What takes a Data Scientist to be a Great Data Scientist?

Over last 13 years, while building Data Science teams from scratch for multiple organizations, I have come across many Analysts, Data Scientists, Data Engineers (add as many titles you know here) from our internal as well as partner companies. Some of them created a strong mark in my life and I would feel proud to have them as my colleague in which ever organization I am at, or to recommend them to other organization where I know they would get good growth path.

At the same time, what surprised me, when I saw the same person, in one organization did so well, but in another organization wasn’t adding enough value even though they are capable of. This made me to think what makes one data scientist to rise up to a Great data scientist?

I spoke to almost 300 of them (Junior/ mid senior to senior) and observed their progress minutely, and the only word that remained common among them were ‘culture’. Many researches have been done how culture improves productivity, so, I won’t go to that generic path. It’s as easy to say the word ‘culture’ but it’s equally difficult to define it. I have found the largest of them – at least the one that matters to Data Scientists – and this can be applicable to Sr leaders who are trying to build a great data science team and for individuals, who aspire to stand out as a great data scientist.

Knowing WHY you work drives HOW you work, which in turn makes you stand out

Let’s first see it from the lens of a Data Science Leader

You want to build a team of great data scientists – who are motivated, takes ownership of the projects, does quality delivery on time, proactively experiment. Well, I found many JDs actually write these are individual behavioral criteria of hiring too. Yes, every individual has to come with certain level of these characteristics as part of their behavioral traits, but a lot depend on the leader, whether those highly talented and highly capable team members will show case those on the job or not. And it only depends on – whether you have been able to create the ‘reason’ for them to do so. It's the same thing that differentiates a great teacher from any teacher. A teacher teaches, but a great teacher creates inquisitiveness among the students, and they then study themselves.

Let me take an example – you have a team of 6 data scientists. 2 of them (Team A) working on a consumer journey funnel analysis. They work on it very enthusiastically, showing where the drops are happening. Whether the journey improving over time or not. It’s a very elaborating work they have done and as you take those to leadership team, they loved it and asked you to update the same analysis every day or every week. Now this team converted their work into an automated visual dashboard and delivered. After few weeks, Product Lead calls you and says that one time visual dashboard is not enough, because the dashboard needs to incorporate regular product changes happening in every sprint release. Thus they need to look at it and update the dashboard regularly. Your team finds that very tedious, plus involves manual works, because it needs them to learn what all changes has happened in the consumer journey, change the funnel, change the measurement once every 2 weeks. They find it not so cool… Slowly they go demotivated. You force them to do, because this is important, they do it just because they have to do, thus quality goes missing over time.

Have you faced this? Do you see what went wrong in this whole process?

In the whole process, you explained them ‘what they have to deliver’, but you didn’t explain ‘Why are they doing this analysis?’

Let’s assume, you instead did the following:

On day 1 when they started the journey analysis – you explain to them that for the business as a whole we have a target to improve ‘consumer conversion by x%, because if we do this, it can improve our revenue by y%’. We need to analyze all the journey stages where high drops are observed, discuss with business how to improve those. Strategies together. Once business implement some new steps, we need to measure and find if the situation has improved. We need an automated mechanism to build real time/near real time triggers if it drops below a certain level so that business can take action immediately. And finally, we have to achieve the conversion target. Let's brainstorm together and see how best plan this?

Believe me, you should have seen a different level of ownership, dedication, proactive collaboration with other business functions, timeliness of delivery, Agile thinking from your team. It just starts with that one alignment – ‘WHY’. It’s like a large ship rowed by many people. When everyone knows that 1 destination and how each of their role help the ship to reach that goal – it works like a self-motivating self-fueling mechanism.

 Now let’s see from the lens of a data scientist – at junior or mid senior level role

At times projects come to you in a very abstract format (where you are not sure how to break it down) or as multiple specific tasks.

For example: Someone asks –

(a)   Run a consumer segmentation model and show what different persona of consumers are using our product/service. – This is called ‘Abstract’ problem.

(b)  Number of consumers coming daily on our platform vs once a week vs once a month. What % of them order from us and what’s per consumer order value? – This is called ‘Specific’ problem.

 In either of the cases you have 2 options:

A)    If it’s specific, then do the analysis and share the data as asked. If it’s ‘Abstract’ then do a first cut as per your understanding and share. If business wants something else they will anyway get back.

B)    Speak to the person who has asked for them, and ask ‘The Context’ behind these analysis.

If you follow (A) you are a data scientist, but if you do (B) you are the one who stands out.

Here is why:

When you ask the context – you learn about the larger business problem. You learn why they are asking for these data. Once you learn the background, you can think better, and you would be able to bring bigger value than just delivering what they have asked. Let’s take the case of (b). When you ask them, they say “well, you know we have to improve our quarterly revenue by x%, and one of the way to do that is by converting as many user coming to our platform to purchase, and those who are regular on our platform, we are trying to improve their cart value”.

Now when you know this, you will not just share the numbers/ information they asked for. You will go extra mile, and do a proper RFM segmentation. You would profile those segments with kind of products each of these segments buy, so that business can target them with similar or complementary products. As a next step you can even do a ‘Next Best product to buy’ model, where for every consumer/ micro segment you tag a set of 3 products they are highly probable to buy. Your analysis adds way higher value for business to take action, and Your value, in turn, to the business team grows by 3-5X. Because you move from a ‘doer’ to a ‘thought partner’. 

Could all of us data science practitioners take a pledge to ask 'WHY', before jumping to any analyses and elevate ourselves to 'thought partner' level?