I recently gave a lighting talk at PyData Meetup London where I talked about ‘Consulting skills for Data Scientists’.
Here are the slides here
https://speakerdeck.com/springcoil/consulting-skills-for-data-scientists
My thoughts
Some thoughts – these are not just related to ‘consulting skills’ but something more nuanced – general soft skills and business skills – which are essential for those of us working in a commercial environment. I’m still improving these skills but these are important for me and I take these seriously. I present some bullet points that are worth further thought – I’ll try to tackle these in more detail in future blog posts.
- Business skills are necessary as you get more experience as a data scientist – you take part in a commercial environment.
- All projects involve risk and this needs to be communicated clearly to clients – whether their internal or external.
- Negotiation is a useful skill to pick up on too
- Maturing as an engineer involves being able to make estimates, stick to them, and take part in a joint activity with other people.
- Leadership of technical projects is something I’m exploring lately – a great post is by John Allspaw (current CTO of Etsy). http://www.kitchensoap.com/2012/10/25/on-being-a-senior-engineer/
- My friend John Sandall talked about this at the meetup too. He talked more about ‘soft skills’ and has some links to some books etc.
- Learning to write and communicate is incredibly valuable. I recommend the Pyramid Principle as a book for this.
- For the product delivery and de-risking projects – I recommend the book ‘The Lean Startup‘ can be really good regardless of the organization you’re in.
- Modesty forbids me to recommend my own book but it has some conversations with data scientists about communication, delivery, and adding value throughout the data science process.
- Editing and presenting results is really important in Data Science. In one project, I simplified a lot of complex modelling to just an if-statement – by focusing on the business deliverables and the most important results of the analysis. Getting an if-statement into production is trivial – a random forest model is a lot more complicated. John Foreman has written about this too.
In short we’re a new discipline – but we have a lot to learn from other consulting disciplines and other engineering disciplines. Data science may be new – but people aren’t 🙂