How to let your data science team succeed (part 1 of 2)

The first part of the “How to let your data science team succeed (or why your data science team will fail) Part 1 of 2

Don’t fit a solution to a problem

This may sound controversial given that I am a data scientist myself. But the myth of data science unicorns that can take your organisation to a magical wonderland is just that: a myth! Let’s talk about why your data science initiative will fail, how to avoid the pitfalls, and what to do in order to give them the best chance to succeed!

Anatomy of the data science unicorn

To begin with, let’s look at the anatomy (not literally…we have a heart too) of a data scientist. Having been an interviewer and an interviewee in equal measures, I have gathered the top requirements in a data science job spec:

• Academic: usually a PhD/MSc in a STEM subject
• Personality: Strong analytical mindset, equally strong curiosity towards unsolved problems
• Experience: Fluency in various programming languages (esp. Python), skilled in cloud computing/big data analytics

Depending on the level of the role, some may require 3-5 years of industrial experience, a proven track record of machine learning projects, and people/process management skills.

Why it’s hard for a data scientist to fit into traditional corporations

With the above requirements in mind, it is not difficult to understand why it’s hard to find data science talent in the first place. Not a lot of academics are ready/willing to leave the ivory tower, and those who did often do not have enough commercial awareness.

The perk of their curiosity is also their pitfall: far too many data scientists dive into the explorative rabbit hole and never get out. Very often they see a business challenge and immediately spring into action, using the most complex and bleeding edge solution to fit to a problem. But what the stakeholder wants is the result: simple, actionable insights; rather than a beautiful piece of academic paper that lacks connection to a real world problem.

From a data scientist’s perspective, if all that the stakeholders want are KPI dashboards and daily reports, it’s frankly not the best use of their skills; your organisation would be better off hiring data analysts/reporting analysts/BI wizards. An actual data scientist will soon get bored with the mundane tasks and start to look for new challenges elsewhere.

It is especially difficult for big traditional corporations going through digital transformations, where the concept of “unicorns” coming to save us jars with a fear over overly academic types vanishing down the reddit hole. In some unfortunate scenarios, I have seen organisations simply restructured their entire BI department and made everyone a “data science analyst” (which makes no sense from our perspective).

Are startups any better as a fit for Data Scientists?

On the other hand, the grass is not always greener for a data scientist in a startup, as very often they may be the only “data” person in the company and end up having to deal with everything,  from KPI reports (reporting – yuck!) to data ingestion (which is more of an engineering task).

Key points to takeaway

In summary,

• DS‘s are hard to find (especially if you stick to the idea of an academic unicorn);
• Just rebranding your analysts won’t help
• But when you find them, they are hard to fit in your org whether its corp or startup as they are easily bored and don’t necessarily focus on what you need!

To remedy this…

In short, from a business owners’ perspective, whether corporate or startup, the first thing to do is to identify what business challenges your organisation has, rather than starting from what talents your organisation wants; basically the good old: don’t fit a solution to a problem!

For a data scientist, again – don’t fit a solution to a problem! Identify the pain-points in the business, and pin-point them with simple, actionable solutions. Your stakeholders don’t need a 99.9% accurate (i.e. overfitted) model, so don’t let perfect be the enemy of good! 

… you can use the best of both worlds to build a high performing team…

Coming up next…

In the next part of “How to let your data science team succeed (or why your data science team will fail)” Part 2 of 2, we will go through some more Blockers of the DS journey, Common DS Projects, and introduce an Ideal DS Ecosystem. Stay tuned; and if this post has struck a chord, drop us a line at connect@leapbeyond.ai and let us help you to build a successful data science team!