UXR, Market Research, and Data Science Walk Into a Reorg, or: Why Your Job Title Has About Three Years Left

Part 2 of a series on what research actually owns from here. Part 1 is here.

Part 1 ended on a claim I had not earned yet: that UXR, market research, and probably the user-facing side of data science stop being separate disciplines. I said I was more sure of this than of anything else in the series, which is an uncomfortable thing to be sure of before arguing it. Putting it in writing was a promissory note. This piece is the payment, and nothing else: no framework, no toolkit, just the argument that the functions merge, what is driving it, and what deciding it badly or well will look like.

Start with why the walls exist at all, because the answer is embarrassing once you say it out loud.

The Walls Are Made of Org Chart

Market research reads attitudes: brand, willingness to pay, the survey and the panel. The user-facing half of data science reads behavior, the logs and funnels and the instrumented everything. UXR reads experience. Three teams, three reporting lines, three vocabularies, and underneath them the same question asked three ways: what do people do, why do they do it, and what would change it. The same pricing question gets a willingness-to-pay survey from market research in March, a usability test on the pricing page from UXR in May, and a funnel analysis from data science in June, and the three findings never meet. Everyone reading this has watched some version of that happen and called it collaboration.

The walls came from history, not epistemology. UXR grew out of usability and HCI and got attached to product. Market research grew out of marketing and landed under the CMO. Data science grew out of instrumentation and got attached to engineering. Each one was born owning a data stream, the org chart calcified around the streams, and then the walls acquired budget lines, conference circuits, and professional identities. Nobody ever decided that understanding customers should be split three ways. It accreted, hire by hire, and now it looks like the natural order of things.

A Merger That Already Failed Once

The obvious rebuttal is that companies have tried gluing these teams together before, and they have. The 2010s produced a wave of insights departments, market research and analytics folded under one VP, sometimes with UXR bolted on after a reorg. Mostly it produced shared managers rather than shared understanding. The teams kept doing what they did under a new letterhead, and the collaboration mechanism was sharing readouts, which runs on the fantasy that the other team reads them. Nobody reads them.

Those mergers failed because they were reporting-line mergers. Each team still interpreted its own stream against its own memory, so the readings stayed local, and locality is the actual load-bearing wall. The walls survived their own demolition because the thing holding them up was never the org chart. No reorg touches the locality of interpretation.

The People Got There First

Two things changed since those mergers failed, and the first one is walking around your building already. AI collapsed the technical bar across the whole space, and the practitioners who lean technical walked straight over it. There is a type of researcher now, and every reader knows one or is one, who vibe codes their own tools, understands system architecture well enough to know what talks to what and why it breaks, and ships working things that would have needed an engineer two years ago. Not production code, but good-enough code, and good-enough turns out to cover tooling, prototypes, pipelines, and most of what a research team actually needs built. From there, light data science work is an afternoon away, not a degree away.

The backgrounds were always closer than the org chart admitted. A lot of UXR comes out of the math-heavy corners of social science, or computer science, or HCI programs that were half engineering to begin with, and those people already read R and Python. For them quant work was never a leap. It was a fence, the fence was mostly tooling friction, and the friction is exactly what the models dissolved. The same thing is happening one wall over in each direction: the technical end of market research is doing work that is heavy quant by any honest definition, and the user-facing end of data science was always half social scientist, hired to explain people while sitting in an engineering reporting line. All three fields run on the same internal spectrum, more technical to less, and the technical ends of all three have started doing each other's jobs without asking anyone's permission.

Part 1 watched this same force push the top of the funnel outward, to PMs and designers. It runs sideways too, and it is not a fixed state of affairs but a direction of travel: every capability the models absorb lowers the bar further, widens who can traverse, and makes the walls describe less about who actually does what. The uncomfortable part belongs in the open rather than the subtext. This advantage is not evenly distributed. The practitioners who lean technical are compounding it, the colleagues whose careers assumed the walls were real are not, and I am not going to pretend a framing exists that softens that.

