How do you go from a Facebook ad stats to a campaign update decision?
Once your engineering team did all the heavy lifting of connecting with the channels Ads APIs, storing the data in a normalized format, and make it available to your analysis, the job is still not done.
There are two phases left to complete the cycle: insights and predictions.
How to convert clean, normalized data into useful insights, that can guide the analyst to make the right decisions, is where most of the ad tech companies are struggling today.
As with any other ad tech project, you can build your solution, or you can use -get inspired by- existing products. For Insights (also known as business intelligence) you have products like Knowi or Tableau.
Although they are moving towards machine learning, you will be better off building your tool here. You can use libraries like Python scikit-learn (here are a few code examples for Facebook Ad stats), or use the AWS Machine Learning service.