For Ben Dias, head of advanced analytics and data science at Royal Mail, there are three non-negotiables when it comes to looking for prospective data scientists. Candidates need to want to learn, and be able to pick things up quickly, be able to program, and have a deep understanding of mathematics.
The latter is not overly surprising given Dias (left) describes himself as a ‘computational mathematician’ and that his undergraduate studies were in mathematics and astronomy. Yet the evident mix between maths and coding cannot be overemphasised. “If you don’t know the underlying mathematics that underpins an algorithm, you could apply it to the wrong data and if you make business decisions on that, you could kill the business,” he tells AI News.
“If you apply an algorithm to data, it’ll give you an
answer,” he adds, laughing. “I’m always looking for people who have a solid
maths background who can program – they don’t have to be software engineers –
and they want to learn. I can teach them to talk to the business, I can teach them
everything else, but if they don’t have those three it’s very difficult.”
Dias has been busy at Royal Mail since joining in 2017, building
up a team of 25 data scientists and analysts. It was an opportunity he had been
looking for, and Royal Mail gave him it on the condition of it being a blank canvas
to work with.
“They had set everything up nicely – there was a platform we
could work on,” he says. “It wasn’t perfect but they had done a lot of the
groundwork, and they said ‘you’re the expert coming in, you tell us what you
need and we’ll give it to you.’ They stayed true to their word, which was
The team works on initiatives split into three broad areas;
revenue protection and recovery, customer experience, and operations. The
latter is of particular interest because it covers both vehicles – optimising schedules,
maintenance, streamlining deliveries – and personnel. The algorithms for
internal use evidently make sense. These include predicting sick absences, as
well as correlating a link between doing too much overtime and becoming ill.
When an employee gets a certain amount of overtime, the model gets flagged.
It’s important here to determine between the B2B side and
the end customer who receives their post, thanks to the data science team, in a
more seamless fashion. “For the business customer it will be recommended
systems, customer segmentation, churn models, that kind of thing,” says Dias. “For
the end customer it will be things like the estimated delivery window for parcels.
We’ve built an algorithm for that.”
In terms of the sheer amount of data available, then the
numbers are unsurprisingly vast; billions of parcels and letters to more than
29 million addresses, collecting from more than 100,000 post boxes and 10,000
post offices. The overall amount encompasses B2B customers, marketing and
operations data, as well as Internet of Things (IoT) data coming from tracking
devices. “It’s a lot of data, covering a lot of asepcts of the company, and it’s
all available to us which is great,” says Dias. “It’s not always in one place
and not always instantly available, but all we have to do is ask.”
The result is what Dias describes as ‘about £50 million of
opportunities created’ in just the first year of building out the team, with
many initiatives now moving out of the pilot stage. This journey of ‘from zero
to data science’ is going to be explored further when Dias takes to the stage
at AI & Big Data Expo Global
on April 26 in London.
“A lot of people are wondering how to set up a data science
function or team and, because I’ve managed to do it in two years, delivered
stuff while building the team out, I thought it’d be useful to share what I did
and the things that worked, and the things that didn’t work, to help others get
there as well,” he says.
“It’s important for us to deliver fast because the hype
cycle is coming to an end – and when people hire they want impact immediately.”
You can find out more about the session here.
Interested in hearing industry leaders discuss subjects like this and their use cases? Attend the co-located AI & Big Data Expo events with upcoming shows in Silicon Valley, London, and Amsterdam to learn more. Co-located with the IoT Tech Expo, Blockchain Expo, and Cyber Security & Cloud Expo.
The post Ben Dias, Royal Mail: On building out data science teams and ensuring deep mathematical understanding appeared first on AI News.