Traversal alone is not a merger, though. Hybrid individuals scattered across three org charts mostly make the turf fights stranger. What turns people who can do all three jobs into one function doing one job is the second thing that changed.

Shared State Finishes It

Shared state is what touches the locality problem from the failed mergers. Run interpretation against one accumulated, dated, confidence-weighted account of what the organization knows about its users, and the boundary dissolves on contact, because the gate from part 1 does not care which department an input came from. A survey, a session recording, and a behavioral log arrive as inputs differing in coverage and confidence, not in tribe. Three separate gates over one organization's beliefs is incoherent, and one gate makes three interpretation teams redundant.

The intelligence world sorted this out decades ago, and it is worth taking the analogy seriously rather than decoratively. Collection disciplines stay proudly distinct: the signals people and the human-source people have different tradecraft, different training, and barely a shared language, and nobody proposes merging the crafts. Assessment is unified anyway, because an assessment split by collection method is not an assessment, it is a turf arrangement with footnotes. The craft stays plural. The reading does not.

And the two forces are not independent, because the practitioners doing the traversing are exactly the ones building the shared state. The merger is being assembled by the people it will absorb first.

The Prediction, With Its Hedges

So, as plainly as I can make it: UXR and market research merge. The user-facing half of data science probably goes with them, though maybe not all of it, and the hedge deserves a sentence because probably and maybe are doing real work. The half of data science that builds models and pipelines is engineering and stays engineering. The half that gets asked why the number moved, the funnel archaeologists and the behavior explainers, is doing interpretation of user evidence, and that half merges whether it wants to or not, because why did conversion drop and why do users abandon at the address step are the same question wearing different dashboards.

The standard objection is that the disciplines are too different, that survey methodology and ethnography and behavioral statistics are separate crafts with separate standards. They are. That is an argument for keeping the collection crafts distinct, which nobody disputes, and it says nothing about the assessment. The people who ran signals work were sure their craft could never sit under one roof with human-source work, and they were right about the craft and wrong about the assessment, and the same split applies here.

I have tried more than once to construct the version where the walls survive contact with shared state, and I cannot make it cohere. I notice that my certainty makes me suspicious of myself, and I have sat with the suspicion, and the certainty survives it.

A Capability Story or a Cost Story

None of this says the merger goes well. There is a version that arrives as a memo: three teams merged for cost, one cheaper department doing three kinds of collection under a new name, with the interpretation layer belonging to nobody, which in practice means belonging to whatever tool is closest. If you have lived through a consolidation, you have seen the memo version, and it looks exactly like the real one in the announcement.

The capability version is different at year three, not at week one. The crafts stay plural, the assessment is one, and the organization can finally answer what do we believe about our users in one voice, with receipts. That version depends on the merged function running on something the three separate teams never had, and what that something actually is deserves more than the paragraph I could fit here.

The consolidation is coming either way; I have stopped arguing with people about that part. Which version arrives is the open question, and it is not a small one, because the two look identical in the announcement and nothing alike three years later.

Two More Promissory Notes

I have now bet an entire merger argument on shared state without saying how the state gets built, kept honest, or run at speed. And part 1 already admitted the uncomfortable fact sitting next to that: a good enough tool wired into genuinely maintained state closes most of the reach gap. Both debts are deliberate. What the merged function runs on, and why the answer circulating on LinkedIn, become better thinkers, is not a plan, is part 3. The question underneath all of it, why the owner of this machine is a person rather than a supervisor of the system, comes later in the series, once the machine is assembled enough that the question has nowhere left to hide. It is not getting more comfortable as we go.

🎯 Part 3 lands next week. Subscribing to The Voice of User is how you catch it, along with essays on UXR, the organizational dysfunction underneath it, and the parts of the AI conversation the vendors are too invested to say out loud.

📖 The operating model underneath this series, the Frame and the machinery around it, is the book, AI-Powered UX Research. It is out this week.