All models are wrong, but some are useful
George Box
Identify the right problems
- Use data to idenfity the key levers in the problem
- Get your data architecture in place so you know what, why and how to track?
Understanding the problem is 80%
- The approach is simple if the problem is well understood.
- Keep the modules/approaches simple, scalable and extendable.
Use the right frame of reference
- A well thought through solution uses the right frame of reference
- This is true across Physics, Electrical Engineering or Statistical Learning