Latest news insight

What does BARB know about Premier League viewing on Amazon Prime?

6 January 2020

This article by Doug Whelpdale, our Insights Manager, was first published in Mediatel Newsline on January 6th 2020.

For the 2019/20 Premier League season, viewers have been offered something new; a way to appreciate English league fixtures that has never been seen before. We aren’t talking about the questionable angles that VAR offers as an ‘enhancement’ to watching the nation’s beloved teams compete, but rather the chance to watch via Amazon Prime Video rather than Sky or BT.

For BARB, this is both exciting and frustrating. Frustrating, because Amazon Prime Video has chosen not to have its audiences measured by BARB. Even though we have the means to monitor such services, we cannot report the average audience figures for the Amazon-streamed matches in the same way as we do for BARB-reported channels. On the other hand, we can get some insight into viewing to such services by looking at the levels of unidentified viewing, which is exciting.

Unidentified viewing includes all TV set use that BARB cannot identify; this could be viewing to programmes more than 28 days after broadcast, gaming, or viewing to SVOD or online video services like YouTube. This kind of viewing now accounts for an average of almost 52 minutes per day for all individuals aged 4+; a 21% share of TV set viewing. The level of unidentified viewing is a reasonable proxy for viewing to SVOD services such as Amazon, because SVOD is a primary catalyst for the growth in unidentified TV set activity.

BARB Establishment Survey data show that SVOD services are increasingly popular in UK homes. By Q3 2019, 13.67m UK homes (48.5%) had a subscription to at least one of Netflix, Amazon Prime Video or Now TV, and 6.4m UK homes subscribed to Amazon Prime Video. As SVOD subscription levels rise, we have seen a strongly correlated increase in the level of unidentified viewing. Placing the relevant datasets together provides an R2 of 0.96 – close to a perfect correlation. So, while SVOD viewing is not the only element that contributes to unidentified viewing, it is a significant one. This means that we can look to unidentified viewing as an indicator of audience size achieved by Amazon Prime Video’s streaming of these Premier League games.

From December 3rd – 5th 2019, ten Premier League games were streamed live by Amazon. Viewing to these games in panel homes was registered as unidentified viewing. However, with six games taking place on December 4th, five of which kicked off at 7.30pm, followed by another game kicking off at 8.15pm, we had the prospect of large amounts of unidentified viewing taking place at the same time with a very likely cause.

The chart below compares the minute-by-minute levels of unidentified viewing on Wednesday December 4th from 6pm to 11pm with the previous five Wednesdays. Unidentified viewing on December 4th is noticeably higher. The average unidentified audience for the five previous Wednesdays was 4.66 million between 7.30pm and 10pm, while on December 4th during this time it was just under 6 million: 1.34 million higher.

Minute-by-minute data also show a pattern that we might recognise from sporting coverage. Unidentified viewing started to build from 7pm when coverage began, and was considerably ahead of the other Wednesdays by the time five of the six matches kicked off at 7.30pm.

Unidentified viewing levels for selected Wednesdays

Source: BARB All individuals 4+

When the sixth match began at 8.15pm, unidentified viewing did not shift greatly, perhaps suggesting that Liverpool and Everton fans may have been watching one of the other matches before switching to follow their team. At approximately 9.05pm, unidentified viewing started to drop, probably as fans whose team happened to be losing abandoned hope of a comeback. Unfortunately, the data can’t reveal to us how many Everton fans persisted in watching their team after they went 4-2 down in the first half. Nevertheless, we can see the audience drained away during the final minutes of that game as Liverpool added a fifth goal, before dropping back to more standard unidentified viewing levels for a Wednesday by 10:17pm.

This was, of course, not the only day Amazon screened Premier League matches. Nine matches were available to view on Boxing Day, with a further match screened on December 27th. Here again, there was a marked increase in unidentified viewing just before the first game kicked off at 12.30pm. This rose further as six more games began at 3pm and dropped away as those games ended at around 4.50pm.

We must, however, acknowledge that Boxing Day is a less ideal comparator given the unique viewing patterns that exist around Christmas. There is a clear increase in unidentified viewing when compared to Boxing Day for the previous three years. But, with more noise on this day and comparing over a longer time period during which SVOD subscriptions have grown significantly anyway, we cannot be quite so confident that football is the sole driving force behind the increased unidentified viewing here.

Nevertheless, the coincidence is difficult to dismiss. Returning to the fixtures on December 4th, we cannot be certain that the increase observed was caused by viewing to Amazon’s coverage of these Premier League games. We can, however, see that the most popular mid-week Premier League games on BARB-reported channels achieve audiences of a similar level. The Manchester derby on Wednesday April 24th 2019, for instance, achieved an average audience of just over 1.7m across Sky Sports Main Event and Sky Sports Premier League. December 4th offered the chance to watch six matches, including a Merseyside derby, but we should bear in mind that this was a new method by which to view, and although our data show that 6.4m homes have access to Amazon Prime Video, this still trails the 12.4m homes that pay for some kind of Sky or cable (mainly Virgin) subscription.

BARB has the ability to measure viewing to SVOD services on the TV set and on other devices. Amazon Prime Video also meets our requirement that any service we monitor has a regulatory licence underpinned by the EU’s Audio-Visual Media Services Directive (Ofcom in this case). Without Amazon’s active participation, we believe we are a year away from reporting viewing data publicly as part of our current panel via WiFi router meters that we are currently rolling out to panel homes. But if Amazon, or any other qualifying service, has the will, we have the means to offer precise measurement much more quickly. That really could be exciting.