Can MMM measure marketing when customers buy through multiple channels?

A storm’s rainfall over the mountains doesn't land in one creek. Some of that rain runs into the river on the east side of the ridge, some feeds a lake to the west, and some soaks into groundwater that surfaces in a spring weeks later. Put a gauge on just one creek and you'd conclude the storm barely mattered, even though it raised the water level everywhere downhill.

Marketing works like that storm for brands that sell in more than one place. A TikTok ad or a CTV spot builds interest in your brand, and that interest drains into whichever place the shopper prefers to buy: your website, Amazon, Target, or Walmart. So, can a marketing mix model (MMM) measure marketing when customers buy through multiple channels? Yes, it can, and an MMM is one of the few measurement approaches that can, because it doesn't need to follow the shopper to the register. The catch is that traditional MMMs often treat each place a customer buys as its own separate problem, which splits apart the demand your marketing created.

That matters because budget decisions follow whatever your measurement can see. If only your website's sales are in view, upper-funnel campaigns can look unprofitable while they're driving a meaningful share of your Amazon and retail growth.

Key takeaways

• Yes, an MMM can measure marketing across your website, Amazon, and physical retail, because it works from spend and sales data and doesn't need to track individual shoppers.
• Pixels, last-click, and multi-touch attribution mostly see what happens on your own website, so they miss purchases made on Amazon or in stores.
• Traditional MMMs often model each sales channel separately or lump all revenue together, which either splits marketing's impact apart or hides where it landed.
• Prescient's MMM treats marketing as building one shared pool of demand that shows up in whichever channel the customer chooses to buy from.
• Prescient measures retail halo effects, the revenue a marketing channel drives at a retailer, broken out by retailer.
• Measuring every sales channel in one model keeps teams from cutting campaigns that look weak on website sales alone but are lifting Amazon and retail revenue.

Customers buy wherever is convenient, and most tracking only sees one place

Most shoppers don't take a straight path from an ad to your checkout page. Someone might see a Meta ad, find your brand again on Instagram, read a review, and then buy on Amazon or pick the product up at Target.

Digital attribution tools have a hard time following that. Pixels, last-click, and multi-touch attribution are built around activity on your own website, so they can't see when a paid social or YouTube ad leads to a purchase on Amazon or in a store. As far as those tools are concerned, that sale belongs to no one, or to whichever Amazon ad the shopper happened to click last.

Omnichannel brands also tend to hear two workarounds for MMM:

• One is to build a separate model for each place you sell. 
• The other is to combine all revenue into one total and model that.

Both come with a cost. Separate models split one pool of demand into pieces that each look smaller, and one combined total tells you that marketing worked without telling you where it showed up. Either way, upper-funnel campaigns judged on website sales alone can look like they aren't paying for themselves.

Why traditional MMMs struggle when sales span multiple channels

Traditional MMMs struggle in omnichannel settings because they treat each sales channel as a separate equation, credit retail lift to the wrong source, and become unstable when many marketing inputs are tied to many sales outlets at once.

Each sales channel gets treated as its own island

A traditional MMM typically explains each sales channel's revenue as a baseline plus the marketing assigned to that channel. Marketing doesn't respect those boundaries. A Meta campaign doesn't drive only website sales. It builds demand for your brand that spills into every place people buy. When each sales channel is its own separate equation, that spillover has nowhere to go.

Retail lift gets credited to the wrong source

When an upper-funnel video campaign scales, in-store sales and Amazon search volume often climb at the same time. A model that doesn't represent which campaigns create demand and which ones capture it can credit that retail lift to organic demand or to a retailer's own promotion. The video campaign that created the demand ends up looking unprofitable.

More inputs and more outlets make the math unstable

Connecting a dozen marketing channels to several sales outlets, such as your site, Amazon, Target, and Walmart, multiplies the number of relationships a model has to estimate. Much of that marketing spend also moves together, especially during promotions. A traditional model can't reliably tell which ad drove which outlet's sales, so small changes in setup can shift large amounts of credit between your website and your retailers.

How Prescient measures marketing across every place you sell

Prescient measures omnichannel marketing by modeling it as one connected system in which marketing builds a shared pool of demand, and each sales channel is a place that demand turns into revenue. Our MMM works backward from what can be observed—your spend and your sales in each channel—to estimate what drove those sales.

