How mobile apps made William Hill a real-time business

William Hill CTO Finnbar Joy gambles on cloud, personalisation, machine learning and the Internet of Things

Even if you're not a gambler, it's difficult to miss how large the betting industry has become. Millions of people have a flutter on sporting fixtures and people can gamble on almost anything ranging from politics, to the potential names of celebrity babies, to personal bets on weight-loss goals.

In a competitive field, William Hill is the largest operator in the UK with a 25 per cent market share and a workforce of more than 16,000 employees in nine countries. Managing the company's IT, therefore, is no small task, and responsibility for it falls to chief technology officer (CTO) Finnbar Joy.

"I'm responsible for ultimately pulling together the technology direction for the company, which means I'm responsible for the architecture development, test functions and building those solutions," he tells Computing.

The fast-moving nature of modern betting, whereby punters expect to be able to use mobile apps to place in-play bets, means providing customers with access to services "anytime, anyplace" is a huge focus for the company.

"We have to respect that our customers are being educated by the best in class [on smartphone apps], which means we have to be as good. We're competing for time with all those other apps, so ours has to be at least as good," says Joy, adding that the shift towards mobile is influencing how William Hill implements technology.

"During the last Grand National, we were in the top-10 downloaded apps. That is huge for us and absolutely orientates our direction," he says.

With major events like this year's Euro 2016 football championships increasing demand for services, William Hill has deployed a cloud infrastructure to better meet such spikes in demand. But, unlike most other companies, it hasn't outsourced this to public clouds belonging to Amazon or Microsoft, but instead built its own.

"We get the exact same benefit as external cloud. But because of our legislative practices, we've got to ensure we're managing data correctly, yet we can still be elastic and respond to events," he says, pointing out how regulations around handling data affect where it can be stored.

"We've got to make sure the online transaction is resolved in Gibraltar [where its online betting is based]. Unfortunately, Amazon doesn't have any Gibraltar data centres."

But while William Hill attracts millions of customers, the company wants to ensure that its punters feel that their personal interests are catered for. Indeed, such is the focus on personalisation that Joy even compares William Hill's capabilities for prediction to Google's.

"You can treat it as a unique, bespoke betting system per customer. This is all about the fact we know you in the same way that Google now knows you and knows what you want to do next; we're actively working with you on what you might want to bet on," he explains.

However, Joy is keen to point out that this is "not all about getting as much money as possible", and that personalisation and predictive analytics can be used to prevent people going "too far" when betting. "This mechanism gives us the ability to spot that a lot sooner," he says.

Of course, any company that collects information about its customers has to be explicit about how it will be used, and Joy describes how that's certainly the case for William Hill.

"We have to make sure that the practices are being made explicit to the customer for each step that they're taking. We don't identify a customer unless they want us to," he says. "It has to be done on a very transparent basis; opt in, rather than assume."

'The biggest challenge is keeping up'

But adapting to the balance between personalisation and privacy isn't the only technological challenge faced by William Hill; Joy describes the "pace of change" as one of the biggest hurdles which needs to be overcome.

"In my experience over the past 20 years, the 'time windows' between material capabilities are shortening. So by far the biggest challenge is keeping up. Any individual challenge you'll reference is underpinned by that," he says.

For Joy and his team, that means being agile is hugely important, so that they can respond speedily to the next big thing.

"We're spending a lot of time with the teams so they can move quick enough to get to the next thing. They are working on things that are discrete and small enough so that they're easy to change. The key is not to get locked into a machine that's going to take five years to migrate," he says, adding that the goal is to ensure "each team is only one or two iterations away from a rewrite".

"When that's true we'll know we're in good shape. We're not there yet, but that's what we're trying to make happen because we don't think we can second-guess which technology is going to be successful," he says.

How mobile apps made William Hill a real-time business

William Hill CTO Finnbar Joy gambles on cloud, personalisation, machine learning and the Internet of Things

The difficulty of predicting the next big thing means any organisation has to be flexible when it comes to development. But rather than seeing agile development as a trendy new thing, Joy tells Computing that it's an essential practice that should be the norm everywhere.

"The whole industry is ripe with that but, for me, I'm less inclined towards the ceremony around agile and, quite frankly, there's too much noise in terms of the Kool-Aid around it," he says. "Right under the hood [of agile] is what I've always regarded as good, core engineering practice."

For Joy, a good engineer is someone who "will always want to build something first" and that's an attitude he wants to encourage throughout William Hill's WHLabs, the company's experimental technology division.

"It's about how the rest of the organisation orientates itself to make that easy. And if we can keep those hops as small as possible, it means we're always in a recoverable shape," he says.

It's this attitude to technology that means William Hill is "always" looking for what Joy describes as "core engineers", because it enables the company to stay flexible.

"I'm less minded to what frameworks are in fashion now, but always oriented towards how we can find excellent engineering talent, because they'll be able to move as the industry changes," Joy says. "That means us being respectful to the communities, that's where we'll find them."

Joy says William Hill works well with open-source communities, with WHLabs sponsoring developer events and releasing code to be worked on during hackathons.

"I'm happy to say at some very large-scale events, some of our practitioners have taken significant prominence, which is great to see. It tells us that we're doing things right," he says, describing hackathons as "huge" for William Hill.

During one recent hackathon, Joy reveals, there were 11 external teams working with William Hill's API. "Out of it has come what looks like a genuinely useful application we can make use of and we didn't have to do anything," he says.

"We've done a lot of work to make our APIs usable and to publish them, but once that's there, it's out there to be exploited by a huge third-party community."

Continuing on the subject of innovation, Joy predicts that 2016 could be the year that the Internet of Things makes its presence felt, especially as people start to move away from smartphones to wearables.

"I think the fact that it's not always going to be about this piece of glass is just a matter of time," he says, picking up his smartphone.

"If you think about our retail operation, we have something like 10,000 machines out there. All of those [devices] connected to the Internet of Things gives us compound capabilities, especially when you combine it with a device," he says.

"Being able to trigger events when our user walks into a shop, that'd be an Internet of Things capability," Joy continues. "We've already got the devices out there and we'll be looking at Internet of Things capabilities to make more use of them so they can be more wired with the customer experience."

But while William Hill is keen to use connected devices and the internet to collect customer data in order to provide them with a better experience, Joy is under no illusions that the web also poses a risk to the bookmaker.

"It's constant," Joy says, on the subject of cyber attacks, "which means we're using a lot of technology to stop those patterns of behaviour. How do you spot a pattern that's different from the normal?"

The most useful method for doing that, he says, is by deploying machine learning algorithms. "It builds up a graph of normal activity and builds that up over time," says Joy, describing how, by using this technique, William Hill is able to streamline its response to cyber-security incidents, and also detect things that people would likely never spot.

"When things are happening outside of the normal patterns, it makes those explicit. Then you can zero-in on them. It's a good example of how a machine can learn what's normal and therefore spot anomalies much more easily than a thousand people," says Joy.

With machine learning being such an important aspect of William Hill's cyber defences, Joy explains that the company's security team consists of "software engineers more than anything else" who spend their time "looking at ways of automating and finding patterns of disruption".

"It's not about classic barrier security, it's about interacting and working with scenarios, and we're certainly using machine learning in that capability as well," he says.

Ultimately, the adoption of technologies such as machine learning and the Internet of Things is part of what Joy describes as "getting us increasingly towards that place where we can move in real-time with the market".

"We're never standing up and saying ‘we're done', we're in the habit of checking back on where we are on our journey now, but recognising it's about continual evolution," he concludes.