Data driven authenticity - scaling influencer marketing with AI.

Recently, Andrew Chen of venture capital firm Andreessen Horowitz self-described a blaring, but commonly overlooked problem in the rapidly growing world of digital advertising: ‘The Law of Shitty Clickthroughs’. Chen documented that throughout history, novelty has served as a predictive attribute for the performance of advertisements. As brands scale their campaigns and competitors begin to mimic well-performing Ads, their creative becomes stale, and novelty dwindles.

To capture consumers' finite attention, brands have resorted to crawling into less occupied channels like SMS and voice assistants. As users consume an ever-increasing amount of content, attention spans have dropped, and tolerance for inauthentic sales pitches is at an all time low.

The overvaluing of short-term, quantitative data has distracted marketers from building trust, preferring manipulative tactics that try to sell, not market. So what’s the solution for brands? Market with people, not billboards.

Brands have already begun allocating resources to the quickly growing influencer-marketing sector. The common belief from outsiders is that influencers can coax their followers into eating up anything they promote. However, many brands that have launched influencer campaigns have found the results to be even more chaotic and unscalable than traditional channels.

Think of influencer marketing like talking to friends. If you were paid to sell products to friends by reading a sales script, they’d pick up on it almost immediately. Inorganic tone, body language, and vocabulary all illicit a defensive response, reducing the chance of a conversion to near-zero.

One of the few solutions to the Law of Shitty Clickthroughs is the frequent rotation of your creative (your ad format, message, and visual). This necessitates the ability to extrapolate the elements of your content that are actually driving conversions, and separate them from the “filler.” This isn’t so straight-forward, and is often only possible through A/B testing at a monstrous scale - an advantage that only the largest companies have. Even if you churn out new creative and perfectly maintain overall performance, you have only succeeded at beating other averages.

This is because every person responds differently to different messaging and visuals. The inverse relationship between the personalization of content and it’s scalability is obvious. A red phone case might convert better than green at scale, but still falls magnitudes behind an engraved variant with the recipient's name, and of their favorite color.

The modern solution to this paradox is to show the same creative to the fewest number of people possible by means of audience targeting. A common technique that brands employ is leveraging audience parameters such as location or interest in order to create and distribute hundreds of ad variations to large cohorts of people - something like “The cereal for Floridians!” to people in Florida, or “The healthiest whole-grain cereal around!” to nutritionists.

This works for creating contextual relevance, but not trust. Trust psychology is extremely complex by nature, even when restricted to the scope of marketing. Brands know about showing testimonials, money-back guarantees, and transparency. But, the decreasing effectiveness of these textbook techniques is a well-documented evolution of modern consumers. They see all Ads as a clever but untrustworthy “spin”, and know they’re being sold to.

The most trustworthy, and unsurprisingly most effective form of advertising is word of mouth. Hearing about a brand from friends and family is not only trustworthy, but also novel. Humanization creates novelty at its purest form. We have this kind of boundless complexity that makes our experiences and how we share them totally unique, and at a rate much quicker than manually tweaking phrases and visuals in digital content. Even better, people get used to our uniqueness the more time we spend around them. Our brains love patterns, meaning they’ll know exactly when you’re being you, and when you’re not.

This is where thoughtful marketers have the potential for massive upside over sales. Some influencers can emulate the deep authenticity of peer to peer. What I mean is, you need the right influencers, producing content that's true to themselves! People love consuming content, but hate ads. You should try and make the two as indistinguishable as possible. If you even get close, you’ll notice a substantial difference not just in conversions, but also in the lifetime value of your customers.

Most marketers don’t even try because they’ve been trained to read data as just that: numbers. This data in reality acts as a proxy for real human behavior - the kind of qualitative data that can’t be fairly quantified. It’s easy to overstate the significance of channel attribution data, performance bidding, and user acquisition modeling because it’s so readily available.

It’s much harder to dig deep, especially at scale. Many brands have found success “hacking” authenticity through repeat campaigns. An influencer might promote a product that’s totally inauthentic, but after the 30th campaign, it’s a part of them and seen as less “salesy.” This is because authenticity is partially a function of averages. The more you do something, the more it becomes a part of your behavioral identity that everyone begins to recognize.

Authenticity is just a single ingredient to building trust. The full picture is a kind of social relation reliant on accountability, something that’s difficult to express through a medium (the internet) compared to in real life. It’s easy for an influencer to rant and rave about how “amazing” a product is in a one-off promotion, and never speak of it again. In the absence of the kind of accountability you might have with a friend or family member, influencers should build trust through credibility. If you find influencers that are experts in their niche and would use your product, there's a good chance their followers will as well.

That’s only part of the equation, but it helps you to think about how your campaigns might progress. If you’re building a casual mobile game, for example, you might start by targeting influencers posting any kind of gaming content - perhaps using your own intuition as a heuristic for where to start. There’s a lot to consider. Should you pay them for a dedicated video, or a pre roll? Should you feed them a script and give them instructions, or should you just let them wing it? If you can afford it, you should test all of these things and optimize your targeting and content specifications based on the results.

It’s important to remember that the “results” will vary per influencer. Details even as microscopic as a change in tone between “normal” content and your promotion can have a substantial impact on conversions. Getting this correct requires collecting a huge amount of baseline data on past content from the influencer. What is their editing style like? Do they use big words? Live reactions, or voice overs? Just like talking to friends, the goal is to be natural.

As mentioned, certain attributes will work great across the board. Maybe influencers showing a face overlay and inviting their fans to play with them is a great call to action for all demographics. All of these things are worth trying.

Aggregating, processing, and drawing actionable conclusions from complex qualitative data is no easy feat, but it’s not insurmountable. At a small scale, much of this content annotation can be done manually. As you scale, it may be worth automating your data pipelines - the most effective way to do this likely being through implementations of AI-driven reinforcement learning, computer vision, and natural language processing.

As a collective, we must recognize that the ad industry is in an unprecedented era of user tracking and privacy concerns, and people are under more pressure than ever to spend. By focusing on positioning our brand seamlessly in the content people love to consume, we can re-establish trust with consumers, and be infinitely more likely to continue growth of our own.