Kieran Sharpey-Schafer
Kieran Sharpey-Schafer and Lucien De Voux on site in Johannesburg, South Africa.

History and Origins

For as long as I can remember, my friend Lucien and I have argued over whose work makes more of a difference.  

We both studied computer science at the University of Cape Town, but we took different paths after graduating: Lucien joined the software team at RMB in Johannesburg, one of Africa’s biggest investment banks, whilst I joined Cell-Life, a small Cape Town-based NGO fighting HIV, one SMS at a time. We often bantered - do NGOs or the private sector have the biggest positive impact on real people?

As we gained more experience in both the US and Africa, we learned how businesses and governments work, and the role data and technology could play in local growth. Lucien saw how milliseconds won or lost trades, how the design of an interface could increase purchase volumes, and how responsible risk-scoring could unlock credit for people who had been ignored by banks for years. Meanwhile, I spent these years in all kinds of clinics with all kinds of health workers - from big city hospitals in Johannesburg, to sitting under a tree in Malawi with a volunteer armed with only a wooden box and measuring tape. This was a decade ago and even then, it was easy to see how phones and data were changing remote parts of Africa, places where the clinics had no doors and the ministries had no money for medicines. For me, it was a dream job. I was able to witness how quickly people could solve their own problems with small investments in their resources, infrastructure, and digital connectivity.

Identifying Problems

Despite all the amazing technologies in our work, there were problems brewing in both camps. In the NGO sector, technology tended to stop at the production of “indicator dashboards” - a pie chart or graph that never quite fulfilled its promise to trigger tangible impact. In the private sector, precise and practical data services increasingly focused on marginal gain use-cases for the richer top end of the market (e.g. improving cab rides or marketing automation), whilst interesting problems for the bottom-of-the-pyramid population were left untouched.

Connecting to Build Solutions

Ten years later, the argument was over in an instant.

Lucien and I got together to catch up one day, and was attempting to explain his new day job at a “big data” company in the US. He explained commercial terms like risk modelling, credit scoring, and identity verification, and how important they were in the US economy, especially for those without a traditional data footprint - like recent immigrants and young people.

After years battling with the development sector’s obsession with data collection and patient identifiers, it became painfully clear to me how far ahead the private sector was. Whilst governments and the development sector were scrambling data together to form a picture of what happened in the past (often years ago), corporations had already evolved to harnessing sophisticated pipes of the real-time data people generate every day. They could predict when you wanted beer, and when you needed nappies. Instead of being fearful of the uncertainty of real-time data and how different people behaved, they’d already embraced these realities and leveraged them to improve how they could sell toothpaste, or influence how you might vote in an election.

The key insight for me linked back to my work in public health: for HIV services, we consistently struggled with the inherent problem of an overwhelming demand for services whilst suffering an under-supply of resources to test, treat, trace, and retain patients. Most projects assumed all patients were the same person, with the same reasons for missing care or treatment. Meanwhile, the private sector had (for up to 30 years!) been developing triage techniques, allowing them to understand how both sub-populations, as well as unique individuals, all behaved differently.

The private sector might care less about why someone did something, but they were much closer to being able to predict or influence it rapidly. It became clear to me that the modelling techniques Lucien used for risk-scoring could also be used on the many HIV defaulter projects my team and our partners were working on in the region. These projects, which addressed "loss to follow-up" (patients who become lost at the point of follow-up in a trial or treatment), sought to understand quickly which folks were more at risk or vulnerable to dropping out of care. If the projects succeeded, the clinics’ limited resources could be targeted more effectively and efficiently for greater impact.

Palindrome is Born

In 2016, my idea won seed funding from the Global Partnership for Sustainable Development Data/World Bank Group innovation fund, allowing Dimagi (where I was working as a director at the time) to team up with mothers2mothers, UCT researchers, and DataProphet to develop and test a model for predicting defaulters.

This funding was critical, as we quickly saw that a) there’s real predictive signal about how patients are behaving beyond the more familiar linear relationships between demographics and outcomes; and b) there’s real practical need at the frontline for tools to help triage and direct resources for higher ROI.

It was a revelation to me, and Lucien was increasingly interested in the potential of these techniques back home in South Africa. The more we investigated, the more surprised we were by the lack of adoption in local tech and businesses in South Africa. Turns out there is lots of interest, plenty of skepticism, and a gap between the hype and real-world applications of advanced data science.

We thus got started on our new vehicle to meet these needs and searched for a name. The first data pattern I ever noticed was in my family tree: my sister’s name, Hannah, looked the same forwards and backwards – a special word known as a palindrome. Palindrome Data was born and we hit the ground running. Years of debate between school friends, comparing work and strategies across sectors, turned out to be useful qualitative research, as Lucien and I now work together to bring private sector solutions to development problems.

Learn more about Palindrome here, and feel free to reach out - we’d love to hear from you.

  

Contributing Partners

Palindrome Data Logo
Palindrome Data is a data science consultancy specialised in predictive analytics and advanced data services for the development sector.