Many would say that the most critical phase of any analytics endeavor is what happens at the very beginning. The way a project is defined and scoped has to be right or else everything that follows is irrelevant. Unfortunately, it can be difficult to find the time to execute a proper scoping process. Not only are analytics and data science professionals very busy, but their business sponsors are as well. There is much temptation from both sides to start “doing something” quickly. Instead of falling into this trap, focus on the simplicity of playing the game of 20 questions.
20 Questions Can Find Almost Any Answer
One of my kids received a gift a few years back that was a small electronics game. It purported to be able to identify anything you might think of in 20 questions or less. To play, you simply think of any person, place, activity, or thing. Then, the device’s screen starts to ask simple questions to which you respond either “yes”, “no”, or “I don’t know”.
The device always starts with very general questions like “Is it a person?” and then gets more specific as the questions progress. The amazing thing is that despite my initial skepticism that the cheap device could guess what I was thinking of in just 20 questions, it did so more often than not. I found I needed to think of something very obscure in order for the device to lose. Given their narrower level of life experience, the game was even more successful with my kids.
Make Your Scoping Sessions A Game Of 20 Questions
To make a scoping effort seem less daunting and time consuming, approach it and position it as a simple process of asking 20 questions. The questions are meant to ensure that the project team is in alignment with the project sponsors on exactly what the problem is, what the sponsor is hoping to learn, and how the sponsor hopes to act.
Approaching scoping this way makes the process feel less open ended and less intimidating while also adding a bit of gamification. Just as with the children’s game, much more can be gleaned from a few simple questions than most people would predict.
Which 20 Questions?
Now we get to the challenge of the matter. While it is true that 20 questions will be enough to get to the heart of a matter in most cases, it isn’t possible to specify a magical list of 20 specific questions that will work each time. It will take a different mix of 20 questions for every case, even if there is a common structure and logic to those questions.
In the children’s game, if I say “yes” to the question “Is it a person?”, then the game will start to ask additional questions relevant to that answer. For example, “Is it a man?” or “Is the person alive today?”. If I answer that what I am thinking of isn’t a person, then the game will ask about other broad categories such as “Is it an animal?” or “Is it a place?”.
The game applies a very logical approach and flow, but the questions it asks will vary each time. Even to the extent that the first few questions are very similar each round, the more questions that are asked, the more specific and atypical the next set of questions will become.
Define Your Hierarchy Of Questions
The secret to the game is to have a decision tree-like structure that is followed as more information is collected. For scoping analytics, start with obvious questions such as “Is the goal to predict something?”, “Is the goal to optimize something?”, or “Are you trying to automate an ongoing decision?”. It is possible to create a library of starter questions that will enable the scoping process to get started down a productive path from the outset. Perhaps as many as the first 10 questions would typically be covered using stock questions from the framework.
However, as each session unfolds, later questions will likely need to be determined on the fly based on the specifics of the request being discussed. It isn’t possible to have a list of every possible question that might need to be asked. After all, we have to win every game here, whereas the cheap electronic version only needs to win often.
Give 20 Questions A Try
Search for a website or smartphone app that provides a variation on the 20 questions game and give it a try. In order to really believe that the game works as well as it does, you may need to see how successful it is in practice.
After playing the game, use your new belief in the 20 questions process to adapt the approach for use in your analytics and data science endeavors. Done correctly, it can enable faster, more consistent, more focused scoping sessions. Better yet, everyone involved will look forward to their next game of 20 questions instead of dreading another open ended conference room meeting.