Seven West Media is using machine learning to predict what 7Plus viewers are likely to watch up to 28 days in advance, forecasts it can use to shape content and pitch for advertising dollars.
Audience intelligence and growth director Andrew Brain told the Databricks Data+AI Summit in San Francisco that the company’s traditional audience and revenue reporting was performed on the previous day’s data.
But with competition accelerating in the advertising-based video on demand (AVOD) space, Brain said the company wanted to access predictive intelligence.
It has used Databricks – and the vendor’s professional services resources – to run a pair of experiments aimed at forecasting view patterns on a seven-day and 28-day horizon and is now expanding use of Databricks to power other analytics use cases.
“In short, what we want to understand is … how can we actually start to predict what our audiences are going to be watching over the next seven to 28 days?” Brain said.
Brain cited use cases of predictive intelligence across the 7Plus business, from marketing and sales, to finance and content teams.
“When we can actually look at what content will look like from a [viewership] perspective, and then we actually attribute revenue against that, that informs whether we are paying the right amount of money for the content, can we afford it, and what’s the long tail opportunity for us to make sure that what we’re actually trying to bring to the platform is monetisable.”
Meanwhile for sales, being able to accurately predict audiences is intelligence that can be used in meetings with brands and advertising agencies.
“When we take this to market and we go and show advertisers, media agencies and brand managers about forecasts, we can give an understanding of what that audience will look like over a campaign period,” Brain said.
“It makes them incredibly excited, but it also gives them confidence to buy with us, because we can understand these audiences over that [campaign] period.
“Whereas if I was to go to the content and the programming team and say, ‘Can you give me an understanding of what the audiences will look like in the next seven days?’ they will just give me that finger-in-the-air approach or rely on gut feel or years of TV experience to come up with some sort of percentage number, which isn’t good enough.”
An initial experiment over six weeks produced a seven-day rolling prediction of audience figures, with an accuracy above expectations.
“When we took that to our executives and said, ‘We can now understand what our audience is going to be from a value perspective on volume and also what they’re going to be watching’, you can understand that’s huge for our business,” Brain said.
“That clear seven-day opportunity is fuelling all the future initiatives through the roadmap. So seven days was great, but it’s a very short window, especially in the world of media and campaigns – not many campaigns only run for seven days.
“We wanted to take this and run it over 28 days, to work on marketing use cases or campaign windows for advertisers or brands.”
Brain said that Seven West Media is experimenting further by adding sophistication to the models. This means, for example, folding in weather prediction data and future public holidays, which could change viewership patterns and the propensity for audiences to return to the 7Plus platform.
Outside of audience prediction, Brain said another use case running atop the Databricks platform is targeting users registered with 7Plus but that haven’t engaged with the platform or viewed content for a while.
“If we can bring them back into an active state, it’s worth potentially tens of millions of dollars for 7Plus,” Brain said.
“What it also does is saves our marketing team from constantly trying to go out and find new audiences – when we’ve actually got audiences, it’s just about how we bring them back to the platform.”
Another package of work over the next 24 months will cover “ad load experiences and engagement”.
In short, Brain said 7Plus wants to experiment with the way advertisements are served to users, based on their time-on-platform.
If a viewer stays longer, “should we actually stretch out their ad pod breaks around their engagement, or because we know they’re going to be longer with us on the platform, do we actually pull ads out of the pod, so do we reward them for the longevity by actually seeing less ads?”
Similarly, Brain said the data might show that another viewer spends only a short amount of time viewing content, potentially put off by the number of advertisements served.
“That’s the thing we need to test: do we give that person an ad holiday when they come to the platform, wait until they achieve a certain frequency and engagement metric, and then start re-introducing ads?” Brain said.
Ry Crozier attended the Databricks Data+AI Summit in San Francisco as a guest of Databricks.