How to Build a Foolproof Data Science Team – Data engineers, data scientists, B.I. specialists and data analysts. What do they do, which skills do they have and how do we fit them together in a productive team? Then you might say, “but the company I work for has no data or it has, but it’s a total mess”. Sound familiar? Then this article is also for you.
We are going to answer these questions by highlighting which roles a company should acquire first according to the maturity of its data capabilities. Also, we are going to provide you with further sources that will help you to deepen your knowledge on these roles that are, more than ever, rising in importance in companies of all kinds
Hi here’s Bernardo again. Let’s do a small exercise. Reflect about how the internet years ago transformed industries and the job market. This is actually quite useful to navigate the current changes that are happening in artificial intelligence and data analytics now. For example, in the 90s we started to exchange digital information with companies. Then, software engineers, web developers, SEO specialists, digital marketers, I bet you know many of them.
They became popular professions surfing the adoption of digital channels in our daily lives. As a consequence, this phenomenon forced companies to adopt at least some form of digital transformation. And this brought us where we are today: a stage in which billions of interconnected devices like smartphones and personal computers, the so-called Internet of Things, generates loads of data.
Companies can use this data to personalise their services and get insights on how to create new value propositions. Not surprisingly, artificial intelligence became much cheaper in the last five years. now we have the Internet giants from before: Amazon, Microsoft and Google leading the supply of cognitive services and online advertising. I like to see this as a long term phenomenon that started years ago with the digitisation process, going later to the digitalisation of information.
Here’s where all digital skills have risen in business relevance. Most companies were pushed to transform their business models into digital platforms and hire people who could do that. Now, because of the competitive advantage that data brings to business, Companies are also being pushed to adopt some kind of data transformation or A.I. transformation.
Maybe both, if the industries they operate rely a lot on predictable and repetitive tasks. And with this comes the first big misunderstanding from the company side: They think they need to first have all the data, properly stored and cleaned before they do any analysis.
This happens because they assume that the maturity of the analytics capabilities is the only driver of productivity. As highlighted in the AI Playbook from Landing AI, and this report from the BCG, focusing first on quick wins and pilot projects to gain momentum is an efficient way to discover which data to store and which features to engineer, so you can improve your analytics.
Examples of these quick wins could be running a data-driven personas analysis with existing CRM data, creating a simple predictive model that already performs better than a human decision-maker alone, or designing a data acquisition strategy for a product that will be launched soon. Okay, so then to run these pilot projects what a company needs to do is to contact a recruiter and find a data scientist with five years of experience, a PhD in machine learning and integrate them into the analytics team, right? Wrong!
If the company still does not have decision-makers who already understand why experimentation is relevant and what AI can do, then hiring data scientists at this stage will only bring frustration for both sides, There will be no actionable data-related questions being posed by management and no clear metrics to optimise.
So, in the first place you want to have decision-makers who have been educated on what AI can do, perform due diligence to identify business problems that AI can solve evaluate possibilities and find which ones are profitable and ethical. This is important not only from the company perspective, but also for job seekers. In order to mitigate professional frustrations, avoid offers from companies that have not yet educated their decision-makers. Once the company has educated the decision makers, it needs to have team members who can analyse data and are capable of, for instance, finding hidden connections or interesting patterns in data.
The skills needed here are a combination of applied machine learning, statistics and traditional analytics. Cassie Kozyrkov from Google uses these skills to define what makes a full expert data scientist. But she points out that hiring the true three-in-one is an expensive option. If you are on a tight budget consider upscaling and growing your existing single role specialists. For example, transforming your BI specialists or data analysts into data scientists.
And this is when another type of mistake usually happens. Companies very frequently don’t know which skills they need to develop in order to improve their data capabilities. They go straight to the job market trying to hire new specialists. Instead, they should map the existing skills, like in these skills matrix from HBR, to identify the potential in-house talent that matches the criteria for upskilling. For example, a BI or a data analyst who could become a data scientist Or a developer who could become a cloud specialist or a data engineer.
In her article, Cassie lists the top 10 roles in data science and suggests an order to hire them. Also, Landing AI suggests an education plan for C-levels leaders and trainees. Check this out later on. Now consider that data scientists need some data to start prototyping. For this reason, they end up doing some data engineering as well. They frequently land in jobs where the company still does not have structured data pipelines.
And this is when another mistake usually happens. Companies hire a business-savvy data scientist when they still do not have much support in terms of data integration. Or the opposite: They hire Type B data scientists – the builders – or skilled data engineers – the plumbers of data science when they actually need professionals that prototype and clearly communicate the results of the analysis to the rest of the organisation.
Incurring this mistake usually leads the company to suffer from the last mile problem, in which most of the data projects are never actually put into production because there was a hard time to explain the stuff to the decision makers. If you need to get insights and communicate the results, what you actually need is a business-savvy Type A data scientist. This professional will improve the momentum of your pilot projects by delivering an effective communication of the insights in a visual and engaging way.
So when developing your pilot projects to gain momentum, think that you have to build these skills, the analyser and the builder, the engineer, bit by bit at the same time. At some point you have to serve the data scientists with a structured data pipeline, and the end user with an API or a web app. Andreas Kretz from Plumbers of Data Science and author of the Data Engineering Cookbook has a podcast where he discusses the phases of the data pipeline that require a data engineer, a type A and a type B data scientist.
Finally, to follow up with projects across different business units, the company needs some dedicated analytics translators. They will facilitate adoption and the cross collaboration of the data team with the rest of the organisation.
In summary, as the internet brought software engineers and web developers as professions to the spotlight, the data transformation that companies are starting right now is also giving rise to more recently created roles. Think about how yourself or your company is going through this journey and what level of upskilling you need. Whether a short course for decision makers, or a longer course for new hires and data talent.
Leave your comments and suggestions below, and see you next time!
reference – How to Build a Foolproof Data Science Team
Web enthusiast. Thinker. Evil coffeeaholic. Food specialist. Reader. Twitter fanatic. Music maven. AI and Machine Learning!