Model
A glimpse under the hood of Sellforte's powerful Marketing Mix Modelling platform methodology
Bayesian inference
Sellforte’s approach on Marketing Mix Modelling (MMM) is Bayesian hierarchical modelling methodology which has been the gold standard for MMM since 2017 when Google’s research team decided to sum up the problems with current methodologies in MMM.
Bayesian hierarchical models enables usage of priors in modelling which helps to manage the noise of the data, but also the model to learn from data when there are statistically significant differences e.g. TV channel Y has higher ROI than TV channel X.
Automatic holidays, seasonality and trends
As business' have different seasons when marketing is more effective than other times, it's important that modeling will take into account these variables. Modeling includes
- Automatic seasonality and trends based on your historical data
- Trend component that recognizes your business trend, i.e. high growth or stable business
- Taking into account special days and vacations per market
Refreshing the model data
As new data arrives every day, week, or month, typically models are re-trained every time. Updating parameter estimates based on new evidence is at the very core of Bayesian statistics. Not all marketing teams are equipped to act on the results every day or every week so sometimes a slower updating cycle makes more sense. If offline data is involved, it typically cannot be provided on demand.
- Updating data as fast as data can be delivered
- If offline media is used, that is updated typically on a monthly basis with a separate .csv file via Data Platform
Response curves
Scenario planing tool, Media Optimizer utilizes response curves learned in the MMM but typically on a less granular level to ensure robust results.
Custom optimization algorithms are implemented through Python and AWS Lambda microservice.
Response curves answer the question how long media will still deliver 1 $ back from invested media dollars.
