Some guidelines for the intimidated business leader to get the most of their data science team Credit: Thinkstock Business leaders and data scientists don’t always communicate particularly well. While some data scientists could do a better job of explaining results and storytelling, business leaders could also be asking smarter questions. The problem is, some business leaders are intimidated by data scientists, they don’t know what questions to ask, and perhaps they don’t know what questions they should ask. Here are my tips for getting the most out of your data science team: 1. Understand the basics Get some basic knowledge of the methods and solutions that your data scientists are using, and the complexities involved. It’s easy to gain a basic level of understanding, because many sites and solutions have easy to access data points and UX, for example. 2. Forget about the math … … but you need to be clear on the level of analysis and insight you are expecting, because data scientists have a habit of going as deep as they can. You need to think about what you want to do with any analysis, and how you want it to be presented, so its actionable. 3. Clear goals The biggest thing to remember, as the product owner, is understanding your end goal. This means knowing what you are trying to achieve, what is needed, and why and what level of accuracy is acceptable and actionable. This will have the biggest impact on how you brief your data scientists. 4. Select the right method for the time and data that you have A basic view of the data analysis methods that are possible, and available to your team, is a very important first step. You need some idea of the level of work that goes into simple methods of data analysis; for example, clustering or linear regression. This knowledge will put you in a good place to assess more complex methods like deep learning and convolutional neural networks (CNN). Then you and your team can decide which one is right for the task you are working on. There are loads of infographics to help you do that (I can even provide some, or just look at my LinkedIn profile). 5. How much and how long? Ask your data scientist how much data is needed for each task, and what the task is meant to achieve. For example, a clustering method will be fast and can get you 80 percent of the way. So, you need to be clear if that’s going to be enough. If it is, stop there. As the product owner, you need to set that level and communicate it to your team. A data scientist will always try to get better and deeper, so be aware of the level of accuracy you need to make the decisions you want to make. That will be one of the first questions your data team asks you. 6. Learn what “features” are … … and what they mean for the task at hand. For instance, a feature is information that describes, or helps a system detect, the item you are looking for. Think of arms, legs, head, and eyes, which are all features of animals. But glasses, hats, shoes, and clothes are probably only used by people. If you take out features that you don’t need, it will help to make the system more accurate and faster. So, ask your data team members what features they are thinking about using in their system and help them narrow it down. It’s also worth having a domain expert on hand to help with this step. 7. Don’t be intimidated by the math Math is not important. Most, if not all, machine learning is broadly the same in terms of the computation of vectors and the distances between them. What’s far more important is your ability to understand what you need to do with any insights derived from the data. Your data scientists will need to preprocess any insights to get it into a format that can be used. This can take time, so ask how much time your team will need for data preparation. They will also need a lot of time to test different parameters, so ask what their initial thoughts are about methods, as previously mentioned. If your data scientists jump to the hard methods that need lots of data first, ask why. Can they take a simpler approach in the beginning? 8. Data Find out how much data you have, and have access to. This is important to understand as some data is a lot more important. Make sure any data you are using is well reviewed for accuracy. Therefore, check what plans your team members have to ensure data accuracy and ask them who is going to cleanse that data. It may not be a job for your data scientists, either, because their time could be better spent on the analysis. Data ScienceCareers