31 Jan 2020, 09:50

Zindi helps attract talent and make better predictions at Kenyan logistics start-up Sendy

“Working with Zindi really improves our profile within the data science community. It lets people know we have data that is interesting to work with, and that we have challenges that make it appealing to work at Sendy,” says Marvin Kirschner, VP for Growth & Data at Sendy.

Sendy is a mobile-based logistics platform that launched five years ago in Kenya with bike couriers, and now provides transport options from bikes all the way up to 28-tonne trailers to customers across East Africa. The company is data-driven, with a dedicated data team that turns that data into insights and builds machine learning models to improve their business.

Zindi hosted the Sendy Logistics Challenge in 2019, when Sendy came to the platform looking for great ideas from the Zindi talent pool on how to predict estimated time of arrival (ETA) for deliveries. They had found that their delivery times were influenced by many factors that their existing prediction algorithm did not account for.

With more than 1100 Zindians signing up and over 16 000 solutions submitted towards a prize pool of $7000 USD, this was the most popular challenge ever hosted on Zindi. Kirschner says this is because they aimed for a challenge that was accessible and interesting to the data science community on Zindi, as well as immediately useful to the business.

“We looked at what data was available and suitable (i.e. a large enough dataset and easy to understand), found something that would produce useful results at Sendy, and made sure it was an interesting challenge,” he says. “Looking at the results, I think we achieved that. We decided on a challenge to predict ETA for deliveries together with Zindi, who really helped us to make the right decision.”

Sendy received the top three machine learning models submitted by the competition winners. These three solutions gave them unique approaches and new perspectives on the problem, which led to better business outcomes. By combining the solutions with their own work on this problem, they are already implementing a machine learning solution to predict confirmation time and arrival at pickup location.

But for Sendy, partnering on this challenge with Zindi has had another major benefit – the company values the exposure to data science communities, and the chance to contribute to growing data science in Africa.

“The two in-person hackathons for this competition, one in Rwanda and one in Kenya and South Africa, brought a lot of people together who later applied for positions at Sendy, which directly created a benefit in finding the right talent and growing our reach to the best applicants. It also had a positive impact on the data science community by providing local data sets that can be used in future to learn and train, and to build better companies in Africa.
“It was a great opportunity for a start-up; not just for algorithms, but also to make sure that data scientists know what you’re up to, what challenges you are facing and to kick of conversations within the community.”