In the age of AI, everything should be For You.
Not just TikTok. Or what you do in web3. Or the AI chat app you talk to most.
Every website should welcome us by name. Every piece of marketing should speak to our own interests and our own past purchases and whatever’s in our closet.
explained Adobe’s Scott Belsky last week in an interview with Bloomberg.
These are things that should be happening but aren’t and it’s because the data layer isn’t there.
Candidate data layers for this purpose are not well understood. We’ll study existing ones and then apply Stratechery’s Aggregation Theory to explore the new power dynamics when users become their own bitch and aggregate the aggregators.
Existing data layers
The existing data layers are either insufficient, bad (in quality or legal compliance) or inaccessible.
Onboardings
To solve the personalization challenge and bolster lacking first party data, some companies have started adding richer onboarding flows, inviting users to share more about themselves to get a more personalized experience. For instance, eBay has added a new discovery section (notably gated behind a login wall) that invites users to share a bit about their interests and specializes the experience based on these choices. This approach has the problem that the more extensive the onboarding, the more it churns. This places an effective limit on the power of onboardings.
DMPs
The leading data layer for this purpose are DMPs like Epsilon, Acxiom or Zeta Global, who have transaction and activity data on over 200 million Amercians. These data layers are hooked into enterprise CDPs. They’re rumored to source data from all the same places, with small differences on the margin (e.g., Zeta Global purchased Disqus in 2017).
The whole who’s who of Corporate America buy this data.
But this candidate data layer is under pressure for three reasons. First, growing privacy laws and actions by the FTC are limiting data broker data access and continued growth. Second, data broker data quality is on the decline. Last year we empirically noticed the decline in data broker data quality. Because data brokers source their (often snapshot) data in secret backroom deals and operate on identity that’s duct-taped together, their ability to stitch together data on individuals from multiple sources is tactically challenging at scale.
Finally, the effective death of cookies makes DMPs hard to use at the individual level. Brands have typically used DMPs hooked into CDPs as a means to wield additional context for individual users even for those who don’t log in. Brands used third party cookies and ID-bridging to identify users and then performed a lookup in the database to see if they could find data associated to that user identity. The death of cookies makes these practices harder.
Big Tech
Data aggregators like Apple, Meta, Google or Amazon could be candidates for this data layer but their data layers are private for really for their own exercise within their own private networks.
Four years ago – per Wayback Machine – you could use the Instagram Display API to access a user’s profile and posts. Last month Meta deprecated it.
There appears no API to access photos on iPhone though Apple syncs photos to iCloud for an estimated 2 in 3 Apple customers and holds encryption keys for most users.
Spotify has a developers API to allow developers to access users’ listening data and favorites but their Terms restrict using that data with an AI model – see Section IV(2)(a)(i).
Amazon has an API but it’s only for merchants and its terms prohibit the use of data extraction tools, reserving the right to terminate your account at its discretion.
Putting aside access, even the internet’s largest aggregators only see a slice of you.
For us, All For You means
- All interfaces
- Based on all context
All For You.
Existing data layers don't work. We need something new.
Aggregating the Aggregators
Aggregators aggregate a lot of things. Google aggregates information. Netflix aggregates content.
But they also aggregate data. Data is critical to securing aggregator power, in that it enables provision of protected discovery mechanisms to sort through the aggregated abundance of supply.
At Crosshatch, we believe this discovery mechanism is best owned by the user. The user is the final boss of data aggregation. The one with Data Gravity. Only the user has the Right To Access all the data collected about her. All other aggregators only know what the user does with them.
This also makes sense from a privacy perspective – third parties should only know what's "reasonably necessary and proportionate" to their use case anyway.
Aggregator Theory
Ben Thompson introduces Aggregator Theory via Clay Christensen’s Law of Conservation of Modularity.
Commoditizing an incumbent’s integration allows a new entrant to create new integrations — and profit — elsewhere in the value chain.
Now consider the case where users aggregate the aggregators: aggregating all their context and AI; and permissioning it to connecting applications.
Previously apps integrated logic and intelligence. User context was modularized, collected through interactions logging or onboarding flows.
But in a case where users aggregate their context and permission their AI Agent – AI and context to be used for a particular purpose – the App becomes modularized. Thompson explains
More broadly, breaking up a formerly integrated system — commoditizing and modularizing it — destroys incumbent value while simultaneously allowing a new entrant to integrate a different part of the value chain and thus capture new value.
This is poised to flip power dynamics of the internet back to the distributor (or even the supplier, who might soon wish to control her own distribution). Thompson explains in an earlier piece
The value chain for any given consumer market is divided into three parts: suppliers, distributors, and consumers/users. The best way to make outsize profits in any of these markets is to either gain a horizontal monopoly in one of the three parts or to integrate two of the parts such that you have a competitive advantage in delivering a vertical solution.
The biggest companies of the internet era came to become horizontal monopolies controlling supply via a virtuous cycle of ever better user experience. And while the internet initially made distribution free (pressuring the pre-internet controls on distribution), today distribution is expensive – the CAC is too damn high.
This is why we’re seeing every major consumer internet company lean into loyalty as a pathway to reduce distribution costs.
And when users can aggregate their own data and AI – be their own bitch – so that incredible personalized experiences are available to any app, power dynamics shift.
While the internet era commodified supply and fed off a flywheel of ever improving user experience through data and proprietary algorithms, an era where consumers
- Aggregate the aggregators
- Be their own bitch
- Bring their own data and AI to their favorite apps
the supply-side gains power.
Ben Thompson continued, before the internet era
incumbents, such as newspapers, book publishers, networks, taxi companies, and hoteliers, all of whom integrated backwards, lose value in favor of aggregators who aggregate modularized suppliers — which they often don’t pay for — to consumers/users with whom they have an exclusive relationship at scale.
But now, consumers aggregating the aggregators, previously candidate aggregators become modularized.
The rise of taste
As we wrote last week, we do not believe this translates into the same value capture as in the internet era. Apps are thick while protocols that deliver personalized AI outputs to applications are thin.
We’re excited to help support this shift back in power to tasteful upstarts or luxurious software who are creating unique supply and personalized experiences based on all the context, and who keep distribution not for their spend on expensive ads, but because consumers just love what they make.
We believe this offers tremendous opportunities for suppliers with taste, who under this new paradigm can offer the same or better user experience as the aggregators, and develop 1-1 relationships with customers who love what they create, without relationships or customer acquisition mediated by aggregators. To succeed in this new paradigm, suppliers should invest in creating great experiences augmented by complete user-provided context, e.g., creating web and mobile experiences that
- Invite users to log in with their own data
- Greet customers by name
- Recommend products or services based on user context
With more context than any individual aggregator, power returns to those with unique or opinionated supply, who can choose how and where to distribute.
With AI and user context commodified, it offers a chance to principally focus on shipping great user experiences and products that only your company can create.
Here’s to the rise of rebels and challengers and the emergence of an internet that’s All For You.
If you want to see what’s possible when you let users log in with all their context – check out our docs or start building.