Ten Rules for Building a Great Data Science Team – with Ben Dias
(Transcribed from an interview with Ben Dias)
- Focus on Retention First
Most companies focus on growth first rather than retention. There are 2 main reasons why it is better to first focus on retention. Firstly, great talent is hard to find, so you are better off trying to keep great people than by constantly hiring new ones. To retain great staff, you will have to build a great team culture. Have a flat structure, empower your people to make decisions, and get stuff done. Give them leeway to work 20% of the time on self-directed work. Secondly, by focusing on retention first, it will encourage you to start your team small but high quality.
- Deliver While Growing
You won’t have time to take 18 months to grow the team you want if you can’t demonstrate value. Take a portfolio approach, a bit like an investment strategy. Have a portfolio of quick wins, medium-term & long term projects. You have to manage this to create a system whereby you can dedicate time to deliver the quick wins whilst at the same time not neglecting the important mid and long term projects. Effective delegation plays a key role in achieving this successfully.
- Create a Culture where failure is a requirement
Data Science by its very nature is innovation, and therefore not everything is going to succeed. In fact you are going to fail a lot. You have to create the culture and understanding, with your team and the business alike, that we are going to fail and we should encourage failure. If not you will only ever make incremental changes and not transformational changes. Failure is Good! However, there are two rules to this failure that must be followed: 1 Fail Fast. 2. Learn from the failure.
- Find your Cheerleaders
Who are the blockers and who are the enablers? Find out who are the people within the organisation, that you can work with easily to get results and who are the ones where it might be more difficult. Why is this important? Because it is the cheerleaders who will be able to give you your quick wins which will help you down the line. They will also influence some of the blockers in the business to make it easier to work with them. Your cheerleaders will be instantly open to working with you and your Data Science team because they have major pain points that only you can resolve.
- Get out of the building
You must do this and you must convince your team members to do this on a consistent basis. The leader can’t be the only one responsible for identifying cheerleaders and projects within the business. Your team members should be incentivised to meet and get to know people in all areas of the business. Only by doing this can you start to drive the data science agenda throughout the business. Remember that this is primarily a listening exercise. Don’t spend the time banging on about the latest model you have built and how great it is. Instead, listen to understand the other person’s job, what success looks like to them, and any business issues or pain points they have.
- No Change, No Gain
However fantastic the models are that your team build, there will be no impact seen unless the change management in the business is done well.
Changing attitudes and perceptions within the business is more important than the quality of the model you build. Some companies have specialist change management teams to work with Data Science to do this. If they don’t it will fall on the Data Science team themselves. It’s important and it has to be done well to succeed in growing the function.
- Create an MVP Culture
Data Scientists and the Business alike should embrace a Minimum Viable Product culture. It doesn’t need to be perfect! If it’s viable, get the business to start using it right away and then iterate if/as required. In the long run this will save hundreds of hours of wasted time by ensuring that you are not building things that people in the business don’t want or need. It also enables you to start demonstrating value creation quickly/early on.
- Build your team brand
Why would the best Talent want to come and work in your team? This is a question that many companies struggle to answer effectively. The best thing to do is take action and build your employer brand in the Data Science community so that you are seen as a wishlist brand to work for.
It is possible to do this in a relatively short space of time by utilising the following: Social & Digital Media, Careers Fairs, Conferences, building relationships with academic institutions, sponsoring meetups & workshops, running your own hackathons, etc
- Start with No Processes
Don’t waste your time building processes for 6 months that are likely to be completely redundant. Every team and business is different. Just start doing straight away and you will achieve much more that way. Build processes as you develop to fix things that aren’t working. That way you know you are only spending time on something that’s worth it!
- Who watches the Watchmen?
You must collect data & gain insight into your own processes and performance. Ironically this is often something many Data Science functions don’t do. You are analysing the rest of the business but not your own team! Collect data & statistics on yourselves; track your own tickets, conversions & run rates, etc. as well as your business impact. You can’t get better and grow effectively unless you are measuring yourself and trying to improve.
Ben Dias is Data Science Director for easyJet, Europe’s leading airline where he will give even greater focus and weight to the airline’s use of data to create insights to enhance customer experience, driving revenue, and operational reliability.
Ben is considered one of the top Data Science leaders in Europe and has successfully grown a number of major corporate Data Science functions. He has been recognised as a Top 50 Data Innovator, and one of the Top 100 practising Scientists in the UK. He is a sought after speaker and contributor at industry conference and thought leadership events.