New entrants make media mix modeling faster and more accessible

Advances in cloud computing and machine learning have helped to democratize marketers’ abilities to use MMM. There’s also just a need for marketers to have greater transparency into the effectiveness of media planning and media measurement. Because of the greater pressures on marketers to spend more efficiently and effectively — take your pick of the reasons for that, from financial and inflationary pressure to greater competition — agencies need to ensure their clients can more clearly understand the outcomes from the work they do on marketers’ behalf.

One such effort, an MMM tool called FutureSight, has made use of cloud computing and machine learning to create a software-as-a-service version for mid-market clients to make use of. An outgrowth from independent media agency Mediastruction, it’s largely the brainchild of Marilois Snowman, Mediastruction’s founder and CEO.

Snowman emphasized the transparency of FutureSight as an important element for clients using it. “The difference is it’s not a black box, it’s a glass box. So the model that goes to attribute your media is curated and customized for the brand,” explained Snowman. “A great way to think of it like it could actually be an ensemble model, which is like a combination of multiple different types of models. The algorithm for a car dealership might not be appropriate for an online retailer, and so we adjust that. That’s the glass box part.”

“It’s a really valuable tool in that it allows you to not only say this is what has happened, but this is what will happen,” said Jen Marino, a fractional CMO who has used FutureSight for work with several clients, including Rockland Trust Bank and health care clients. “And the the MAPE [mean absolute percent error] has in my in my experience has always been very low, less than 10%, so it’s very accurate in terms of what it can forecast.”

Marino said her time as CMO of Rockland Trust resulted in savings and efficiency thanks to the optimization FutureSight enabled — but it’s also helped her increase spending because of the effectiveness uncovered.

“We were able to demonstrate the impact of marketing so that we could see what channels worked most effectively,” added Marino. “We made decisions about the investment across different vehicles based on that model and and what it told us. So we were more efficient with the total budget year over year 30-plus percent because we were able to optimize each of those channels with a full attribution.”

At media agency Media Matters Worldwide, machine learning and AI play an increasing role in developing metrics and modeling. Its agile mix modeling (AMM) provides clients with weekly campaign ROI readouts across channels, as traditional attribution becomes less useful with walled gardens and third-party cookie deprecation.

For instance, compared to traditional methods, AMM uses automated data collection every 24 hours instead of manual data collection, which is typically done on a quarterly basis. The time needed to process the manually-collected data also requires an average of four weeks and comes with a high chance of errors, explained Sara Owens, svp of analytics at Media Matters Worldwide. Forecasting through AMM also becomes more accurate, and the models train themselves in hours — compared to weeks with traditional MMM that requires manual modeling that takes anywhere from four to six weeks to train.

Client Sierra Nevada Brewing Company, which was heavily investing in social media like Facebook that doesn’t allow third-party measurement, didn’t have a single measurement solution for campaigns that showed how each channel and partner was driving sales or ROAS.

Using AMM to capture what was previously unmeasured behind walled gardens, as well as a model that refreshed monthly to give timelier and more accurate metrics of the campaign’s incremental revenue and ROAS by channel, the brand was able to get “a comprehensive understanding of their media effectiveness, leading to more informed decision-making, optimized budget allocation, and higher ROAS and revenue incrementality,” Owens added.

AMM showed that one the brand’s highest-performing channels was local radio, which was a small part of its media mix before. It also provided insight to the client about paid social underperforming, while streaming music was performing well. The tool ultimately led this client to use an optimized media mix and forecasted revenue with double the ROAS.

Meanwhile, Keen Decision Systems, a firm that focuses on marketing mix modeling, which, as noted above, delves into the broader marketing mix than just media channels, is offering a free trial use for the first time for its five-year SaaS marketing mix modeling tool — another effort to bring the discipline to small marketers, said Brad Keefer, the firm’s chief revenue officer.

“If we put a tool that is intuitive in your hands and you have the ability to gain insight, you can see that, in fact, even though you’re a brand that spends $500K a year, the model will produce results that are valuable to your business,” said Keefer, who said clients include Athletic Brewing and beverage brand Poppi. “The folks that we talk to are looking for a tool that can help them forecast, help them hold their agencies accountable, that can measure and prove their value and can do it at a faster pace.”