Why William Hill is betting on open source rather than traditional IT vendors for digital transformation
The 80-year-old betting firm is now using tools such as Hadoop, which was developed by staff at Google; distributed database management platform Cassandra, which originates from Facebook; and Kafka, which was created at LinkedIn.
"The frameworks that these guys are spitting out pretty much as their exhaust fumes - that is the interesting thing for this particular corporate in terms of where we source our technology from," said CTO Finbarr Joy, addressing journalists in London yesterday.
"We used to be that entity that would rock up to the IBMs, the Microsofts the Oracles and seek 'what is our answer'. And we would buy a great big box and place it in the centre of the business and it would crunch all of the numbers and do everything."
According to Joy, the move to open source tools is being driven by a stronger focus on customers, which he says has been lacking from the major IT vendors. This differs from the intentions of the various open source communities and their 'bottom up' approach.
"The demeanour of that whole sector is to suit the IT buyer - there is no part of that conversation which is about the customer. We are certainly increasingly customer oriented, where we are minded to look more towards those communities that are building purely for the customer, and certainly coming out of initiatives specifically from the likes of Facebook, Google and Twitter," he said.
"So it is this bizarre element of the open source - the bottom up element."
Investing in digital
Such tools are helping William Hill deal with huge volumes of data it processes each day. "It is giving us the benefits of, for example, parallel distributed computing. You wind back a few short years ago and that was still in the realms of research. Now it is literally available for download," Joy said.
And the challenge it faces is significant - gaining an edge of stiff competition in the online gambling space means gaining better insight into the huge volumes of data generated by its customers.
According to Joy, William Hill supports 400 customer transactions per second, with the added challenge of prices changing five million times a day. This results in up to 160 terabytes of data flowing through its network during major sporting events.
Managing this volume, variety and velocity of data has been a challenge for William Hill and its competitors, and Joy said the industry has tended to lag behind other sectors in terms of its digital propositions.
However, he added that the use of open source tools is enabling the company to improve customer engagement, using analytics and machine learning to provide more personalised services.
"Now thanks to the power of these frameworks [such as Spark and Hadoop] we can actually deal with all of them [volume, variety and velocity]," Joy said, adding that it can now offer promotions based more closely on what customers want, for example.
"We are ending up with a specific betting system for each one of our individual users getting the experience that they prefer. And it almost becomes an agent acting on behalf of the customer."
One example of this is its new recommendation engine, a data aggregation platform which relies on Spark, Kafka and Cassandra, providing a real-time picture of each customer.
"It is attempting to look at everything that is happening in real-time - what the market is doing, what that customer's typical interactions are, presenting them with a recommendation for their scenario based literally on what might happen in the last twenty minutes. For instance across hundreds of races on a Saturday afternoon."
Evolving IT teams
William Hill's adoption of new technology platforms and application architectures has coincided with a change in the way that the firm's IT teams function and interact with the wider business. This means operating more dynamically, he explains, and means the company is no longer following five-year IT strategies.
"It leads us to a different way of operation. We have started to evolve how we work and certainly evolve what we used to call the IT department looks like, and that makes root and branch changes to how the whole business operates," he said.
"For instance, our structures are classically hierarchical, very much top-down, centrally driven, especially as we evolve these technology sets and then sort the best ways of applying them. Now we start to get into more of a network model, where you can literally see things evolve in something like a Darwinian basis."
In the past, technology projects would be continued "until the damn thing got over the line, regardless of how it was costing, regardless of how much of a doomed failure it was". Now the company is running "much smaller exercises, much smaller and coordinated teams".
A move away from centralised, monolithic applications to more complex architectures also means its software is much more reliable in terms of uptime.
"Any human intervention in any system is going to have at least 1% chance of failure," Joy said. "We now have several thousand things operating on our infrastructure and in our network, so a 1% chance of failure means we have failure pretty much repeatedly, but nobody notices and we change the practices and develop how we deal with that."
Joy added that the open source tools it uses are suited to use as part of its core platforms.
"We will see more and more businesses using [open source tools] in mission-critical scenarios. We certainly are, and we are seeing the material benefit in how well we can run mission critical applications compared to what we used to do. Because of the multiple instances it is actually easier to do that."