Lenskold Article Series

by Jim Lenskold

Three Critical Questions to Get the Most from Marketing Mix Modeling

Marketers interested in measuring effectiveness and ROI need to understand if marketing mix modeling can work for them and how to get the most from it. Modeling measurements are often considered a bit complex and mysterious by marketers – many of whom did their best to survive any required statistics courses. Marketing mix modeling (MMM) has long been a reliable measurement for consumer package goods (CPG) companies and is now proving to be a great planning and measurement tool that increases profits in many diverse B2B and B2C industries. The challenge is that marketers are missing opportunities to fully benefit from marketing mix models as part of a comprehensive measurement and management process.


Marketing managers need to understand if marketing mix modeling will work in their environments and what to expect from the outputs. For companies already using marketing mix modeling, they need to assess its value in driving marketing decisions and determine if they are truly tapping into its full potential.

Will MMM Work for You?

Modeling of marketing effectiveness is dependent on good historical data capturing, at a minimum, your marketing activities and sales performance. The quality of the data and the level of detail drive the accuracy of the modeling. There is a significant difference in having data on a weekly basis vs. monthly. We won’t get into the technical requirements for modeling but will establish that companies without quality data are not good candidates for marketing mix modeling.

There are additional considerations once data availability is confirmed.

  • Do you have the resources – either internal expertise or budget – to complete the modeling? Using more sophisticated methodologies does have a cost, especially if there is work required to clean up and consolidate the data prior to developing the models.
  • Do you have large volumes of sales each month? Modeling accuracy is generally better with larger volumes that can establish statistical validity. As volumes decrease and marketing mix modeling becomes ineffective, there are other modeling and analytic techniques that can be applied.
  • Do you expect that marketing is driving at least a moderate portion of the total sales activity? For marketing organizations that play a minor role relative to sales-driven organizations or retail presence, it is possible that the contribution of marketing may fall below detectable levels.
  • Will your data reflect marketing impact that aligns with your marketing objectives? For example, if marketing contributes to generating new customers expected to have long-term purchasing behaviors, modeling the impact on short-term sales volume will not reflect the complete contribution.
  • Can you include data points for external factors that may also influence sales activity, such as competitor marketing, economic trends, environmental conditions (weather), or changes in your own operations and customer contacts? Modeling that includes more diverse and relevant data points will provide greater insight while a lack of analysis on external influences could leave gaps that distort the results.

What you should expect from a high quality marketing mix model in terms of both benefits and gaps:

  • The analysis will identify correlations that indicate the incremental contribution to sales/revenues/profits generated from each marketing channel. More detailed models will show the impact at a promotion level with some insight into campaign strategy.
  • Smaller-scale marketing programs are often below detectable levels; however, this should be acceptable since you typically want to put your strongest measurement effort against the larger initiatives. Other forms of measurement are more appropriate for smaller, strategic initiatives.
  • One of the most appealing benefits from the perspective of marketers is that this measurement comes after the fact and does not require modifications to the marketing that is typical with measurements such as market testing.
  • One of the greatest shortcomings is that modeling a full year of historical data will typically deliver results after three months of analysis, providing results over a year after the marketing initiative concludes. This limits the speed and ability upon which to act on the insight.
  • Identifying the influence of external factors on your sales activity provides valuable insight on historical performance, but also lets you apply adjustments to your real-time performance tracking that follows as you see these external factors change.
  • It is important to understand that factors such as seasonality effect sales even without marketing programs.
  • Marketing mix models can be used to develop simulation models and optimization models that apply the calculations from the historical analysis in an interactive tool to support planning of alternative marketing programs.
  • Marketing mix modeling plays a specific role in measurements – it establishes correlations between marketing initiatives and results (typically but not always incremental sales activity). What it does not provide is any diagnostic information as to what parts of the overall strategy are driving the marketing program’s success or perhaps causing it to fall short.
  • Once the model is completed, there will be a large portion of the sales that are considered “base sales” in addition to the sales associated with your marketing programs. This is sometimes referred to as “brand impact” and mistakenly associated as sales driven by the value of the brand. What this base represents is the uncorrelated portion of the data, which includes a combination of impact not appropriately correlated to current marketing activities, sales activity from repeat customers not captured through modeling, and your base sales from your overall brand and customer relationships.

Is MMM Driving the Right Decisions?

