From the course: Predictive Analytics Essential Training: Data Mining

Addressing organizational resistance

From the course: Predictive Analytics Essential Training: Data Mining

Start my 1-month free trial Buy for my team

Addressing organizational resistance

“

- [Instructor] As data scientists, we're just trying to help everybody out. Why would there ever be resistance? Well, in short, we're in the business of organizational change, you can't inspire change without communication and trust. So the next element is that, unfortunately, you can count on at least some organizational resistance. I remember one engagement where an employee explained to me that the reason he was so distrusting of the data science team is that he associated them with finger-pointing. Data science for him was all about blame. Now it wasn't true, but communication had clearly broken down. The trick to combating this is to remember that you're building the model to be used, but you won't really be the one that's using it. Go find those that will be affected. They will be the employees out in the field, performing repairs or investigating claims, and so on. As a result of the model, their typical day might change. Even their workload might change. A great way to respect their role is to do a test run. A colleague of mine is somewhat famous in a particular software user community for producing a very successful auto part assortment model. He was trying to figure out what car parts should be stocked at each individual retail location. The model looked very promising, but most of the local retail management rebelled and didn't want to be part of the trial rollout. Even though the rollout was going to be a hundred days, and just a small percentage of the stores, senior management had to get involved all the way up to the C-suite. It turns out it was a happy ending. As it was so obvious that the model was excellent that within 30 to 60 days, all the retail stores wanted in, not out. That model went on to have a nine-figure ROI. The details may change, but this will happen to you. After all, if a model with those kinds of numbers faces resistance, then any model might count on it. So plan for a trial rollout, involve the whole team and involve them early. Anticipate concerns, make sure to examine things from their point of view. If you do all that, you can look forward to having a successful deployment.

Contents