June 3 - 6, 2024
Commentary
Why Brands Need To Break Out Of Black Box AI
- by Eric St-Jean , Columnist, June 11, 2018
Marketers are hearing about artificial intelligence everywhere, and many technology providers are claiming some form of AI capability in their products — and for good reason. AI algorithms can churn through incredible amounts of data about customers, their behavior, and a business, ultimately determining the best course of action at a scale and pace that humans cannot.
For a marketer, AI can be a vital asset to a marketing stack. That is, if it works the way it should. But knowing if AI will work — or not — isn’t always easy, especially when it hasn’t even been tested or deployed in your organization. Marketers might ask themselves a few questions before adopting AI products:
- How do you know what works well and what does not?
- Can it assist in developing a strategy for the entire business?
- How can insights be gleaned and analyzed?
advertisement
advertisement
The problem with many of the tools available today is that these questions aren’t answerable. Instead, they create a “black box” effect, where the marketer may be seeing the results of what is happening, but does not get the information needed to understand why. That crucial “why” is important because it is from there that marketers get the insights that help to improve processes and iterate on programs.
These tools are often presented as flashy business automation solutions, but can only be helpful to a point if the marketer isn’t learning from them. That’s the opportunity that breaking out of the AI black box presents. Marketers can glean insight on why the tool made the decision it did, and can make strategic decisions based on content and be creative because of it.
Another pitfall of the black box effect is that it creates barriers to the adoption of AI in the first place. The misperception that marketers have of this tool — that they can’t learn anything from it and it won't help them do their own jobs more effectively — can lead to apprehension around testing out new technologies. According to a recent McKinsey survey, only 20% of respondents at traditional companies in healthcare, retail, and telecom industries said they had adopted one or more AI-related technologies at scale or in a core part of their business, and only 10% reported adopting more than two technologies.
Transparency and real-time analytics are necessary to boost confidence in AI and allow marketers to leverage the technology in a meaningful way. That was not previously possible, but is now — with the right solutions. By providing the ability to peek into the platform and its complexities, marketers have the tools they need to inform a customer’s next experience, as well as have more insight into the customers and what makes them tick.
Companies must break through the black box barrier to more effectively understand their customers and deliver the individualized experiences customers crave. AI-powered technology — especially in the world of personalization — will soon be table stakes. It presents marketers with the opportunity to not only better understand what content is working, but also why it is and what the customers it’s working with look like, which is a holy grail of information. This focus on insight is what’s helping early adopters get ahead, but the advantage won’t last, as others will catch on before they fall behind.
4 comments about "Why Brands Need To Break Out Of Black Box AI".
Read more comments >
Next story loading 
Well stated, Eric. It reminds me of the big debate about what ad effectiveness metric should be used for evaluating the impact of TV commercials three decades ago. One side contended that ad recall and persuavisness indicators as well as detailed viewer playback should be used, the other favored a much simpler approach, namely whether the viewer "liked" the commercial. Of course there was a corrrelation between "liking" and responding, however the actual effects varied all over the place and advertisers who adopted this simplistic barometer, rather than digging into the subject in more detail, were often stymied. When they obtained a favorable "liking" score, they didn't know why it was attained and what elements used in the successful message might be best applied to others. Just as with AI, you've got to know why it works---or doesn't---- in order to make the best decisions about how to employ and integrate it into your marketing programs.
Thank you Ed! I never really thought of the TV ad reward loop in fact, that's really interesting. I would guess that the feedback cycle would also be important here. For instance, "liking" is something that can be immediately measured, whereas recall, by definition, would entail a longer loop. But recall would also imply (or measure, even) a lasting effect, which is important in TV ads since the desired commercial effet is rarely immediate (or, at least, wasn't back then - things have changed now that everyone watches TV with a smartphone in their hand!). It would seem to me that one would want a metric that can be both measured soon after, and also have a lot less subjectivness to it.
Eric, having made an extended study of how branding ads---as opposed to direct response ads---work, it is clear that most of these are seen as "campaigns" rather than as immediate response efforts. Hence, the goal is to establish the brand's basic selling points among those who are most likely to be responsive so that this will have an effect at some time in the immediate or long term future when buying decisions are made. It is also evident that even if this is not the primary goal, that all advertising promotes the product category as a whole----including rival brands--- as well as the advertised brand.
It has also been demonstrated that current brand users respond best as its ads reinforce their existing convictions about the brand. The next best results, not surprisingly, come from former brand users who now buy another brand while those who have never used the brand are the hardest to convince. It has been estimated by scanner panel researchers that upwards of 75-80%% of the effects of packaged goods ad spending stimulates current brand users to restock while the remainder draws in new customers or brings back those who once used the brand but switched to another.
This is why, most branding advertisers with any savvy monitor how well their basic sales points are driven across and how this tracks over time. At a certain point, most campaigns wear out their welcome in terms of convincing more people to swicth to the advertised brand---but even so, they are often continued due to their ability to promote sales among those who buy into the sales pitch or, most important, current brand users.
That is incredibly insightful, Ed, thank you so much. Basically, what I'm reading is that long-running campaigns become a lot less amenable to machine learning, at least to how it's being used today, and very different strategies would need to be used to introduce it into the mix.