Despite many half-baked efforts to standardize metadata and contextual intelligence in advertising, IBM is clearly taking our opinion onboard about this not happening any time soon. High on hubris, the company has decided to utilize its famed Watson supercomputer to drive targeted advertising in OTT on its own terms.
IBM’s Watson Advertising Accelerator is now looking further afield than display, with the company hoping that its AI capabilities can optimize OTT ad creative. Accelerator will be able to generate and assess individualized variations in video ad creative to achieve KPIs like video completion rate (VCR), various brand site actions and app installs.
Most importantly, its datasets appear almost entirely self-contained, with audiences segmented according to a rapid process of trial and error with ad creative variants. This insulates customers using the Advertising Accelerator from fragmented and dying metadata formats.
So, what is the Advertising Accelerator actually offering? First, it requires that advertisers load its libraries with as many variable creative assets as possible. These could be background elements, headline variations, different voice overs, sound effects and compositions. Watson will then assemble these assets as it best sees fit for each individual viewer, with the aim of driving audience engagement at scale.
Advertisers will be happy to do this if they can charge a premium for the content, but it might take a while to get creative departments on board with this new format of delivery.
IBM argues that AI can be used to improve the dynamism and creativity of ad creative by tailoring it specifically to individual viewers. The detail on the AI and ML capabilities are typically slim for PR materials using those buzz terms, but IBM says that Accelerator analyzes campaigns from multiple perspectives to rank content as high or low performing.
IBM argues that the other major failing of current digital ad campaigns is their reliance on contextual information, which is still too often the ID-based third party data assets that are slowly going extinct, such as cookies. But context can also mean less controversial data sets – simple metadata such as programming guides and genre labels for titles, for example. Is IBM planning on ditching those fundamental sets of tools?
Very possibly. It looks as though the Advertising Accelerator builds its own audience segments through a process of trial and error. It suggests creative variants that will drive outcomes from certain audiences and then puts this to the test, segmenting individuals according to their responsiveness to each variant. If repeated rapidly, these shots in the dark can quickly become a well-informed and self-sustained picture of audiences that will be highly valuable to advertisers in the shaky world of ad tech data formats.
So far, it seems the method has only been trialed on mobile advertising. IBM pushed its own Storm Radar app over 23 days and saw app installations increase three-fold, while advertising budgets were slashed by two thirds.
So, it seems to work, but IBM now needs industry confidence to scale these capabilities rapidly. It may seem like a Catch-22, but Accelerator has already partnered with AT&T’s ad tech unit Xandr to drive adoption among streaming platforms.
Meanwhile, the Accelerator can apparently manage without context by harnessing its powers of ‘prediction’, a conveniently vague marketing term that sheds little to no light on how these predictions are made.
But IBM does have a point that the traditional data cues from beloved by digital advertisers are on their way out, largely thanks to sweeping moves from governments and Big Tech.
One clear benefit of the Accelerator is that campaigns can be adjusted on the fly in real-time in response to early performance. Current OTT advertising procedure – what IBM labels Dynamic Creative Optimization (DCO) – is a much slower, more staggered affair, where a creative campaign can only perform upon existing assumptions at the time of starting, with performance assessment only following completion of the campaign.
Despite its claims to be an earth-shattering supercomputer, achieving much success in more industrial verticals, IBM Watson has barely made much headway in video. The rare occasions upon which Watson has graced the pages of Faultline tends to be for its voice-recognition capabilities when we assess the role of voice assistants in the connected home ecosystem.
We cannot help but wonder whether the Advertising Accelerator will be harnessed beyond OTT video platforms and advertisers. As Faultline has noted in recent months, there are other stakeholders now trying to get a footing in targeted video advertising, notably operators and OEMs. Both are well positioned with existing infrastructure and hardware footprints to facilitate targeted advertising and – in the case of OEMs – a holistic view of audiences on an individual level.