ntwrk/
Scope the 30-day POC
Build to Keep
FIELD NOTES

The report got the architecture right. It got the owner wrong.

The new martech map is right about where everything is going. It is quiet on the one layer worth owning, and in Australia the disclosure behind it now has a date.

Written by Wes Fischer, founder and CEO, NTWRK. · .

One owned tile on the composable canvas A grid of eight neutral, interchangeable tiles forms the composable martech canvas, all assembled from the same shelf. One tile is set apart: the record of why, the single layer you have to own. THE COMPOSABLE CANVAS data identity audiences agents activation analytics governance the record of why OWNED, NOT RENTED
Everyone can assemble the same canvas from the same shelf. The one tile you have to own is the record of why.

In March, Scott Brinker published the clearest map I have read of where martech is going. The stack of boxes is dissolving into what he calls a composable canvas: one data foundation, with custom agents built on top as the real edge. Source: chiefmartec.com I do not have a quarrel with any of it. We have been building on that premise for years. The data stays where it lives, the intelligence sits next to it, and the advantage goes to whoever can shape it to their own business.

He is right, and the market has caught up to him fast. Databricks now calls the warehouse the CDP. Gartner's latest quadrant rewards warehouse-native and agentic. The CDP Institute's own numbers show the composable vendors growing several times faster than the rest of the field. The direction is no longer in question. So I am not going to spend a thousand words agreeing. I want to talk about the one thing the report names and then walks straight past.

Of the data classes Brinker lays out, the one that decides who actually owns the system is the record of why. He calls it the context graph: the decision traces, the approvals, the reasoning behind what the machine did. He treats it as architecture, a useful thing to capture on the way to real-time everything. It is not architecture. It is governance. The "what" lives in any system you care to name. The "why" is what a board, a regulator, or a customer asks for when a decision goes wrong, and it is the thing almost nobody is keeping on purpose.

In Australia, that gap now has a date on it. Brinker's report calls the context graph the keeper of the "why". Australian law is about to put those decisions on the record. From 10 December 2026, many businesses have to disclose, in their privacy policies, the kinds of decisions they let machines make, and the kinds of personal information behind them. The Privacy Act reforms tighten what consent even means. The Consumer Data Right keeps widening. The report describes a best practice. I am describing a legal requirement, with one distinction worth keeping sharp: the law asks for the disclosure, not for any particular architecture. The recommendation is mine, and it is blunt. The record behind that disclosure is easiest to stand behind when it lives in infrastructure you own. A document written for a global audience, and sponsored by a data platform, was never going to lead with that. It is not their deadline.

Here is where the architecture and the ownership part ways. You can assemble a composable canvas out of rented parts and it will run beautifully, right up until someone asks you to explain a decision. At that point the answer matters more than the speed, and the answer lives wherever the parts live. If the explanation sits in a supplier's account, you do not own it. You can request it. You can pay to keep querying it. You cannot put your name to it without trusting someone else's. A decision you cannot explain on your own terms is a decision you are renting, and renting your own audit trail is an odd place to be standing when the regulator is on the phone.

It runs beautifully until you have to explain it A composable stack of rented parts runs smoothly. When an explainability request arrives, the answer either sits in a supplier account, which you can only request, or in your own environment, which you own. RUNS FINE, UNTIL THE QUESTION rented audiences rented model rented activation ASSEMBLED, RUNNING "explain this decision" SUPPLIER ACCOUNT the answer lives here you can request it, not own it YOUR ENVIRONMENT the record of why is here yours to produce on demand A DECISION YOU CANNOT EXPLAIN IS ONE YOU ARE RENTING.
The stack runs fine until the question lands. Then it only matters whether the answer is yours to produce or someone else's to release.

It is tempting, at this point, to reduce ownership to infrastructure: our cloud means ours. The boundary is rarely that clean. You can host a system nobody in the building understands. You can also run on a managed service while holding strong access, portable records, and independent evaluation. Control is what you can do, not where the software sits. Can you retrieve the record? Can you follow the logic? Can you test it without the supplier in the room? Can you change supplier without losing the learning? Does the owner have a name? When I say own, that list is what I mean.

I want to be honest about what we do and do not do, because the report's end state is enormous. We are not the firm that builds your whole canvas. That is a multi-year, whole-of-business programme, and there are much bigger firms for it. What we do is narrower, and I think more useful right now. We build the pieces that have to be owned, and we hand them over with the keys. The consolidated data. The shared definitions, so "high-value customer" means one thing in marketing and the same thing in finance. The first agents that act on it. And the record of why, written in your environment, in your name, ready before anyone has to ask. You do not have to own the whole canvas this year. You can own the first piece this quarter.

Own the four pieces first The composable canvas with the rest of the tiles left neutral, and four pieces grouped and highlighted as the achievable first step: consolidated data, shared definitions, the first agents, and the record of why. NOT THE WHOLE CANVAS. THE FIRST PIECE. activation analytics audiences identity THE REST CAN WAIT consolidated data shared definitions first agents record of why OWN THESE FOUR THIS QUARTER
You do not have to own the whole canvas this year. The first piece is four pieces: consolidated data, shared definitions, the first agents, and the record of why.

So I read Brinker's report as good news, mostly. It is the best argument yet that the thing we have been quietly building is where the whole market is heading. But validation cuts both ways. When the architecture is something everyone can buy, from the same handful of platforms, assembled the same way, the architecture stops being the edge. Within a year, composable is table stakes. The edge is the part you cannot buy off a shelf and cannot rent back: the layer that has to explain itself, kept somewhere you control.

Everyone is about to own the same composable canvas. The only edge left is owning the layer that has to explain itself.

Wes Fischer, founder and CEO, NTWRK

FIELD NOTES

The notes keep pulling on the same thread: which layer is still worth owning once everyone can buy the same canvas.

Free. Unsubscribe in one click.

The first piece is the one you own.

NTWRK builds the foundational pieces that have to be owned, the record of why included, and hands them over with the keys.

See what you keep Scope the 30-day POC