How to build High Performance Data Science Teams – Mike Hyde

This morning Mike Hyde (former Director of Data & Insights Skype. Previously pioneered Big Data projects at: AlixPartners, Opera Solutions) gave our Community a wonderful breakfast briefing full of advice on how to build the best data science team.

He started by reminding us it’s all about Build Measure Learn…not Build Measure DENY!

Skype’s data science team was 12 people managing 1 Petabyte of data and 1,000s of engineers. Mike explained how building his team was like an Opera “you have to think what skills you need in the round.” His advice was for you all to make your own Hex…

Starter for 10 – these are the slices of the Hex Skype focused on: New data and new metrics and governance, data processing and quality, data interrogation and access, analysis hypothesis and impact, Applied Data Science, Personal Leadership.

image1-6

He went onto discuss the Hub and Spoke vs Embedded model and how you have to find the right model for your company. Mike’s advice was “embed data science people where there is a critical need but also make sure you are teaching the buss unit to fish for themselves.” Also take caution, ensure that the central team retains the only CDO position and moves fast enough to avoid feefdoms of CDOs in each business unit. Either way you need a community of data people – wherever they sit in the organisation make sure they come together socially to share ideas.

img_0481

Advice on technology skills: “Every year there is a better mouse trap” technology is moving so fast it’s better to hire people with the right technical aptitude and ability to learn rather than the specific coding languages.

The most asked question was “who would you hire first” – Initially¬†Mike answered “Hire the CDO, hire the Unicorn” but tempered his answer and explained “it’s important to understand your problem first and focus your hiring on where the gap is – it’s very sexy to hire data scientists but you shouldn’t be in a rush, build the foundations of the house first!”

Published in
0 Comments

Leave a reply

Log in with your credentials

or    

Forgot your details?

Create Account