If you’re looking to intensify your data science efforts, it’s pretty fundamental for you to sit down with your current team and really ask yourself how you’re all going to tackle this.
If you’re growing a data team, your first reflex is to hire new talent of course; and get yourself a team of statisticians and unicorn data scientists to make your company data driven overnight. You should take a few moments to consider these three key points before rushing in.
This is a fundamental pre-requisite to building your data team because:
1.It will help you clearly define your objectives, the areas you want to focus on, to decide on a plan
2.It will help you set up attainable goals and communicate internally on what you plan to do
Every Company is different and has different data, as well as different goals for their data team. Start by asking yourself simple questions to define where you should put your efforts to grow the perfect team for your business.
You should aim to be as specific as possible when answering them. Your data team shouldn’t be about “leveraging your company’s data to create value and innovate.” It should be about “optimising the production line by reducing machine failures by 10%”, “increasing retention by one point by clustering clients and building a churn prevention model" or “analysing support ticket content to automatically determine subject and improve support response times by half a day."
These questions, and many others that come from them, will be the basis to defining WHY you’re building your team.
Choosing your stack and tools is going to be crucial for your team’s future growth of course.
So be careful and stay away from the hype. Always go back to your objectives, and what your team will be working on, to decide what technologies you’ll be needing; just because you hear talks and read papers on how revolutionary spark streaming does not mean you need to have a system that supports streaming, or live training. In most cases, you don’t.
You don’t have to completely rethink your stack every 6 months just because some new technology has come out.
Think about scalability. You can anticipate to a certain level how much your data is going to grow, and what new use cases you’ll want to set, or new data products you’d like to build. Of course, you can never predict it all. Check out this great talk by Criteo on how their system evolved to what it is today!
Plenty of others have written on which storage system, or database management system is better for what so we won’t go into details. However consider that you’ll always have an arbitrage between accessibility and stability. If you want your data to be extremely available to perform machine learning for instance, you’ll have to accept some loss of data and precision.
The next step is tools of course. Scalability is important in this case as well. You want a tool that works with different technologies you have available, and that’ll evolve to include future ones. Think about it, once you’ve gotten your team to work on a tool, it’ll be very hard to get them to use a new one just because your stack has evolved faster than your tool.
Having a tool that helps different technologies work together, will also help different people and data profiles to work together, whether its data scientists using R, Python, Scala, or MapReduce, you don’t want to have to refuse an interesting hire because his skills don’t match your tooling.
Speaking of people. Technology isn’t what will make your team successful though. The people in your team will.
Once you’ve defined what your objectives are for your data team, and what technologies will be involved, translate these in terms of skills you need, and skills you’ll be needing in the future.
Identify what skills you have, and what skills you already have available, and hire to strengthen them. Prioritise, and build a hiring game plan of how to bring future talent. At this stage, keep in mind that, as our CEO always says, the goal is to bring in people that are smarter than you, with skill sets that you do not possess.
If you are building a project to boost your marketing with a churn prediction project for instance and want it put into production fast, you’re going to be looking for someone with analytical skills of course, but also strong coding experience, maybe even a software engineer. They’ll be the ones to get ROI and proof of concept to the rest of the team fast. They should be working along analytical business users to get practical business knowledge and make sure the projects are operational.
If what you’re looking for is less urgency, and a more in depth understanding of your data to get actionable insights, you’ll be looking for more of a statistics oriented data scientist, with strong skills in data viz and communicating about data.
Always remember what makes a great data team is lots of people with very different skill sets, experiences, and tools, brought together by a clear understanding of the business goal of their projects.
For your data team to be successful, it shouldn’t just be great, with brilliant people working on what they love, it needs to be integrated within the company.
There is no miracle of Big Data. Even with the best people in the world, they have to work within the company to be efficient. Your data team has to involve all the other departments so their projects are understood and eventually used by the business end users.
That’s why it’s a key factor of success to have analysts working with the various departments within the company to bring solid business experience to your projects. Your data team shouldn’t just be data scientist superstars with extensive experience with MapReduce or Spark. For their work to be efficient, it can benefit greatly from analysts working with excel and building smart features with operational teams. They’re the ones that can make results accessible for business end-users throughout the company.
This is also the key to keeping your data team happy. You want to hire people that are aligned with the tasks they’ll perform. If what you’re looking for is an in-depth analysis with a lot of data preparation, you don’t need to bring in a 100k-a-year-data scientists, you can hire an analyst with basic SQL or even just Excel skills, and use their business sensibility, while growing their tech skills.
Before you go ahead and start interviewing for dozens of unicorn data scientists, look for people with a strong analytical mindset who can build their skills.
PS: We would like to apologize to all the serious companies' whose pictures we used to make this a serious looking article. Get in touch on twitter for any complaints.
Please fill out the form below to receive the success story by email:
How can we come back to you ?