A century ago, John Wanamaker (1838–1922) famously coined the phrase “Half the money I spend on advertising is wasted; the trouble is I don’t know which half”. A multitude of primary and secondary media research techniques have emerged ever since to guide marketing strategy. Yet per Hubspot, about 40% of marketers still identified, in 2017, proving the ROI of their marketing activities is their top marketing challenge.
Marketers know they are missing out on key information if they’re not measuring marketing effectiveness which is at the core of addressing marketing spend questions:
- How much media is enough?
- Which medium is most effective? Online or offline?
- Is it better to aim for reach or frequency?
- When to use flighting or continuity in media scheduling? When are ads worn out?
There is no doubt that ROI-centric marketing management and marketing accountability are critical for organizations today. The real question is how.
Marketing mix modeling is a statistical technique of linear or non-linear regression that uses historical information, such as point-of-sale data and companies’ internal data(i.e. campaign calendar), to quantify the causal relationship between sales and various marketing activities. This is achieved by setting up a mathematical model with the sales volume as the dependent variable and the marketing activities as the independent variables. Needless to say, data integrity and granularity will be crucial for any robust MMM.
Once the variables are created, the model works to calculate a base sales volume that reflects historical sales trend and seasonality but is insusceptible to changes in any of marketing activities. And with the rest of the volume, termed as incremental volume, the model slices and dices it in a way that maximizes the correlation between a volume portion and changes in one of the marketing activities. However, this decomposition process is subject to analysts’ intervention as the calculated correlation may not make perfect business sense, or could be attributable to some other unknown element(i.e. competitors’ campaign) that analysts are yet to discover. The fine balance between automated modeling tools crunching large data sets versus the artisan econometrician tweaking parameters and coefficients is an ongoing debate in MMM domain, with different agencies and consultants taking a position at certain points in this spectrum.
What’s certain is multiple iterations need to be carried out before the model is any good for future simulation. And further validations are necessary, either by using a validation data set or by the consistency of the business results. Once a quantitative relationship can be established between volume and marketing activities, it’s like solving ax+by=z, once we figure out the coefficients a and b, we can plug in any value of x and y to see what value of z it will yield. We can then deploy this learning to fine-tune marketing tactics and strategies, optimize the marketing spend levels and also to predict sales while simulating various investment scenarios. For big budget activities like television advertising, attribution could be conducted down to ad copy level, i.e. marginal incremental volume that can be obtained by increasing the respective marketing element(i.e. ad copy) by one unit.
Half the money I spend on advertising is wasted; the trouble is I don’t know which half
Marketing Mix Modeling closes the loop in terms of the customer journey and outlines the path to the optimal distribution of spends and improved return on marketing investment. However, MMM is not without its limitations. Tying back to the importance of data integrity and granularity, MMM models have a clear bias in favor of time-specific media (such as TV commercials) versus less time-specific media (such as ads appearing in monthly magazines). Biases can also occur when comparing broad-based media versus regionally or demographically targeted media. Lastly, the focus on short-term sales can significantly under-value the importance of longer-term equity building activities.
In closing, before you dive into MMM, you need to ask what your top marketing priority is. Is it proving the ROI of marketing activities and securing budget? Is your data comprehensive and granular enough to run a weekly time series model? Are your marketing activities primarily driving sales or building brand equity? Once you understand that, you will have a crystal clear idea about whether and why you would need MMM.
Read my other thought pieces here.
