Channel 4 test shows AI-driven contextual ads double recall rates

Samantha Adcock explains the methodology behind the Channel 4 tests

Channel 4 has been conducting laboratory tests with viewers that have demonstrated the potential to dramatically improve ad recall, and also the ability to raise brand awareness, when ad breaks show brands that are directly relevant to programme scenes that preceded the ad break.

The company used AI (artificial intelligence) to identify what the broadcaster calls ‘contextual moments’ where a category of goods is used or talked about in a positive manner during a programme. It found that when strong and positive contextual moments were followed by a relevant brand advertisement, ad recall was 101% higher than the recall amongst a control group.

Speaking at the ‘Big Day of Data’ conference in London this week, Neil Taylor, Lead Data Strategist, Commercial, at Channel 4, admitted that his team could not believe the uplift themselves, so checked the figures and methodology multiple times. But the data was solid. “We realised that context was important but didn’t realise how important,” he declared. “Ad recall doubled across all the brands that we tested.”

The contextual opportunities are identified automatically but humans also review the context to guarantee brand safety. All content is pass through a software programme that uses scene analysis to recognise objects and scenery in shows.

Subtitles are also analysed to understand the audio (and therefore the script), with natural language processing. Opportunities are then rated according to ‘strength’ based on the extent to which an object is in the foreground or background, and then positivity.

Neil Taylor at Channel 4 explains the neuroscience behind ‘Contextual Moments’

If a pen is clearly visible in the foreground (in a simple example given by Channel 4), and someone says, ‘Who uses pens anymore?’, that represents negative sentiment towards the object in question. But if the pen is in the foreground (and so rated as strong on the scale of prominence) and someone says, ‘That’s a timeless classic’, then you have both strength and positivity and therefore what the broadcaster defines as a ‘contextual moment’.

Channel 4 wanted to understand the results from its tests and partnered with the Neuroscience team at Durham University, who at the time were studying their own hypothesis that content that has recently been ‘accepted’ into your brain is easier to code as a memory. In effect, our brains can be warmed up to receive more similar information.

At the Big Day of Data, the UK broadcaster showed a scene from ‘The Big Bang Theory’ where Sheldon decides that another household member has broken a flat-mate clause and that their friendship should be terminated. He makes his colleague ‘sign’ a release contract by using finger ID on his tablet.

The Durham University theory, which Channel 4 believes explains the massive leap in ad recall, is that when viewers saw this scene in the tests, their neuro-network relating to tablets was activated. It was easier to activate that same neuro-network when an advertisement for a tablet was shown in the break that followed the scene, which in turn helped memory.

In its consumer testing, Channel 4 also found that ‘contextual moments’ generated erroneous recollections that are very favourable to the brands whose ads were presented afterwards (and which were relevant to the scene). After the ‘Big Bang Theory’ scene, one of the ads shown in the next break was for Samsung, focused on a tablet.

When viewers were later asked which brand of tablet Sheldon had been holding, with a multi-choice list, a significant number named Samsung when the tablet was, in fact, an Apple product.

This tendency was demonstrated during a scene from the Channel 4 sitcom ‘Catastrophe’ where one of the two lead characters, Sharon, is told she has cervical dysplasia, a pre-cancer of the cervix. During the break that followed, viewers were shown an ad for the cancer support charity, Macmillan.

As part of the test, viewers were later quizzed about the programme scene itself, and specifically about the content of a poster that was in the consulting room where Sharon received her news.

The poster was about pregnancy support but many of the test viewers who saw the Macmillan ad in the break that followed the scene incorrectly recalled it as being a poster about Macmillan.

Referring to the market for products and services in general, Taylor said: “This is interesting for us, as a sales house, because this represents an opportunity for challenger brands to ‘hang on the coat-tails’ of major brands.”

Across all the brands that Channel 4 studied in the test, spontaneous awareness from relevant ads that followed ‘contextual moments’ was higher than from normal advertising. Compared to a control group that was not exposed to any advertising at all, normal ads played to test viewers resulted in 23% higher awareness.

For contextually placed ads, awareness uplift was 34% compared to viewers who saw no ads. Brand perception also benefited, with an average uplift for all the brands whose ads were contextually placed.

Taylor believes the recognition of ‘contextual moments’ inside shows is an opportunity for brands. He used the example of a dating app which, rather than just advertising around dating shows, could advertise in breaks that are close to scenes in sitcoms that feature dating or romance.

Samantha Adcock, Senior Research Executive at Channel 4, explained the laboratory tests in more detail. Consumers were shown three different programmes in their entirety, including the idents and promotions to mirror a real in-home experience.

One of these programmes, ‘Catastrophe’ contained contextual moments relating to pregnancy, smoking and cancer (the moment Sharon is told she has cervical dysplasia) during the first part of the show. Smoking and cancer were also identified as contextual moments in the second part of the show. A family planning ad was shown during the first ad break, and Macmillan had an ad in two advertising breaks.

Taylor summed up the opportunity for the broadcaster and ad buyers. “This is automated but brand safe. It [the success of the ‘contextual moments’ model] is driven by the ability to boost memory coding. It enhances effectiveness for advertisers.”