How to measure marketing when most sales happen in stores
If most of your sales happen in a store, your ad platforms can't see most of your sales. Measure marketing against all of your revenue, in store and online, with a media mix model. Use platform reports and store-visit estimates as signals for running campaigns, not as the answer to whether advertising works.
Key takeaways
- Ad platforms see clicks and online conversions. For a store-first business, that's a small slice of revenue.
- Store-visit numbers from ad platforms are estimates, and none of them can say which sales would have happened anyway.
- A media mix model measures advertising against total sales, and separates the baseline you'd have without ads.
- Clean, consistent data matters more than a fancy method: weekly sales, spend by channel, and a calendar of promotions.
- Use the model to plan budgets, and refresh it often enough to keep up with your decisions.
Why ad platform reports fall short for stores
Digital ad platforms were built around clicks. They're very good at counting what happens on a website or in an app. For a business that sells mostly in stores, three gaps open up:
- They can't see the register. A shopper sees an ad on Monday and buys in a store on Saturday. Most of the time, no platform connects those two things.
- Store visits are estimates. Some platforms model how many people visited a store after seeing an ad. That's useful for bidding, but it's an estimate of visits, not sales. See YouTube ads for store visits.
- No platform can say which sales would have happened anyway. Your regulars were coming in regardless. Each platform still counts the ones who saw its ads.
The result is a familiar problem: every partner's report looks good, the reports add up to more than you sold, and nobody can say whether advertising is working.
What to measure instead
Measure against all of your sales, in store and online, because that's the number that pays the bills. Then ask the question that matters: how much did advertising add on top of the sales you would have made anyway? That's incrementality, and for a store-first business the most practical way to estimate it across every channel is a media mix model.
A media mix model doesn't need to follow anyone from ad to store. It looks at weekly sales next to weekly spend by channel, controls for seasons, holidays, promotions and the economy, and estimates what each channel added. It can include TV, radio, out of home and print alongside search and social, as long as you have weekly spend for them and that spend moves around over time.
The toolkit, and what each piece is for
- Site analytics (Google Analytics, Google Tag Manager). What people do online: visits, store-locator use, online orders. Keep it clean so online results are right.
- Platform reports. Running campaigns day to day: which audience, which ad, which keyword. Treat their sales claims as a starting point.
- Lift tests. A clean answer for one channel at a time, when the platform or a geo test allows it.
- A media mix model. The budget question: what every channel adds to total sales, and where the next dollar should go.
- Dashboards. One shared view of all of it, so the team reads the same numbers.
Getting your data ready
Most of the work is in the data, not the math. Here's what a model needs:
- Sales by week, in store and online, ideally by region. Two or more years is a common starting point.
- Spend and impressions by channel, by week, for every paid channel, offline ones included.
- A calendar of promotions, price changes, store openings and closings, and holidays.
- Consistent naming in your ad accounts, so spend lands in the right channel every time.
Turning the model into decisions
A model only helps if it changes what you do. The useful outputs are simple: does advertising add sales, which channels earn the most on the next dollar, and how long the effect takes to show up. Refresh it often enough to keep up with your plans. The model informs the calls. The people who buy each channel still make them.
A retailer's first real answer
A Florida retailer asked us a simple question: is our advertising actually working? Most of their sales happen in their stores. Ad platforms could only estimate store visits, and partners' own reports claimed returns that looked too good to be true.
We built a media mix model, with Google's open-source Meridian, that measures advertising against all sales, in store and online, and we update it every month. For the first time, it showed that their advertising was working: it estimates the mix adds more than $10M a year in incremental revenue. We used it to make the case for spending more, and the next year's digital ad plan came in more than 20% higher. Now the client plans the mix with it. See the full case.
Common questions
Can you measure in-store sales from digital ads?
Not person by person, in most cases. But you can measure them in total: a media mix model compares how store and online sales moved with what you spent in each channel, and estimates what the ads added.
Do we need a loyalty program or customer-level data?
No. A media mix model works from totals, like weekly sales and weekly spend. Customer-level data can add detail, but the model doesn't depend on it.
Our ad partners already report great returns. Why measure again?
Partner reports are scored by the partner, from its own point of view, and they can't see which sales would have happened anyway. In our retail case, partners' own reports claimed returns that looked too good to be true. A model of total sales gives you one consistent yardstick for every channel.