For far too many marketing organizations the first question is, sadly enough, are you driving any decisions with your modeling outputs. Too often, marketing mix model results are used for reporting and justifying marketing investments already made with little or no change in decisions going forward.

The modeled output, in the format of reports, simulation tools, or optimization tools, should support your decisions to re-allocate budget and achieve greater returns. It is important to recognize that the model delivers insight and does not generate “the answer” in terms of how to invest your marketing.

You need to consider the following in your decision process:

  • The analysis is based on your historical performance, which reflects the complete strategy of each campaign (i.e., targeting, offers, messages, integration), even though the results may lead to comparisons by marketing channel. You will need to apply additional measurements or your own knowledge to understand what is making one portion of your marketing investments more effective than others.
  • Along the same line of thinking, the model output assumes you can replicate your success, so the more you understand what drives your success and how external conditions influence your impact, the better you can prioritize and implement the right strategies and tactical mix.
  • The models will guide you to replace (or shift budgets between) marketing channels and offer little insight into fixing marketing effectiveness. You ultimately need a mix of re-allocating budgets to marketing initiatives that are generating better returns and actions to improve those marketing initiatives that may fall short of your expected returns.
  • Higher quality marketing mix models will also capture the combined impact of multiple channels working synergistically and offering better returns than the marketing channels working independently. This is important since we expect the most effective marketing programs to leverage the strengths of multiple marketing channels and not just one top-performing channel.
  • Modeling is not limited to correlating marketing activities with just sales volume or revenue (the most common dependent variables). Where detailed data is available, model the marketing impact on customer behaviors that drive revenue (such as first time buyers, growth in customer value, or reduced churn) or to customer actions (such as inbound inquiries, new leads, or website visits). This insight becomes more actionable and better supports strategic planning. It also helps build credibility with non-marketing executives who can understand the influence on customer behaviors but find it difficult to accept the analytic “leap” to influencing revenue.

Are You Getting the Most Insight from MMM?

Once you understand how to use marketing mix modeling more effectively to better support your decisions, you can then go further to draw even greater insight. These opportunities generally come as you incorporate measurement objectives into your marketing plan, and as you integrate multiple measurement methodologies together to accelerate and deeper your insight into improving marketing effectiveness.

Here are a few suggestions to get more from your MMM:

  • The underlying statistical methods used to create the correlations are dependent on having a broad range of values – so the change in values as marketing activity increases and decreases can be correlated to the increases and decreases in sales activity (or other outcomes) that follow. If your marketing activity is fairly consistent throughout the year, the correlation will be difficult to detect. You can enrich the data and the insight from modeling with more extreme differences in your activities over the course of the year. This is especially helpful if there is some debate over the appropriate spending levels. You might, for example, trim spending in one quarter back by 25% – 50% and increase the levels by 25% – 50% in the following quarter to provide better data that will help prove or disprove the differing opinions.
  • Building on this approach, instead of changing spending levels nationally or globally, which may be a risk to sales performance, run these changes in select markets. You get the immediate insight into the market test results, plus your data is much richer for your year-end modeling.
  • Once marketing mix modeling is run, the conclusions and subsequent decisions should be validated by adjusting budgets and strategies on a limited basis, and measuring performance through market testing or data tracking. As the expected results are confirmed, more significant changes in step with the model outputs can follow. This not only confirms the model accuracy, but also reinforces its value with other key stakeholders as you demonstrate the ability to replicate the conclusions.
  • Depending upon your business model, analysis that just tracks incremental short-term volume is likely to miss out on the long-term customer value generated from that specific marketing initiative. A combination of modeling to identify the incremental customer behaviors (such as new customers or certain product sales) and an ROI analysis with a more complete view of customer value, will drive the right decisions.
  • Marketing mix modeling should be designed as part of a comprehensive measurement and analysis plan that extends one year or longer. It is important to complement the insight from modeling with more diagnostic measures that come from market testing, quantitative research, qualitative research, and internal data tracking.

Measure Smart to Manage Profitability

The best way to maximize and manage marketing ROI is to prioritize your measurement objectives based on supporting the most critical strategic and tactical decisions and having the greatest impact on profit performance. Companies that have good data and are not using modeling need to consider marketing mix modeling or other forms of regression or time-series analyses. Create a measurement plan that leverages the strengths of different measurement methodologies and provides the marketing organization with reliable intelligence they can act upon. There are always more information needs than can be actually measured so be smart in where limited resources are prioritized.