Same game, new approach – how Fox Cricket used AI to predict wickets last summer
Cricket has been the nation’s sport for some time, but the problem is viewers drop off when nothing’s happening. Here’s how Fox Cricket used predictive AI to let fans know when the action was coming.
This article originally appeared in The Madtech Brief, Marketing‘s second print issue for 2019
Campaign: Monty – the greatest cricketing mind ever created
Client: Fox Cricket
Foxtel is Australia’s premium subscription TV network, home of Fox Sports, Australia’s leading sports broadcaster and publisher. The Australian subscription TV market is competitive. More than 70% of Australian homes are willing to pay for subscription TV and competitive subscription services have doubled in just the last 12 months.
In this incredibly crowded category there’s one product that cuts through: live sport. Foxtel’s broadcast deal for cricket was a strategically vital investment to secure a competitive advantage over the next six years, adding a strong summer offering to its existing leading broadcast rights for winter sports.
Expectations for the cricket launch campaign were extraordinarily high:
- acquire new subscribers while also reducing CPA by 40%, and
- steal share of free-to-air viewing from channel Seven.
This would be an intimidating brief at the best of times. To make matters worse, thanks to Australia’s ‘anti-siphoning’ sport licensing laws, we were asking fans to pay for a product they could still partially enjoy for free through a free-to-air competitor broadcaster, Channel Seven. We had to create a cricket experience unique to Foxtel, a new way to watch the game that would set our subscription offering apart, which also laddered into Foxtel’s consumer promise to launch a new Fox Cricket channel and to broadcast ‘cricket like never before’.
To create something new for fans, we had to understand how they currently enjoyed the game. Cricket is the soundtrack to the Australian summer. A constant background chorus of commentators, bowlers’ cries and classic catches compilations. But a typical match runs for five days and has a maximum of 40 wickets in 2700 balls. Each wicket lasts less than half a second. This all equates to just 18.2 seconds of truly meaningful moments in 30 hours. And with the lure of the beach, barbie and bathroom, we found that most people missed these key moments in live play and had to watch replays.
This was our key insight – most key moments in our audience’s favourite sporting pastime had never been seen truly live. Armed with this, we knew we had to deliver more value for the subscription and transform the viewing experience from a passive, lean-back experience (reacting to what you’ve missed) to a lean-in moment, primed for something crucial to happen.
In short, we had to predict wickets.
Cricket is a hugely complex game, with almost impenetrable rules (to some) and a huge amount of variables – pitch condition, weather, air temperature, moisture levels, speed of delivery, bowling speed, bowling type, right or left handed batsmen, googlies (if you don’t know, ask someone who does), slower balls, bouncers. The list is almost endless, making wickets almost impossible to predict… for a human.
Introducing Monty, the world’s first AI ‘predictive commentator’, which can tell fans when a wicket is about to fall and make sure they are watching when it does.
For those machine learning fans reading this, Monty is technically a supervised classification model. For Australian cricket fans, he’s pure magic. First, we trained Monty, feeding him an entire year’s worth of Australian cricket team data – every ball faced and every ball bowled across test matches, T20 and One Day International (ODI) cricket – each ball with 83 data variables attached to it. To ensure the model was match-fit we then challenged Monty to recreate some of the most seminal series in recent history.
Once we were confident Monty knew the game inside out, we created an API (application programming interface) output that could be used to trigger time-sensitive adverts across video, digital outdoor and display, directing subscribers to tune in to watch a wicket and maximise their subscription and directing non-subscribers to sign up to catch the wickets.
Instead of buying media inventory with our media budget, we bought data, investing 3:1 into data against standard media. The risk allowed us to play by our own rules. ‘Game changer’ is used a lot in our industry; however, we feel this is the first time it’s warranted as Monty has changed the live sport experience in Australia, offering fans a totally new way to watch the national game when it really matters.
Once Monty was trained, we let him loose on live games where he tracked 83 variables for every single ball during live play, overlaying his training data and the game’s momentum to project what would happen in the next five minutes. When he spotted a wicket coming, he triggered dynamic creative across pre-roll video, mobile display ads and outdoor billboards with a call to action to tune in to Fox Cricket and watch the wicket fall – making sure subscribers got maximum value and alerting the rest of the public to the fact that Fox (via Monty) could ensure they never missed a big play.
Through a Google integration, fans could also ask Fox Cricket for ‘Monty’s Call’ through the Google Home Assistant during live play and an API was also integrated into the Fox Cricket app for instant predictions and enhanced visualisations, which fans could use to help with suggestions on drafting their ultimate fantasy team. The true measure of effective innovation’s impact is whether it changes not only business outcomes but culture and consumer behaviour as well.
This project for Foxtel was an unparalleled success in this regard. It created an entirely new way to watch the national game and allowed fans to witness the best moments of every match as it happened live. Rather than offering data for targeting or as a visualisation after the event, we used data to predict sporting events – putting the viewer ahead of the game for the first time ever.
Our aggressive strategy to invest in data over inventory, and create an entirely new way to watch the game, paid off.
Monty over delivered against every core objective:
- Acquire new subscribers and reduce CPA by 40% – CPA was 61% below average (53% above target).
- Steal share of free-to-air viewing – Channel Seven share of P16-54 dropped to 20.8% from 49.9% (on Channel Nine – previous rights holder) year prior.
* * * * *
To purchase a subscription to Marketing magazine, visit the online shop »
* * * * *