Companies that effectively manage, analyse and understand their data can make the right business-critical decisions with speed and confidence.
There are very few businesses these days that don’t have a significant amount of data or acknowledge the need for effective data analysis and management and this had led to the exponential growth in demand for data scientists and other data professionals in recent years, along with the number of people adding a ‘data’ descriptor to their CV.
However, despite this significant rise in demand, the subject of data science remains a mystery to many, making it hard for them to identify exactly what they need and therefore recruit the right person/team to deliver results. It also makes it easier for candidates to slip through the net based on a spattering of on trend keywords in a CV designed to catch the eye of the recruiter. R – tick, Python – tick, SQL – tick, interview – not necessarily.
When I first started interviewing and recruiting data scientists, predominantly for my own needs, I made mistakes. I allowed myself to be blinded by the impressiveness of the CV and ignored my instinct as it questioned the style and cultural fit of the individual. It meant I ended up with what I consider to be the academics of the data world; individuals who were hugely intelligent, able to write incredibly complex code, but without the commercial nous, respect of deadlines and self-starting initiative that I needed. They did a great job doing what was asked of them, but never stopped to consider if we were asking the right questions of the data. They disappeared off into the quiet spaces and emerged with work when it was finished rather than when it had been asked for.
Since then I have gained significant experience in both recruiting data scientists and understanding how to work with them best. Here are some of my insights:
1. It is really important to correctly define needs up front
2. Setting candidate homework makes for a much more effective interview
3. They like an intellectual challenge
4. They need the time to get it right
5. What is not on the CV is more important than what is
6. There are as many different personality types as there are for any other role
Through my experiences, I have developed a very refined sense of what makes a good data scientist which I have deconstructed into a four-stage process. Like any other role, it starts with challenging what exactly is the business need and how has it arisen. Then it is about understanding what kind of person and culture is successful in the business. Thirdly, I work to understand the team environment and dynamics in which the person will be operating. Finally, I list the skills and expertise required. Note, qualifications and years of experience do not come into it. One of my best hires had a CV you would skip over – non-relevant degree, most of his career working as a police officer, no formal analytics qualifications and only two years of data experience. However, I chose to interview him because he had taught himself Python and various other languages ‘for fun, in his spare time’. He had all the attributes I was after – able to think about the problem from a business perspective, think about what would be interesting to find out and would proactively come up with more insight. He demonstrated a great deal of initiative throughout our time working together which was exactly what the role needed and gave me the confidence that we were maximising the use of the data we had to benefit the business.
I recently had a very poor experience with my bank when trying to set up a new business account with them. I will spare the ins and outs, but each team appeared to operate on a different system so there was no single view of contact with me, processes did not seem to consider existing customer data (leading to a bizarre request for me to provide evidence of a personal bank account despite having banked with them for about 15 years and already passed that part of the checks). People can hide behind ‘legacy systems’ excuses, but that is not an excuse and that is where the right data scientist(s) come in to play. A bank holds a huge amount of data about each customer and their transactions. Our expectations are that this is used to create a seamless customer experience. The right data scientists can deliver the architecture to make this happen. They will take the time to understand the problems currently preventing a joined up view and then work through each and every one tirelessly to design an effective solution. Without them, established banks will fail. The challenger banks have already architected around idealised data maps and recognise the benefit of extracting every last ounce of value from the data they hold. This is what customers expect. This is what all banks must recruit to deliver.