3 ways to get more out of marketing mix modeling amid cookie loss & regulatory change

I couldn’t resist the opportunity to continue practicing my budding Spanish, so asked him what he thought of the guidebooks and tourist apps that enabled local businesses to advertise to travelers. He shrugged and said he didn’t use them, predominantly because he felt he had no way of telling if the ad spend would be worth it.

For a local family-operated business, it’s certainly a challenge to measure this kind of impact – not to mention the impact of word-of-mouth. Yet, multi-billion-dollar global brands struggle with marketing measurement, too.

While measurement is vital in proving marketing’s value to the rest of the organization and optimizing campaigns, channels and tactics, it has become increasingly difficult in recent years due to a few key forces.

For one, efforts are more indirect: Marketing influences customer attitudes, which then cascades into customer action and purchase outcomes. And a volatile, uncertain world only makes influences even more indirect.

Secondly, and most notably, data depreciation, the phasing out of third-party cookies and regulatory shifts put a damper on marketing measurement efforts. These realities have chipped away at directly-linked digital advertising and measurement approaches like multi-touch and view-through attribution (Google offers more than 6.9m results for the term ‘data deprecation’ alone, so we won’t cover them here).

Both marketers and measurement providers have returned to marketing mix modeling (MMM) as a solution to these challenges. MMM measures the impact of advertising, promotions and brand across channels while accounting for the influence of external factors outside of a brand’s control, such as competitive activity or consumer sentiment. When used properly, it’s an essential tool for guiding budget decisions, which are increasingly scrutinized as the threat of an economic recession looms.

Examples of mix modeling insights include: the forecast impact of online marketing spend on physical store sales; the optimal flighting patterns or sequences and volumes of ad exposure required to drive sales; and the optimal total mix of investment across online and offline advertising, given a target goal.

MMM is buffered from forces such as data deprecation because it uses aggregate time series data as an input — for example, daily display impressions for a given campaign in a given geography. Methods that rely on stitching together digital journeys have increasingly more holes.

Leading brands recognize MMM as an increasingly important measurement tool because of marketers’ need to assess the impact of emerging channels, shifts to the digital tracking landscape and changing consumer behaviors. The most effective marketers combine MMM with incrementality methods.

When applied to marketing, incrementality measurement quantifies the unique effect a campaign, experience, tactic or ad had on one or more business outcomes. Incrementality measurement can be done through a broad set of techniques, from relatively simple go-dark tests to more complex universal control groups to advanced marketing mix modeling. All three methods aspire to the same goal: isolating the changes caused by marketing.

For example, in advertising, it‘s common to compare the behavior of consumers exposed to ads compared to those not exposed to ads. The difference is an incrementality measurement for advertising. Analyses that do this accurately excel at making sure the two groups are as similar as possible. The only thing that should vary between them is the exposure to marketing.

How do the best brands do this? They generally have three fundamental approaches:

1. Test the validity of MMM’s predictions

Treat the marketing mix model as a credible suggestion, but not as a mandate that must be followed. How should marketers handle those suggestions, if not as a mandate? Conduct tests to validate models’ predictions.

One example of this is matched-market tests, where marketers make spending changes in one market but not another, then wait and compare the change in results between the two markets. If the test market increases sales or profits more than the control market, that result validates the MMM recommendations. In that case, the model has passed the test, and its influence on marketing decisions should increase.

2. Use simulations and optimization to generate better marketing plans

At its core, MMM is a predictive tool. Most marketing mix companies provide a simulation interface or tool. Beyond enabling the adjustment of marketing levers, scenarios can modify external factors included in the model. These planning tools try to account for marketing elasticities and the fact that each marketing activity is on a different part of a unique response curve.

3. Approach strategic marketing measurement as a journey

Measurement will continue to change as web browsers, walled gardens and regulators evolve their policies. As brands mature analytically, they often require different insights. These two forces should encourage changes to organizations’ marketing measurement over time.

CMOs should focus on improvements in the long run, such as testing new data sources or conducting analytics deep dives to better understand specific channels.

For large advertisers, data deprecation impedes digital-first marketing measurement efforts. While marketers and the measurement industry have returned to MMM, marketers who take MMM only at face value miss out on much of the value if they don’t follow these steps.

Matt Wakeman is a senior director analyst in Gartner’s marketing practice.