2022-08-26 | Article | Insights
As is well known, many roads lead to Rome. In the case of evaluating marketing efficiency, there are two approaches that are available: Attribution and Marketing Mix Modeling. With each of these approaches, it is possible to identify the relevant marketing channels that have contributed to past conversions and to calculate their individual value contribution in terms of conversions and/or revenue. These insights form the valuable basis for future budget decisions and are therefore essential.
In recent years, Attribution has been the preferred approach in the fields of digital analytics. For example, the reports integrated in Google Analytics are based on this concept. However, with the problem of collecting 3rd party cookies, it is becoming increasingly difficult to build the necessary database to a sufficient extent. For this reason, we see a strongly increasing relevance of Marketing Mix Modeling in the future. With this approach, it is possible to calculate marketing efficiency without having to use the information of individual click paths.
The big commonality of both approaches is the goal they pursue. Historical data is analyzed to learn which marketing channels and campaigns have contributed particularly well to the target KPI. This knowledge forms the basis for future budget allocation decisions.
The differences of Attribution and Marketing Mix Modeling are manifold, but they can mainly be broken down to three different levels.
The first level is the underlying data basis: Marketing Mix Modeling uses the daily spendings or impressions per marketing channel as input variables. Ideally, at least one year's worth of data is included in the modeling. In contrast, Attribution leverages aggregate customer journeys, i.e., the sequence of marketing touchpoints that lead to a conversion. An important advantage of Marketing Mix Modeling lies in the variables that can be taken into account in addition to the on- and offline marketing channels. For example, seasonal factors are regularly included in those models.
The respective data basis also results in a difference with regard to the method of analysis. In Marketing Mix Modeling, regression analysis is used to identify the significantly influential marketing channels and other factors and to calculate their proportional influence on conversions. In contrast, Attribution can be both rule-based and data-driven. The most popular rule-based attribution model is last non-direct click, which assigns the entire conversion value to the last marketing channel in the user journey. If this last touchpoint is direct, the previous non-direct touchpoint is used instead. Shapley Value dominates among the data-driven attribution algorithms in digital marketing. This approach originates from game theory and distributes the conversion value to the respective touchpoints depending on the actual involvement of the marketing channels in the user journey. As a rule of thumb, the more the model is supposed to represent reality, the more complex it is. On this scale, rule-based attribution models are at the beginning and Marketing Mix Modeling at the end.
A third difference is the actuality of the results. Since Marketing Mix Modeling is calculated based on at least a year's worth of data, daily model updates would not produce much change in the results. Rather, monthly model updates make sense. This means Marketing Mix Modeling is more suitable for medium and long-term budget decisions. A budget scenario calculator based on the model can be used to compare the outcome of different budget allocation options. Attribution is more flexible here: Models are calculated on data from shorter time periods. For example, data from two to four months is enough. Based on this, new attribution results can be calculated automatically on a daily basis, which leads to a higher actionability. However, it is not possible to predict future scenarios regarding budget efficiency.
As with many things in life, there is no one solution when evaluating marketing efficiency. Marketing Mix Modeling and Attribution each have their advantages and disadvantages. Depending on individual requirements, intensity and diversity of marketing activities as well as the available database, one or the other approach is more advisable.
In a perfect environment, both approaches are used in parallel, as they certainly complement each other ideally. However, it is clear that the available database for Attribution is becoming increasingly difficult due to the 3rd party issue. This means that the marketing touchpoints of the individual customer journeys will no longer be fully available in the future. Hence Marketing Mix Modeling will become much more important as an approach to evaluating marketing efficiency.