Marketing builds one pool of demand that every channel draws from

Instead of building a separate model for your site, Amazon, and each retailer, our model treats awareness and buying intent as something your marketing builds across the whole market. Your sales on each channel are different views of that same underlying demand. A shopper buying on Amazon instead of your website doesn't make the ad that persuaded them any less effective.

Demand-building and demand-capturing campaigns are treated differently

Our MMM represents the difference between campaigns that create demand and campaigns that capture it. Upper-funnel tactics, such as YouTube, CTV, and TikTok, build overall demand. Lower-funnel tactics, such as branded search and Amazon Sponsored Products, capture demand at a specific place to buy. That's what lets Prescient credit a campaign for its halo effects, which is the revenue it drives indirectly through another channel. Upper-funnel campaigns tend to produce more halo effects, but any campaign can.

Retail halo effects are measured by retailer

Prescient measures the halo effects that flow from digital marketing channels to physical retailers, broken out by marketing channel and retailer. In practice, that means seeing how much revenue your TikTok campaign drove at Target versus Walmart, alongside what it drove on your own site and on Amazon.

Response isn't forced to flatten

Our model doesn't push every campaign onto a fixed saturation curve. If more digital spend keeps lifting sales across your site and your retailers at a steady rate, the model lets that response stay close to a straight line, and it shows diminishing returns only when your data supports them.

What affects how clearly an MMM can read omnichannel impact

An MMM reads omnichannel impact most clearly when sales data from every outlet comes in regularly and marketing activity varies enough to tell effects apart. Four factors make the biggest difference.

On that last row, Prescient can incorporate a test result into its MMM as a prior. It shows you how the model performs with and without the test, and you make the final call.

But there is a limit here: A retailer that doesn't share sell-through or point-of-sale data can't be measured directly, so the read is only as complete as the retail data coming in.

What omnichannel measurement changes for the business

Measuring every sales channel in one model changes which campaigns look profitable, because their full return finally shows up in one place. That shows up in three ways.

• One view of marketing return. Marketing and finance can look at the same numbers across your website, Amazon, and retail instead of reconciling separate reports that each tell part of the story.
• Protection for the campaigns driving retail growth. Awareness campaigns that look weak on website sales alone get credit for the Amazon and retail revenue they drive, so they're less likely to be cut.
• Better footing for retail and Amazon planning. Knowing which marketing lifts which retailer gives you evidence to bring to retail co-op conversations and to Amazon growth targets.

Where Prescient comes in

Prescient's MMM measures omnichannel brands in one model across DTC, Amazon, and retail, including Target, Walmart, Ulta, Sephora, Nordstrom, and other retailers that share sell-through or point-of-sale data. It reports campaign-level Modeled ROAS and breaks out retail halo effects by marketing channel and retailer, with online models updating daily and retail models updating weekly.

From there, our Media Forecaster feature shows what's likely to happen when you scale a channel's spend or shift budget into it from other channels, so you can plan around total business growth instead of website sales alone. If you'd like to see what your marketing is driving across every place you sell, book a demo and our team of experts will walk you through how the platform uncovers your performance.

FAQs

Can a marketing mix model measure Amazon and retail sales?

Yes. A marketing mix model can measure Amazon and retail sales because it works from aggregate data, such as spend and sales by channel, and doesn't depend on tracking shoppers. The model needs sales data from each outlet, which for most retailers means sharing sell-through or point-of-sale data.

Why can't pixels or last-click attribution measure retail sales?

Pixels and last-click attribution rely on tracking activity on your own website, so they can't see a purchase made in a store or on another company's site. When a shopper sees your ad and later buys at Target, those tools have no record connecting the two.

Should I build separate MMMs for DTC, Amazon, and retail?

Separate models tend to undercount marketing, because each one sees only the slice of demand that turned into sales in its own channel. A single model that includes every sales channel can credit a campaign for all the revenue it drives, wherever the customer chose to buy.

What is a retail halo effect?

A retail halo effect is the revenue a marketing campaign drives at a retailer, even though the campaign never sent anyone there directly. A TikTok ad that leads someone to buy your product at Walmart a few days later is one example.

What retail data does an MMM need?

An MMM needs sales data from each retailer, usually sell-through or point-of-sale data, along with your marketing spend and your direct and Amazon sales. Most retailers report weekly, which is enough for an MMM to measure retail impact.