Skip to main content
Beyond the CDP: How the Agentic Context Platform (ACP) Puts Your Customer Data to Work
BlogBeyond the CDP: How the Agentic Context Platform (ACP) Puts Your Customer Data to Work

Rally is a Sr Product Marketing Manager at Postscript and one of the few Canadians on the team. She has been an early employee in multiple tech startups and joined the ecommerce industry in 2017. On the side, she has also worked as a SaaS startup consultant.

Access premium content. Become an SMS expert.

Join over 10,000 other marketers who get weekly insights delivered directly to their inbox.

We are big fans of privacy. See our Privacy Policy for more information.

Product Updates, AI

Beyond the CDP: How the Agentic Context Platform (ACP) Puts Your Customer Data to Work

When Bar Bruhis founded Boostcous, the first 500 orders shipped out of his garage. He knew every customer by name, why they bought, and exactly what they said in their review. Then orders climbed into the thousands, and that knowledge stopped fitting in one person's head.

Most retention teams are living some version of this. You know a lot about your customers: what they bought, what they browsed, what they wrote in a review or said in a text. Very little of it reaches the message they actually get.

The Agentic Context Platform (ACP) is built to fix that. Here's how it differs from the CDP you might already have, and three standout features that show what it does: Unified Profiles, Memories, and Multi-model Attribution. Together, these core features address some of the most persistent and expensive problems faced by ecommerce merchants.

A CDP Stores Your Data. The ACP Puts It to Work.

A customer data platform is a filing cabinet. It holds what happened and waits for you to do something with it, usually to power dashboards and reports. The work of turning that data into better segments and personalized messages is still yours.

The ACP doesn’t just hold your data, it decides what to do next with it. It ingests and reads three kinds of context:

  • Customer context: what each customer bought, browsed, and told you, across every channel, tied to one profile.

  • Brand context: your voice, offers, catalog, and goals.

  • Performance data: results from the experiments running across the AI products in your account, so decisions are grounded in what's worked, not guesses.

It uses structured data, like purchases, and unstructured data, like a post-purchase survey answer or a text conversation. Then every channel and AI agent acts on it, and every send feeds back in so the next one is smarter.

Oh—and did we mention? The ACP is included in every plan on Postscript 2.0. Not a paid add-on like most CDPs.

Unified Profiles: One Customer, One Profile, Every Channel

Take Debbie, a Boostcous customer who ordered twice: once from a work email, once from a personal one. Nothing about the two addresses says they're the same person. On most platforms, that's two customers, each holding half the story.

In Postscript, those profiles merge automatically in the background. SMS, RCS, and email all read from the same record, so you're looking at all of Debbie, not two profiles with half the information each.

That matters for two reasons:

  • You pay for the customer, not the copies. On profile-based pricing, every duplicate shows up on the bill. As Bar put it: "You're basically getting punished for growth."

  • Your first-order discount gets used once. When a shopper comes back with a second email for another welcome offer, the ACP already knows it's the same person.

One profile per customer is also the foundation for everything else. You can't personalize for someone you only half know, and you can’t attribute your revenue correctly either.

[embed wistia video: Memories]

Memories: Context That Knows What Matters Now

Customers tell you things about their shopping preferences all the time. In a text conversation, one Boostcous customer mentioned her daughter-in-law has celiac, so she's always looking for better gluten-free options. Bar's take: "Customers want to talk to us. They're volunteering all this zero-party data. Before, it was just going into a black hole."

Memories pull the important details out of conversations, reviews, and other sources and save them to the customer's profile. Then they weigh how long each one stays true:

  • Durable facts stick. "Daughter-in-law has celiac" shapes messages for as long as it's relevant.

  • Passing ones fade. "Recently married" is nice to know, but it won't drive your messaging a year from now.

You could never build a segment for "shopping for a celiac daughter-in-law." With Memories, you don't have to. The next back-in-stock message can ask how her daughter-in-law is doing, not just announce the product.

Lalo saw the same thing. Memories surfaced a customer expecting twins, something the team would otherwise only guess at after a double high chair order.

[embed wistia video: Multi-model attribution]

Multi-model Attribution: Revenue Counted Five Ways

Most retention platforms only show last-touch attribution. But there isn’t a one-size-fits-all attribution model: your preferred model depends on many internal choices and stakeholders. So what merchants really want is the flexibility to choose their attribution model, or to switch between them as needed. And we believe that this flexibility should come standard with all plans, not gated behind an advanced analytics plan.

With any Postscript plan, you can switch between five models and compare them side by side:

  • Last touch: full credit to the final interaction before purchase.

  • First touch: full credit to the first interaction before purchase

  • Linear: credit split evenly across every touchpoint.

  • Position-based: most credit to the first and last touches.

  • Time decay: more credit to touchpoints closer to the purchase.

Set your own lookback windows, see how each channel gets credited under each model, and drill into any order. A customer who came in from a Meta ad and clicked an email four minutes before buying shows exactly that, step by step.

Bar's take: "As a business owner I'm fighting for margin, and now I actually know how much revenue email and SMS are driving, so I can scale it accordingly."

Every Customer, Known as One Person

Unified Profiles tell you who a customer is. Memories tell you what they care about right now. Multi-model Attribution tells you what's actually driving revenue. The ACP ties them together, so every send across SMS, RCS, and email is decided against the full picture of each customer.

Or, in Bar's words: "We still have our five segments. But now every customer is a segment of one."

See ACP in action: Watch the Postscript 2.0 launch event where we demo the ACP