Five Design Principles That Will Maximize The Impact Of Any Data Science Project
Few associate Design with Data Science.
It is much more common to associate it with the design of clothes or magazines. The truth is, the design process (or Design Thinking, as it is more commonly known these days) applies to any type of problem.
It is a known fact that Data Science projects rarely end up having an impact. The reasons are many: from a bad translation of business needs to technical solutions, to a poor understanding of the business problem in the first place.
The solution? Apply the design process to Data Science projects.
A great way to start is to incorporate some principles of design into our projects. Here are five guiding principles that every Data Scientist can apply to their job and that will maximize their impact on the business.
- The customer is in the center: think about the customer first and above all else. What are his needs? What circumstances does he find himself in? how is he going to use the solution you have to come up with? There are few things more frustrating that developing a “ “solution” only to find it doesn’t fit into de customer’s circumstance.
- Design the problem as much as the solution: Don’t take your customer’s word as sacred. You have to design the problem as much as the solution. Think of it this way: A well posed problem is already half solved.
- Build a portfolio of solutions: Don’t think you have to come up with just one solution. If a problem is challenging enough (and if not why are you working on it in the first place…) it deserves to be attacked from many angles. Try as many solutions as the resources will allow and see what is working and what isn’t. You can even combine solutions afterwards.
- Prototype and test: a big principle in Design Thinking. Before you sink considerable resources into a potential solution, build a prototype and test it. It is a good exercise to think what does this mean in terms of model development, but at the very least, try to build a model quickly with what data you have at hand, and do a small test before deploying significant resources into putting things in production.
- Be curious: Perhaps the principle that underlies all the previous ones. Don’t take anything for granted. Not what the customer signals as his problem, not even what you think might be the solution to it. Ask questions and test everything.
