Under the Hood Part 5: What is really automated in Proof?
What is really automated when we talk about automated marketing mix modeling? Traditional marketing mix modeling catches delayed effects and has a high accuracy. The downsides with traditional MMM are the high costs, lack of talent and it's time-consuming. Automated MMM takes weeks instead of months, it's a fraction of the cost and it is a smarter use of talent. The automation in MMM is the significance test of each variable which allows you to easily introduce new variables into the model and create new ones.
The reason why marketing mix modeling has been so popular and still is among the biggest brands in the world is that it catches the delayed effects and it produces very high accuracy compared to any other kind of marketing and PR modeling. The drawbacks with a traditional modeling is that you have a very high ticket – it’s very expensive. It takes a lot of time, normally weeks to months to produce and to define the models and then to execute the models. The results for the decision maker is taking quite a long time. The third aspect is that there is a lack of talent, two-thirds of the people doing marketing mix modeling are PhDs or higher. The amazing thing with automated marketing mix modeling is that it takes minutes instead of weeks or months. It’s a fraction of the cost and the smart PhDs can focus on deeper modeling and instead introduce people that are not PhDs to do the modeling. Because it’s easily done within the software.
What is really automated in Proof?
People tend to ask us so what is really automated in Proof? Normally the execution for a lot of companies is actually automated so that’s nothing new. In an automated marketing mix model you can decide if you run the model in an automated mode or not. The significance test on each variable in traditional modeling is consuming a lot of time so many PhDs are shoveling coal in the mine doing a very time-consuming significance testing on a long list of variables. Within an automated marketing mix modeling the significance tests are actually automated which is the definition phase of the model, not just the execution phase. Also thanks to automation you can introduce variables really fast. If you have done the modeling, it takes a few minutes, then one week later a person is saying ” I have this new data set”. You introduced three models automatically and you have to restart the whole process go nine weeks back in time and restart it. It immediately takes that into account.