How to start your big data project

Experts at Computing's recent Big Data & IoT Summit discuss how to launch a big data project. Boards expect results, but not all projects deliver, so how should you deal with that?

Firms should start small with their big data projects, aim to deliver at least something early on, and ensure that they have buy-in at a senior level in the organisation.

That's the advice of a panel of experts speaking at Computing's recent Big Data & IoT Summit.

"One thing you need to ensure is you deliver something, even if it's just small," said Gael Decoudu, head of data science & digital analytics at Shop Direct. "So start slowly, and ensure your team and technology delivers on a regular basis, and then you can build on that. So many projects aim at the moon and don't land anywhere," he added.

Decoudu continued to give an example of an early big data project his organisation ran, which didn't follow this advice.

"We ran a project that was so big the first deliverable was only due in a few years. But of course things changed and it got deprioritised after six months, so it never completed," he said.

Earlier, the panellists admitted other mistakes they had made in their nascent big data projects.

Paul Pardoe, business architect at insurers BGL pointed to the need for executive sponsorship.

"You need lots of support from the executive board to invest," he began. "We appointed data directors in each business unit, and they created data teams which then had to then demosntrate value. So from there it was about picking the right projects.

"We want to get the most competitve prices for our customers, so foru us it was all about using data smarter. It was about taking data we already had and using it differently to improve pricing accuracy. That way, we demonstrated improved value, it's not that we did something entirely new," Pardoe explained.

"So it's all about the meerkats?" came a shouted joke from the audience.

"It's all about spending a big stack of money to work out who to give the meerkats to, so they come back," Pardoe retorted.

Dr Kevin Findlay, IT & digital board director, Complete Cover Group explained that not all projects will deliver.

"Boards want to see the predicted results, but data analysis is a science, sometimes it doesn't work," he argued. "You don't know that it'll be successful, but you can say 'there's a pile of data over there and it's not being used.' The board wanted a strategy to improve our pricing, and only way to do that was using better data," said Findlay.

Jude McCorry, head of business development at The Data Lab explained that big data projects are more likely to be successful when the board better understands what it is, and technologists and data scientists are better able to communicate.

"We learned the hard way that you need senior buy-in. Big data isn't just a fad that data scientists wants to play with, you need board sign off.

"We had issues in exec education around what data scientists are for, so we looked into how to do data storytelling. Also, giving your data scientists the skills to verablise exactly what their project can deliver is very helpful. You've got 15 minutes to ask for a spend of £1m, what's the best way to do that?"