Have you ever had that experience where you try to explain a relatively complex concept to someone and you can see their eyes glazing over? It could be helping a teenager with their maths homework, or explaining the results of some detailed and advanced analysis to a work colleague. If you could read their minds you would probably hear "Wow - this sounds complex...I think I'll just do what I usually do".
This is what often happens when a statistical model is provided to an end-user. The model could be near perfect and tick all the boxes from a statistical theory point of view, but to the end-user, it looks like the type of complex formula that Sheldon from The Big Bang Theory has created. The result is that the person you most wanted to use the model won't. They see it as rigid, confusing and more importantly, it is someone else's. The user doesn't understand it, and it is hard to take ownership of something they don't understand.
Ownership is often one of the biggest impediments to the use of an advanced statistical model.
" I didn't really understand that - so I did my own in Excel" is a common phrase.
So how do you ensure the model is accepted?
The trick is taking the end-user on the journey in developing the model.
If the process is logical and explainable they can get a far greater understanding of how the model is derived. The user will be able to see why a specific driver has a more significant effect, or why a baseline behaves differently to when there are events.
If the process is quick the end-user can quickly see how the model is affected by the data. They can see the effect of a missing price, or what effect an out of stock has had on the model performance.
If the process is intuitive the end-user will be able to understand the insights that were uncovered. They may be able to confirm their gut-feel, or uncover learnings that may improve the way the business operates.
If the process is agile then the end-user will be confident that the results are current and always useful.
Want to learn more about how this can be achieved - contact Cauself.
Comments