A couple of weeks ago, I was in Chicago for the start of the uptake.org data fellowship. It was a thought-provoking, stimulating and productive week. I came back inspired with new ideas on data and tech for good. I share three here, not because there were no more but because ‘357 lessons I learned from a week in a tech start up’ would have been somewhat off-putting for the reader.
First things first: What is the Uptake.org data fellowship?
Uptake is a fast growing start up that develops data science software for industry. Their philanthropic arm, uptake.org, created a programme to empower data scientists in the social sector through mentorship.
The fellowship consists of a week in Chicago where fellows meet their mentors, attend data science workshops, share previous fellows’ experience and talk through their data science problem. Following that week, fellows get six months of mentorship from three Uptake mentors. At the end of it, everybody gathers in Chicago again to report on what’s been achieved and share their experiences with the next cohort of fellows.
1) Lesson one: Think beyond impact measurement
The most common reason for voluntary organisations to collect data is impact measurement. This ranges from the simple monitoring of attendance to services, all the way up to complex experiments that scientifically demonstrate programme impact.
Having spent a fair share of my professional life looking at impact, I am not disagreeing with the importance of investing in monitoring and evaluation, learning from programmes, scaling up what works and scaling down what doesn’t.
But could this data go further?
Peter Bull from Driven Data presented a compelling case study. An organisation in Morocco has developed fog nets, which, placed at the top of a mountain, transform the fog into water to then serve communities in the desert that would otherwise have to walk miles to a well.
Using the data from the nets, as well as weather data from nearby meteorological stations, they created a predictive model that optimises the location of the next net.
This illustrates that a data-driven sector is not just a sector that can quantify its impact but a sector that embraces data and information technology to deliver the best service to people who need it most.
2) Lesson two: Learn to manage expectations
The use of machine learning for social good is quite divisive. It provokes reactions ranging from “I’ve been in this job for 20 years and I don’t see how a computer could help” to “surely, you can just make a thing that will solve all these problems”.
While both ends of the spectrum hold some truth, the reality often lies somewhere in the middle.
As a small data science team (sometimes, a team of one) in a busy organisation, it can take a lot of demystification and advocacy to get stakeholders on board with reasonable expectations. We were given some tools and food for thought by Matt Gee in a ‘demystifying machine learning’ workshop and by Liz Gagné who shared her tips to advocate for data science.
It may not be written in the job description but data scientists in charities have a role to play in growing the data science capacity in the sector to ensure that all the information collected is put to good use.
3) Lesson three: Get involved
With a team of two data scientists, NCVO seems to be way ahead of the game in its sector. Most people in the fellowship are lone wolves in their organisations, having to deal with everything to do with data and the IT infrastructure to hold it.
Having a team of eight fellows and three mentors to bounce ideas, learn with and get inspired from, was a really enriching experience. I came back to the office with a new perspective on the machine learning behind the Almanac and clarity on what is left to do and how to achieve it.
If you are a data scientist or an aspiring data scientist, get involved in the community, the likelihood is that you’ll find people who have already solved similar problems to yours and are happy to help. If you work for a charity or are interested in data for good, DataKindUK is certainly a great place to start: they match volunteer data scientists with charities who have data but not enough data science capacity in house. There are also active user groups, like this one for R and this one for Python, where you’ll find inspiration and guidance to improve your coding skills.
Finally, if you want to be part of the uptake.org data fellowship, check out this website to see when the next wave of recruitment starts!
Claire Benard is senior data scientist in the research team at NCVO. She collects, processes and analyses data to grow the evidence base on the sector. One major part of this is the UK Civil Society Almanac. 
