“Data-driven thinkingis written by members of the media community and contains fresh ideas about the digital revolution in media.
Today’s column is written by Alice K. Sylvester, Partner at Sequential partners.
Marketing today is too complex to be decoded with the naked eye. Machine learning and statistical models have evolved to provide an extremely sophisticated understanding of the contribution of each marketing investment and the interactions between them.
These models are actually black boxes. They cannot be easily judged or questioned, leaving brand managers and CMOs to blindly follow the results or ignore them altogether. Models can leave marketers vulnerable to faulty decisions and underutilization of data and analytics, which prevents them from taking full advantage of the insights the model can provide.
Fortunately, marketers don’t need a PhD in econometrics to get the most out of marketing analytics and modeling. There are a few simple marketing mix and attribution modeling questions that all marketers can — and should — consider.
Is the model complete and solid?
Does the model include all marketing investments and drivers of sales and results? Otherwise, the model cannot be trusted to accurately indicate the contribution of each investment. It is important that the model takes into account spending on digital and TV walled gardens, linear or analog media, and many non-marketing market factors (eg economy, weather) that influence sales. Attribution modeling typically doesn’t include these factors, but it’s still important to recognize exactly which channels are considered in the model. If something is omitted, the model will overestimate the contribution of the media included in the model.
The strength of the model also matters. In general, marketing mix models should fit outcome data with an R2 in the 90% range, indicating causality between the model and the outcome data, and an MAPE (Mean Mean Percent Error) of less than 5%, relative to an unselected sample.
Marketers also need to determine whether the results of specific models match the results of past analysis or market testing. If not, marketers should pressure the modeler to find possible explanations for the inaccuracies. The model’s prediction is based on what has happened in the past. How is the situation different today?
Does the model take into account external factors, such as brand and sales effects or initial inputs?
Marketing mix models must capture the complexity of the market. It is essential to take into account interactions such as the effect of advertising on price elasticity, the halo effects of advertising on other brands and the less immediate carryover effect of advertising (adstock) .
The deferral also exists in the attribution – this is the attribution window. Without accounting for these effects, advertising’s contribution to sales or other results is distorted and blurred.
Data inputs can also impact results. Often, the data sources used for the model are not the same ones used to guide your business on a day-to-day basis. To ensure your data strategy is holistic, consider both internal and external sources. Are the results data KPIs (e.g. category penetration, brand shares, sales, website visits, traffic trends) correct? What about marketing investment data? It is imperative to ensure that the data entries correctly reflect the brand and category.
How clear and actionable are the results?
Everyone wants near real-time attribution results. But can your operation realistically process daily results data? Or even weekly or monthly? Does your marketing mix modeling match your business cycle? It’s important that modeling information arrives when you need it and that you can act on it.
It is also important that the modeler can communicate in simple terms and is experienced enough to interpret the results based on your specific business opportunities and risks. Are their explanations of how your marketing works relatively close to your assumptions? Do they make sense? Are they triangular with other business/test results you see? This is the reality check phase. Make sure the results make sense.
Once all the information is in place, it’s time to act against the model information. At this stage, it is crucial to have alignment between all stakeholders within your organization. Otherwise, modeling is just an interesting, time-consuming and costly exercise.
Modeling is an important feature of sophisticated marketing today. Marketers need to ask the right questions to understand inputs, outputs, and implications. And there may be even more questions to ask regarding your business growth and specific revenue needs.
Follow Ad Exchange (@adexchanger) on Twitter.