Maintaining stakeholder buy-in
By Eyram Adjaku
In my previous blog, I wrote about using a sales approach to get stakeholder buy-in for your data science project. The next important step is to set clear goals for your project because any project without explicit objectives is likely to be derailed. Data science projects are inherently complex and this makes setting clear and measurable goals critical for all involved.
Set realistic expectations
You need to define clearly what the business goals are and understand how progress will be measured against these objectives. It helps to define what a “good enough” solution looks like and agree on this metric with the business. While most stakeholders will judge what this is with their gut, be careful of chasing results that are unachievable.
With your goals and metrics in hand, the next step is to experiment within the proposed optimization parameters. One way of doing this is with A/B testing. Such tests allow you to compare the new optimization technique against the existing process. This helps answer a very important question, namely, “To what extent does the new process affect my bottom line?” This is where the proverbial rubber hits the road because now, the model accuracy often touted by data scientists, doesn’t really matter anymore. The business is simply interested in knowing how the model performs in the real world.
A/B tests demonstrate to the business how it would have performed with and without a particular intervention. Interventions can range from the type of messaging sent to a customer to finding the optimal position of a “buy now” button on an ecommerce website. If constructed well, A/B tests are easily understood, and especially when they speak to the bottom line. And improving a business’ bottom line can go a long way to building trust in your solution.