Measure the Effectiveness of Marketing Activities with Marketing Mix Models

2022-10-11 | Article | Insights

Summary

Marketing mix models measure the effectiveness of marketing activities without having to rely on 3rd party cookies. This Insights Article describes the data basis required, the statistical methodology of regression analysis, and the expected results of the model, which form the basis for future budget allocation decisions.

Challenge

Companies face the daily challenge of measuring the impact of their marketing activities. Many of them run in parallel, making it difficult to evaluate the effect of individual channels and campaigns on company-relevant KPIs. In addition to this methodological challenge, it is also difficult to evaluate the effect of marketing activities with regard to the measurement of 3rd party cookies - that is required for user journey attribution.

Approach

A time-tested method to measure the marketing impact e. g. on conversions is the marketing mix modeling (MMM). This approach not only allows marketing to be evaluated in the overall context and the main drivers to be identified. It is also very flexible in its setup, allowing all company and/or stakeholder specifics and requirements to be taken into account. Moreover, it uses a database that is independent of tracking regulations. The relevance of marketing mix modeling in the digital environment will increase significantly in the future, as this method does not require the use of individual customer journeys.

At its core, there are three components of a marketing mix modeling. (1) The underlying data that is given to the model, (2) the model itself, and (3) the model output from which the insights are generated.

The most critical part is the data input. Here as in all data science projects, the following applies: Garbage in garbage out. In the conception phase, it is important to determine the level and thus the correspondingly necessary marketing channels and campaigns for which an evaluation is required with a view to their value contribution to conversion. Depending on the target group, the requirements may differ. A budget decision maker, who is responsible for the entire marketing, is more interested in a rough grid of marketing channels, where both online and offline channels should be considered. A channel manager, on the other hand, would rather measure the impact of individual campaigns on conversions or brand awareness. Finally, in addition to media factors, daily seasonal information, such as days of the week, public holidays, vacations can be modeled. These different perspectives can each be served in the modeling, making MMM extremely flexible as a method.

However, the most important criterion is data quality. Daily spendings or impressions per defined input variable, as well as daily conversion and revenue data are collected for at least one year. All the data collected and given to the model should be correct and clean.

The modeling is based on the statistical method of regression analysis. The procedure is used to predict a numerical value, the target variable. This involves learning from the various historical variations of input and target variables and calculating the incremental impact of each input variable (e.g. marketing channel) on the target variable (e.g. conversion). The resulting final model is better the closer the actual value of the target is to the modeled value of the target. Simply put, reality is compared to the model. Minimising the difference between reality and model is used to find the best possible model.

Applied to marketing mix modeling, this means that this procedure is used to calculate the significant influence of each individual marketing channel and each other influencing factor on conversions. The aim is to ensure that the conversions estimated in the model correspond as closely as possible to the actual conversions.

The main model outcome is called decomposition: This is a break down of conversions to the proportional value contribution of the input variables, i.e. marketing channels and other influencing factors. This allows the most relevant marketing channels to be identified. Firstly, this is done on an aggregated level for the entire analysis period and secondly, then also on a daily basis, so that the effect of the budgets of particularly influential channels in the respective weeks or months can be interpreted particularly well.

For each channel considered, further insights are available: Saturation is used to determine the extent to which an increase in spending would lead to an increase in conversions. With this, it becomes clear for which channels the budget should be increased or minimized in the future. The carry-over effect describes how many days after the impression of an advertisement, there are still positive effects on conversions. On the one hand, this gives advice for the subsequent media strategy and campaign planning. On the other hand, the carry-over effect can be compared across channels. This makes it possible to sort them according to their aftereffects in a funnel visualisation.

One of the most important outcomes for data-driven marketing decisions, esp. optimization of media budget allocation, is the knowledge of the Return on Ad Spend (ROAS) per channel. This can be derived from the model as well. All relevant KPIs per channel are summarised in an evaluation table so that the value contribution to revenue and the ROAS for each channel can be assessed at a glance. This table, together with the information regarding saturation and carry-over effects, forms the basis for the optimization of the future media mix.

Result

Ideally, the results of the MMM are used to take actions with regard to optimize the budget allocation and campaign planning. This leads to new data and interrelationships that can be analysed in a subsequent MMM. A model update allows the success of the implemented strategies to be verified. It is therefore advisable to automate the MMM setup and to calculate updates on a regular basis - monthly or quarterly - in order to get into a flow of data-driven budget decisions.

Do you need more Info?

Contact us