Marketing Mix Modeling & Incrementality: Answer Measurement After Cookies with Confidence

Airbnb once cut a large share of performance marketing and discovered something uncomfortable for every marketer: not every click that gets credit actually creates demand. That is the heart of post-cookie measurement - stop asking “who clicked?” and start asking “what changed because we spent?”

• Marketing Mix Modeling estimates how media, price, promotions, seasonality, distribution and external factors affect sales using aggregate data.
• Incrementality asks the causal question: what extra sales or conversions happened because of the marketing, versus the counterfactual?
• Cookies and device IDs made marketers over-rely on user-level attribution; privacy shifts make aggregate and experiment-based measurement more important.
• MMM is broad and strategic; incrementality tests are sharper and causal; the best teams use both together.
• Core MMM ideas: baseline demand, adstock or carryover, saturation, control variables and response curves.
• Core metrics: incremental lift, iROAS, marginal ROAS, cost per incremental acquisition, MAPE and holdout leakage.
• The interview trap: confusing attributed conversions with incremental conversions. Always ask, “Compared with what counterfactual?”

The Big Picture: Measurement Moves from Tracking People to Measuring Change

Before privacy changes, marketers tried to follow individuals across websites and apps. After cookie loss, Apple ATT, walled gardens and India’s DPDP Act 2023, the stronger mental model is measurement by cause and effect: spend changes, market conditions change, sales change - then we estimate what part was truly caused by marketing.

Core Explanation: MMM Tells You What Drove Sales, Incrementality Tells You What Caused Them

Marketing Mix Modeling, or MMM, is a statistical approach that connects business outcomes such as sales, revenue, app installs or leads to drivers such as TV, digital, search, influencer spend, pricing, promotions, distribution, seasonality and competitor activity.

The powerful idea is simple: sales are not only caused by ads. A mango drink may sell more in summer even without extra media. A beauty brand may spike during a sale event. A fintech app may grow because of a referral campaign, a cricket sponsorship, payday timing and app-store ranking all at once. MMM tries to separate these overlapping effects.

The Five Building Blocks of MMM

Adstock means media has memory. A TV ad, YouTube masthead or influencer burst may affect sales for days or weeks after exposure. Saturation means returns flatten. The first ₹10 lakh on a channel may reach fresh buyers; the next ₹10 lakh may hit the same people again.

Incrementality: The Counterfactual Question

Incrementality is the extra outcome caused by marketing compared with what would have happened without it. The counterfactual is the missing world: same brand, same season, same market conditions, but no campaign or different spend.

You estimate incrementality through experiments such as geo holdouts, conversion lift tests, ghost ads, PSA tests, switchback tests or matched-market tests. The key is to create or approximate a clean comparison group.

The Practical Measurement Stack

A mature team does not choose MMM or incrementality as if only one can exist. MMM says where money appears to work at portfolio level. Experiments test causality. Attribution provides fast but weaker operating signals.

Metrics You Must Be Able to Explain

A Tiny Worked Example: Calculating Incremental ROAS

Suppose a brand runs a geo experiment for a new YouTube and influencer burst. The test region and control region had similar pre-period sales.

The campaign creates ₹60 lakh incremental revenue and ₹21 lakh incremental gross profit. Since spend was ₹20 lakh, it barely clears profit before overheads. The interview answer is not “ROAS is 3, so good”; it is “ROAS is 3, and it must be compared with margin, CAC targets and alternative channels.”

For a brand like Nykaa, demand can be shaped by app pushes, creator content, marketplace visibility, store expansion, sale events, festive seasonality, UPI or COD payment behaviour and repeat cohorts. The so what: in India’s app-heavy, privacy-conscious market under the DPDP Act, aggregate MMM plus clean lift tests can be more reliable than chasing a broken cookie trail.

Definitions You Can Say in One Breath

Marketing Mix Modeling is a statistical method estimating how marketing and non-marketing drivers affect business outcomes over time or markets.

Incrementality is the additional outcome caused by an action compared with what would have happened without that action.

Attribution assigns credit for a conversion across marketing touchpoints observed before that conversion.

A counterfactual is the estimated outcome that would have occurred if the marketing action had not happened.

Case Study: Airbnb’s Shift from Click Credit to Incremental Demand

Airbnb reduced reliance on lower-funnel performance marketing and leaned more into brand-led demand creation, showing why last-click credit is not the same as incrementality.

Situation: Like many digital marketplaces, Airbnb could see plenty of measurable clicks from search and performance channels. The danger was classic: if users were already planning to book, the final paid click might receive credit for demand that brand, word-of-mouth, habit or PR had already created.

The move: During the pandemic period, Airbnb materially reduced performance marketing and later communicated a stronger emphasis on brand marketing and public relations. The strategic logic was not “performance marketing is bad.” It was more nuanced: measure how much demand is truly incremental, then avoid paying repeatedly for demand that would arrive anyway.

The result and lesson: Airbnb continued treating marketing as a mix of brand building, direct traffic, product experience and targeted acquisition rather than a pure last-click machine. The primary driver was a stronger read on incremental demand, supported by a distinctive category position, high direct traffic, earned media, marketplace network effects and a product that people actively search for when travel intent rises.

The memorable takeaway: if a channel gets credit only because it is closest to conversion, it may look profitable while adding little true demand. MMM and incrementality protect the budget from that illusion.

How AI Changes Marketing Mix Modeling and Incrementality

AI does not remove the need for causal thinking. It makes the measurement workflow faster, richer and more dangerous if used without discipline.

Student workflow: Use NotebookLM or ChatGPT with a company’s annual report, investor presentation and recent campaign notes. Ask: “List likely sales drivers, control variables, experiment options and MMM risks for this brand in India.” Then convert the answer into a 60-second interview structure.

Interview Relevance

“Third-party cookies are disappearing. You are advising a D2C brand whose Meta and Google dashboards show strong ROAS, but the CFO doubts whether sales are truly incremental. How would you measure marketing effectiveness?”

Use this sentence in interviews: “I would not kill a channel because attribution looks weak or scale it because last-click ROAS looks strong; I would test its incremental contribution and then use MMM for portfolio allocation.”

Common Mistake

The mistake is treating attributed conversions as incremental conversions. It costs candidates because they sound dashboard-driven rather than business-driven. One-line fix: always ask, “What would have happened without this marketing spend?”

What to Revise Next

Now that you understand how marketers measure causal impact after cookies, revise the two tools that make the idea practical at customer and campaign level.