Should I spend more on Digital?

Am I spending enough on TV?

Is my Distribution and Pricing working?

How far can I hike my price?

Am I getting sufficient ROI for my marketing spends?

These are oft-asked questions surrounding business, category & brand custodians, globally. Market Mix Modelling (MMM) helps address such questions, by quantifying the impact of several marketing inputs on the sales volume and/or market share. Furthermore, it measures the contribution of each marketing input to sales; and how expenditure on these marketing drivers can be optimized.

Using Multivariate Linear Regression, MMM extracts actionable insights about category & brand utilization of marketing inputs. MMM establishes an equation between the dependent variable (sales volume/ market share) and independent variables (distribution, price, ATL & BTL spends). This relationship can be linear or non-linear. Variables like TV Spends/GRP, Digital spends, print spends etc are treated as non-linear variables since every unit increase in these variables does not lead to a unit increase in sales. To consider these kinds of variables as modelling inputs, they are transformed into Adstock.

Adstock is a model of how response to advertising builds and decays in consumer markets. There are two dimensions to Adstock:

1. Carryover effect or decaying effect: Advertising has prolonged or lagged effect on consumer purchase behaviour. This is called carryover effect.
2. Diminishing returns: A linear increase in advertising exposure does not cause a similar (linear) effect on sales. Typically, incremental amounts of advertising cause progressively lesser effect on sales. This is referred to as diminishing returns.
3. Marketing mix models may divide sales (dependent variable) into two components –
4. Base Sales: The natural demand for the product (‘pull’), driven by economic factors like seasonality, long term trends, brand awareness, brand loyalty etc. Base sales are usually fixed, unless economic or environmental factors change.
5. Incremental sales: The ‘push’ component of sales i.e. sales driven by marketing and promotional activities. It can be further attributed to individual marketing components like television, radio, print and digital advertising, and promotions like display promotions, temporary price reductions etc.

Typical MMM studies provide insights about contribution, ROI, and effectiveness of different marketing activities to help businesses optimize spends. Once it is established which media are working better than the others, budget reallocation from low-ROI media to high-ROI media is possible; thus maximizing sales.

ThinkBumblebee value-adds to the classic MMM, by running such models through a multi-layer lens. Such an approach quantifies impact of both, direct and indirect impacts of each variable, on a brand’s market presence. When evaluating your current MMM, be sure to ask:

• Does it identify and track high-impact communication?
• Does it provide accurate and benchmarked measures of ROI?
• Does it enable your decision-making, with recommendations to maximize value uplift?
• Does it provide concrete inputs on SKU-related tradeoffs?

Sources

1. Towards Data Science
2. LeoWang on Medium

‍