One of the biggest challenges on MMM is getting up to speed quickly with an unknown dataset. Based on that, during the last month we developed a MMM Framework using state of the art libraries that allow us to do a quick exploratory analysis of multiple variables at once and determine which paths are worth being followed with a more detailed approach. 

Fig 1. (above) Model measuring automatically correlation between 10 variables.

The model is also set to output standardised analyses to detect and correct any inconsistencies on the raw data and help exclude outlier data points which can skew the result.

Fig 2. Standard analyses of the dataset to generate a prediction model

Along with this, we’re working on structuring and updating common datasets such as global economic variables (GDP growth, the Consumer Confidence Index) and other data sources (reading GRIB weather data from Swedish and Finnish stations that cover all Scandinavia and Baltics).

Fig 3. GRIB Output example of temperature variation from Finnish station

Having these analysis scripted makes them replicable (if the same customer needs an update) and for every new customer the codebase will expand to incorporate more variables, reducing the overall project time and increasing efficiency. Currently we’re applying this model on an ongoing project with a casino client (using Swedish-based data) and several other prospects for demoing purposes.

If you have more questions about the framework, do contact us by clicking on the button below.