Has your CEO ever asked you to use AI to anticipate customer needs? Has someone on your product or marketing team called out the difficulty of activating the data in your CDP to create experiences that help convert customers?
In 2018 before we founded Crosshatch we worked on a personalization team at Walmart. Our mission seemed simple: use AI to keep the American consumer’s refrigerator stocked, by using predictive AI.
Walmart is well positioned to actually deliver on deep personalization.
- There is a Walmart store within 10 miles of 90% of all Americans.
- Walmart is a consumer-staples business. Unlike discretionary businesses like travel or real estate, customers shop weekly to stay in stock of staple items they need.
- We had tons of consumer data.
Peers on our team were literally rocket scientists and ex-quants experienced in using data to predict behavior. We had access to every online grocery order: a pristine time series of customer ordering behavior. Transactions data constitute what economists call “revealed preference” — you best know what people want in the future based on what they have spent money on in the past.
We had great tailwinds: rich data, great infrastructure, and an incredible team.
What we found was that even with these tailwinds we couldn’t reliably predict grocery orders that helped families save time.
If rocket scientists and the world’s best data couldn’t readily anticipate customer needs with AI in a constrained case like grocery, how will the rest of the internet?
Introducing Headless Personalization
If you want to anticipate customer needs, you simply need more accurate data.
Most businesses don’t have the data resources of Walmart. Those that do are under pressure by lawmakers who are demanding transparency about what data companies collect and how it’s used. The systems that govern the customer data companies do have are often tightly coupled with the application of it, making personalization today an expensive and tightly coordinated exercise involving product, marketing and engineering.
This is why we’re introducing Headless Personalization. Headless personalization is a data interoperability model that allows you to use any language model to generate personalized text with a consumer's private permissioned context. It works by separating the application layer from the back-end data and governance. This allows customers to share (and un-share) their data to applications they love, while giving businesses the flexibility to transform and infer from it on demand. Apps permissioned to perform this headless personalization use inferences from connected consumer context and don't directly handle the underlying data that powers it. Since consumers have more control over how and what they share and for what purpose, data sharing becomes ... chill.
We contrast this with existing personalization solutions, where a business’ ability to personalize is given by whatever data they happen to have. Headless personalization separates these concerns, allowing consumers to bring their data with them and businesses to activate it into whatever context drives mutual value.
We’re betting big on this architecture because it unlocks creative freedom to build entirely new personalized experiences that are typically not possible with first party data alone. Developers are eager to go headless because it offers a unique level of developer and consumer control and gives them the freedom to leverage composable tech stacks without the overhead of building data integrations, tracking plans, and security regimes.
Going headless allows you to activate permissioned customer information in multiple front-end experiences for different customer touch points on any data a consumer is willing to share. Your web and mobile can talk to a single personalization system via the permissioned API layer, which allows emerging startups to be as personalized as customers want in a way they’ll trust.
Going headless unlocks true hyper-personalization. Hyper-personalization is only possible when services can respond to consumer needs exactly as they are or given by as much data as a consumer is willing to share. Headless is how that data are activated in a chill way.
Signs that Headless Personalization could benefit your business
Brands starting to explore Headless Personalization often see themselves in one of the following scenarios:
- My leadership team is creating a new AI strategy and I’m worried we may not have the data to support it.
- I already have an established CDP but its ROI is not what we had hoped for.
- We shipped a CDP but we’re worried about customer journeys degrading when cookies deprecate later this year.
- I want my customer experience to be tailored to each customer individually even when they log in for the first time.
- I want to build a unique customer experience but I’m worried I’m at a disadvantage to incumbents who have way more data than I have.
- We’d like to scale the data we have about each customer, but we’re cognizant of respecting privacy and costs of compliance.
Consider the Costs
As you decide if and how to go headless, consider costs and time, especially relative to other options. Enterprise CDPs tend to cost in the tens to hundreds of thousands to millions. Some come with enriched data and identity resolution services but are not prepared for cookie deprecation happening this year that could threaten the efficacy of those solutions. Implementing a rich CDP can be an expensive and frustrating exercise.
Going Headless with Crosshatch
Crosshatch is supporting businesses looking to make their services more personalized without heavy engineering lift. Developers and marketers leverage Crosshatch’s suite of headless personalization services to build experiences that speak to each user individually, built in less time and lower cost. Crosshatch’s headless solution allows businesses to add inferences from user context wherever it’s needed and with whatever AI tools they’re already using.
Crosshatch’s headless personalization services include
- Proxy API: add consumer context to any LM we’ve secured
- Query API: test what context you can add
Let’s take a look at how each enables businesses of all sizes to move toward a headless architecture.
Proxy API
Proxy API is the foundation of our headless hyper-personalization platform. It provides access to the full breadth of AI resources and unified representation of data consumers are willing to share. This enables capabilities critical to driving retention, engagement and conversions like
- Optimized search + recommendations
- Tailored interfaces
- Details that put products and services in context
The Proxy API is agnostic to LM inference provider, orchestration stack and evaluation platform. This gives developers the freedom to use the tools they already use and love, while simultaneously experimenting with new hyper-personalized technologies that add consumer context wherever it’s needed.
“Ultimately, the power of generative AI or any technology is only as good as the data that powers it”
Walmart’s Doug McMillon said last fall.
Query API
Crosshatch’s development stack also consists of the Query API, which gives developers a clear path for developing hyper-personalized applications with rich consumer context.
The Query API exposes the query interface for use on test data, allowing developers to see how queries resolve as stringified context in the Proxy API.
Ready to go headless and hyper-personalize?
Whether you’re an established business with existing architecture and user profiles, still building out enterprise architecture, or a new business looking to deliver new services with rich consumer context, a headless approach to hyper-personalization could be right for you.
Similarly, if your product or marketing operations over your customer data are becoming complex or unstable and you want to compete on customer experience, headless might be in your future.
Unlock total creative control with headless hyper-personalization. Learn how.
