ad without clicking, then searches your brand a few days later. That search might show up as organic traffic, or it might get captured by your branded search campaign. Either way, the social campaign that created the interest won't get credit in platform reporting.

Advertising can influence direct visits and marketplace searches

The same thing happens outside of search. People who saw your ads may type your URL straight into their browser or search your brand on Amazon. Those sales look organic because no ad click led directly to them, but they're often halo effects of campaigns running elsewhere.

Here's how those patterns tend to distort what you see:

The danger of baseline leakage

Baseline leakage is what happens when a model blurs the line between demand that already existed and demand your marketing created. It's a structural problem—built into how many models split revenue, not something better data alone can fix—and it can push results in either direction.

When models assign existing demand to advertising

The first version of leakage hands advertising credit for sales that would've happened anyway. This is especially common during peak periods like Black Friday and Cyber Monday, when brands raise spend at the same time demand naturally climbs. If a model can't tell the holiday lift apart from the marketing lift, it may conclude your campaigns are far more effective than they are. That can lead brands to overspend during exactly the periods when budgets are biggest.

When models assign real marketing effects to baseline

The opposite mistake is just as costly. Some models absorb real marketing impact into trend or seasonality, which makes your brand look naturally strong and your campaigns look weak. This often happens with campaigns whose effects build slowly or show up in other channels, like awareness campaigns that drive later branded searches. Brands that trust these results may cut the campaigns that are actually keeping their baseline healthy, then watch "organic" revenue fade a few months later.

Why this changes estimated ROAS

Because ROAS is revenue divided by spend, even a small shift in how revenue is split between baseline and marketing can change a campaign's return dramatically. Here's a hypothetical example of how three different baseline estimates could change the story for the same campaign with $100,000 in spend.

Same campaign, same spend, and three very different decisions. That's why the baseline question isn't a technical footnote, since it sits underneath every ROAS figure your measurement produces.

Experiments don't automatically fix this, either. An incrementality test can describe what happened in one window under one set of conditions, but if the model's structure blends baseline and marketing together, calibrating it with a test adjusts the numbers without fixing the underlying problem.

How MMM estimates baseline and marketing contributions

Marketing mix modeling tackles this question by estimating what your revenue would've looked like under different conditions. Instead of following individual shoppers, it starts from the revenue you can actually observe and works backward to estimate how much came from baseline demand and how much came from each campaign.

That process generally involves three pieces of work:

• Modeling expected demand, or what revenue would look like based on your brand's history, trend, and typical patterns.
• Accounting for marketing inputs and other influences, including spend and impressions for each campaign, seasonality, promotions, pricing changes, and outside factors.
• Estimating response patterns and uncertainty, including how each campaign's impact builds, fades, and changes with spend, plus how confident the model is in each estimate.

Model expected demand

The model first needs a sense of what your revenue looks like when marketing isn't the main thing moving it. That includes long-term growth or decline, the weekly rhythm of your sales, and the way demand rises and falls through the year. This becomes the starting point the model compares everything else against.

Account for marketing inputs, seasonality, and other influences

Next, the model looks at how your spend and impressions for each campaign rise and fall over time alongside your revenue. It also accounts for seasonality, promotions, and other events that change demand on their own. The tricky part is that these things rarely move independently. Brands spend more during busy seasons, run promotions alongside campaigns, and launch products with big pushes, so the model has to work out how much each factor contributed when several are moving at once.

Estimate response patterns and uncertainty

Marketing impact doesn't happen all at once. Ads keep working after they run, and their effects fade at different rates for different campaigns. Some campaigns lift other channels too, like a YouTube campaign that drives branded searches. A model that forces one fixed pattern onto every campaign will misplace credit, often pushing slow-building effects into baseline.

Good models also show how confident they are. When several campaigns always scale together or demand shifts for reasons the model can't see, estimates get less certain, and you should know that before making a big budget call.

Why organic sales don't necessarily mean advertising had no influence

This is one of the most common misreads in marketing measurement. A sale labeled "organic" tells you something real, but it doesn't tell you what most marketers assume it does.

Purchase classification and demand origin are different concepts

Purchase classification describes how a shopper arrived at the sale, like an organic search result, a direct visit, or an ad click. Demand origin describes why they wanted to buy in the first place. Those are two separate questions, and the answer to one doesn't settle the other.

A shopper who arrives through an organic result may have been sent there by an ad they saw days earlier. If you read that sale as proof the ad didn't matter, you're treating the path to purchase as the reason for it.

Today's organic demand may reflect earlier marketing

Organic demand also carries the effects of marketing you ran months or even years ago. The awareness you built last spring, the first purchase that turned into a loyal customer, and the review a happy buyer posted after a campaign all keep paying off long after the campaign ended. That's part of why your baseline isn't fixed. If you stop investing in the marketing that builds it, today's healthy organic revenue can shrink over time.

Questions to ask an MMM provider

Every marketing mix model makes choices about how to separate baseline from marketing, and those choices shape every number you'll see. These questions can help you understand how a model handles the problem before you build budgets around its results.

How is baseline demand modeled?

Ask how the model separates the natural rhythm of your business from your marketing, especially during peak periods when the two move together. It's also worth asking how the model's accuracy is checked, since a model can fit past revenue well while still splitting it incorrectly between baseline and marketing.

How are branded search and marketplace ads handled?

Because branded search and marketplace ads like Amazon Sponsored Products sit so close to the purchase, they're where demand capture is most likely to get mistaken for demand creation. Ask whether the model gives these ads full credit for the revenue they touch, or whether it estimates how much of that revenue existed already or came from other marketing.

How are long-term marketing effects distinguished from baseline?

Ask how the model handles marketing whose impact shows up weeks or months later. If delayed effects get cut off too early, that impact gets absorbed into baseline, and the campaigns building your brand will look weaker than they really are.

Where Prescient comes in

Prescient's marketing mix model was built to separate the demand your brand already has from the demand your campaigns create, at the campaign level and with daily updates. We treat branded search and Amazon advertising the same way: our MMM separates the revenue those ads contributed from the rest, then estimates how much of that other revenue was driven by outside marketing like Meta, YouTube, or CTV through halo effects. Delayed effects and decay can vary from campaign to campaign, so slow-building impact doesn't get filed away as baseline.

That gives omnichannel brands a clearer view of which campaigns are building demand and which are mostly capturing it, along with a model confidence score that shows how much weight each estimate can carry. If you want to see how much of your revenue your advertising is actually driving, book a demo.