18.9 Garbage In, Garbage Out

If you were a chef, you would do your best to ensure you had the freshest possible produce to work with. After all, a delicious meal is only possible if the person cooking it has both skill and quality ingredients.

The same principle applies to analytics. Analytics is not magic – it is the application of statistics to business problems.  Sometimes, reliable, solid solutions can only be obtained when the analyst has access to quality data and sufficient time to ‘slice and dice’ the numbers.

That said, it is possible that you, the analyst, will not get access to the full set of data.  This can happen for many reasons, including time and cost limitations, or access permissions.  If this happens, it is important to note the limitations and assumptions in the final report and presentation to stakeholders. 

You may consider creating your own data to demonstrate an idea, since the prototype helps stakeholders visualize an outcome.  If and when you do so, keep in mind that your simulation needs to be realistic. For instance, if you are creating a set of US household income data, the data distribution would need to be skewed right.

Should you contemplate feature engineering as a way to reduce dimensionality in the dataset, we recommend explaining and discussing the approach with key stakeholders before proceeding.  If they are on board with your recipe from the outset, this “buy-in” may make things easier for you down the road.