With a stream of metrics being recorded every few seconds and intermittently spat out by servers hosted in data centres around the world we ran the risk of being data rich and information poor unless the significance of what was being gathered could be visualised and deviation from the norm could trigger intervention before catastrophe.
UK broadcasters have long relied upon overnight viewing figure information from BARB which collects data from set top boxes in over 5,000 homes and delivers reports to the desks of channel controllers at 9.30am each morning. TV bosses have long measured success based on reach (the number of people watching) and share (the proportion of people watching). More sophisticated metrics like the Audience Appreciation Index slowly joined the stable of indicators.
The advent of video on demand, second screen apps, short content shared or promoted on social media have revolutionised the possibility of what audience behaviours and actions can be instrumented and measured. The challenge for the broadcasters is to avoid drowning in this sea of numbers.
Pedro Cosa is speaking at the Big Data Belfast conference on Thursday 2 June. He is Channel 4’s deputy head of data analytics. He explained to me that the broadcaster began seriously mining its data five years ago.
TV in itself as a medium has changed very dramatically in the past few years … the way that people are consuming is driving change. Content is available across many different platforms and many different formats. Live TV is still really big, but young audiences – that are more likely to watch Channel 4 – have new ways of consuming content which is starting to transform the whole industry …
We are changing from a broadcaster’s perspective which is ‘one to many’ (broadcasting to as many eyeballs as possible) to a ‘one to one’ relationship and we are now entering this world where you can engage on a one to one basis with each one of your viewers.
How is the availability of richer data changed Channel 4’s decision making?
Channel 4 has always been very open and ambitious with data. We wanted the data to be used across the channel, not just for pure commercial purposes (which in itself has got a massive advantage). Across every single part of the business we’ve been trying to investigate how data can help.
We’ve been creating a structure where we’re understanding how the business works, what the processes are, how decisions are made … and where we can identify where [each] opportunity is.
As well as the obvious feed into commercials/ad sales, Channel 4 are using the data they gather to better target their own marketing and promotion of content. But it doesn’t stop there.
We are also using this to enhance the relationship with our viewers. We now have 13.5 million registered viewers, one in two of the UK population aged 16-34 …
Pedro was clear that Channel 4 take their customer data seriously, with a “promise” to viewers around the use, retention and right to delete each individual’s records.
The most challenging aspect for his team is to “help the creative side of the business” with proofs of concept to demonstrate that data can lead to better decisions. “TV is very creative,” explains Pedro. Bringing TV and data together is like “a clash between art and science” with data sometimes seen as preventing creativity.
Many of the tools and techniques have been invented by Channel 4 as they ploughed their data furrow ahead of other broadcasters. Rather than collect data and then figure out what to do with it, Channel 4’s methodology is to start with a business challenge – something to change or fix – and then look at how the data might help.
Training and developing talent in this area is something we have to do for ourselves and for the industry. We’re conscious that Channel 4 may become the power house of big data for TV.
Channel 4’s own ‘IT Crowd’ aren’t hidden in the basement of Horseferry Road. Instead, the team deliberately sit up in the middle of the second floor amongst the rest of the business. The broadcaster works with universities like UCL to create a flow of graduates, with sponsorship of masters and PhD students.
Is serendipity lost when machine recommendations take over? Pedro admits that their video on demand service has so much content that it’s “almost impossible to navigate”.
Because we are a public service broadcaster we have this remit to fulfil. For us it’s not just about generating more views and more money, it’s also about fulfilling that remit. That means we wouldn’t be using normal recommendations to maximise the number of views, revenue and return.
So they’ve built their own in-house recommendation engine to avoid simply promoting already popular programmes and making them even more popular.
Are there surprises in the data?
More than surprises there are things you didn’t know existed but when you see them they make sense. We’ve been trying to understand relationships between programmes, especially with video on demand you can see how people are self-selecting what they watch (as opposed to the more linear schedule on TV where there is a legacy and you keep watching).
They’ve created a huge universe of relationships around how programmes are watch and “generated micro-genres”. As opposed to the well known genres like drama, factual, entertainment, Pedro’s team can see how very specific niche programmes cluster with each other (eg, teen American drama with a bit of comedy). This is helping drive commissioning decisions to fill gaps or feed interests.
Visualisation is a big part of our data programme. For every single thing that we do we have some sort of deliverable that is a visualisation. Normally these are the more interactive online tools so you can start moving things around and exploring.
When Pedro speaks at Big Data Breakouts, he’ll bring the perspective of an organisation that didn’t just start to exploit existing data but innovated to generate new data and use it for the benefit of the organisation and its customers (viewers).
Up and running for three years, the data analytics team stopped being a cost centre and became a profit centre within Channel 4.
We paid of all the investment in terms of our IT systems and people … It’s a very successful story [about] how we managed to quickly delivery back value to the business …
The team has learnt to say no, only producing reports when it’s clear what the business transformation will be on the back of the data crunched.
[To make] business impact and transformation we tend to avoid reporting for the sake of reporting, being just happy delivering a report and placing it on someone’s desk.
We want to make data actionable and see the results coming from it.
More information on the Big Data Breakouts website and Twitter feed.