“Data-driven thinking”Written by members of the media community and has new ideas about the digital revolution in the media.
Today’s column is written by Alice K. Sylvester, Partner A. Sequence partner.
Marketing is very complicated to decode with the naked eye today. Machine learning and statistical models have evolved to provide a more sophisticated understanding of the contribution and interaction between each marketing investment.
Those models are actually black boxes. They cannot be easily judged or questioned, leaving brand managers and CMOs to either blindly follow the results or ignore them altogether. Models can leave marketers at risk for misleading decisions and the redundancy of data and analysis, which prevents them from realizing the full value of the insights that the model can provide.
Fortunately, marketers don’t need a PhD in economics to get more out of marketing analysis and modeling. There are some simple questions about marketing mix and attribution modeling that all marketers can consider – and should – consider.
Is the model complete and powerful?
Does the model include all marketing investment and sales and results drivers? Otherwise, the model cannot be trusted to accurately indicate the contribution of each investment. Digital and television wall gardening, linear or analog media, and many non-marketing marketplace factors that affect sales (e.g., economy, weather) are important for cost calculation models. Attribution modeling does not tend to cover these issues, but it is still important to identify exactly which channels are responsible for the model. If anything is omitted, the model will exaggerate the contribution of the media included in the model.
Model strength is also important. In general, marketing mix models should fit the results data with an R.2 In the 90% range, indicates causality between model and result data and a MAPE (average percentage error) of less than 5% compared to a holdout sample.
Marketers should also consider whether the results of a particular model match the results of past analyzes or in-market tests. If not, marketers need to press Modeler to find possible explanations for the error. The model forecast is built on what has happened in the past. Is the situation different now?
Is the model responsible for external factors such as brand and sales impact or initial inputs?
The marketing mix model must capture the complexity of the marketplace. It is important to consider the effects of advertising on price elasticity, the halo effect of advertising on other brands, and the low-instantaneous carrying effect (advertising) of advertising.
Also exists in the carryover attribute – this is the attribution window. Without accounting for these effects, the contribution of advertising to sales or other results is misrepresented and modified.
Data inputs can also affect results. Often, the data sources used for the model are not used to manage your day-to-day business. To ensure your data strategy is holistic, consider internal and external sources. Does the result data KPIs (e.g., category penetration, brand share, sales, website visit, traffic trends) look accurate? What about marketing investment information? Ensuring brand and department accurately reflected data input is essential.
How clear and actionable are the results?
Everyone wants closer to real-time attribution results. But can your operations actually deal with daily outcome information? Or even weekly or monthly? Does modeling your marketing mix fit your business cycle? It is important that you reach out to modeling insights when you need them and work on them
It is also important that the modeler can communicate in the position of the general public and is experienced enough to explain the consequences for your specific business opportunities and risks. What are their explanations for how your marketing is working comparatively close to your expectations? Are they understandable? Do they triangle with any other business / test results you see? This is the reality-check episode. Make sure the results are meaningful.
Once all the insights are in place, it’s time to work against the model insights. At this point, aligning with all the stakeholders in your organization is crucial. Otherwise, modeling is just an interesting, time consuming and expensive exercise.
Modeling is an important feature of sophisticated marketing today. Marketers must ask the right questions to understand input, output and impact. And there may be more questions to ask related to your specific business growth and revenue needs.
Follow AdExchanger (@ Addexchanger) On Twitter.