Problem gambling is moving up the agenda for gambling operators, as new data science techniques open up the possibility not just of a better understanding of what drives gambling behaviours, but also enabling operators to intervene and prevent consumers getting into trouble.
Mr Green is the latest company to announce the deployment of a new predictive tool that analyses gaming patterns and detects risky behaviours at an early stage of their development. Speaking at the release of the company’s latest results in late July, chief executive Per Norman said the Green Gaming Tool would allow the company “to a greater extent” to adapt its offering to risky customer behaviour in accordance with the company’s ‘healthy customer, healthy revenue’ principle.
The news followed on closely from the announcement earlier in the month from Sky Betting and Gaming which said it was funding a £5m academic programme, in partnership with William Hill, that will seek to develop a ground-breaking data science-based predictive modelling effort.
The company said that recent advances in data science are now being used to apply machine learning techniques and predictive algorithms to vast data sets to put even more customers in contact with information, techniques and controls to effectively manage their gambling online.
These techniques include suppressing marketing, setting deposit limits, self-exclusion, customer care and bespoke web pages that automatically make gambling support messages more prominent.
The company said that as part of the daily data crunching, the model takes into account all activity from the previous day and generates thousands of proactive outcomes. This includes a proportion of customers (circa about 5 percent of customers a day) receiving interventions ranging from less marketing, to emails about the tools available or a call from specially-trained staff.
The Sky/William Hill collaboration with psychologists will also frame and identify customer behaviours through complex modelling of the data to provide further learnings.
James Waterhouse, Sky Betting and Gaming’s head of data science, said the significant data science resource dedicated to this task was a sign of how seriously the company took its responsible gambling commitments.
“Academic input has proved invaluable, and we’ll continue to collaborate to improve our model accuracy to better steer our operations, CRM and marketing teams in their proactive messaging and processes,” he said.
Similarly working on data techniques and player behaviour, is real-time data analytics company deltaDNA. Chief executive Mark Robinson says the company’s system collects granular game data and then provides the toolsets to analyse it, and set rules for targeting player segments in real-time.
“These engagements can include triggered messages that warn players, or delivers incentives that break ingrained behaviours like burning all credits or taking increasing risks to win back losses,” he says. “These engagements can escalate to include placing limits on play for these accounts.”
Robinson says that building the predictive models to identify these players requires a data science capability. “Each game needs its own predictive model, and can use the behaviours of existing players who have self-excluded as a benchmark for determining the predictive factors and coefficients.”
He adds that implementation and testing are important too, as it is an iterative process, which ultimately leads to fewer self-exclusions, as players are supported towards healthier playing behaviours.
“We can provide as much or as little help as is needed,” he says. “We can set the in-game automation, and just as importantly, the exclusion lists for out of game offers, as you don’t want to mitigate play in the game, and then email an incentive.”
Another company working with a team of academics on problem gambling solution is BetBuddy which is working with City University in London on a platform to understand the pathways towards gambling harm.
Simo Dragocevic, chief executive and founder, says the system takes a three-tiered approach looking first at the exhibited behaviour using data collected directly from the player account management system; then it looks at the declared behaviour from a self-assessment test; and finally it looks at inferred behaviour and creating predictive models based in actual data gleaned from problem gamblers.