LLenny's Podcast
← All frameworks
LeadershipJudd Antin (Airbnb, Meta)

The Embedded Insights Partner Model

Success metric: they won't have the meeting without you.

Difficulty
Advanced
Time to result
~months to results
Steps
5
Confidence
94%

Antin's structural cure for the vicious cycle that gets researchers laid off. Instead of research as a reactive service function called in at the end, embed a researcher in the product process from beginning to end, give them the same OKRs as the team (never separate research OKRs), and measure success by whether the team can make a decision without them. It also supplies a counterintuitive staffing rule: don't spread researchers thin — concentrate them, create visible pain for the teams that lack one, and let those teams argue for headcount on your behalf.

Origin

Judd Antin, from his time as head of research (twice), head of design for guest products, and head of the design studio at Airbnb, plus building the research practice at Facebook.

Core principles

  • 01As long as research is a service discipline, it will stay stuck.
  • 02Research teams should never have their own separate OKRs — one set of success metrics for PM, engineering, design, and research.
  • 03The real success metric is influence: they cannot hold the decision meeting without the researcher.
  • 04Staff by relationship, not by product-space coverage.
  • 05Headcount grows from the pain of loss, not from abstract arguments about research's importance.
  • 06Silos between insights disciplines are stupid — unify them into one insights machine.

How to run it

  1. 1

    Break the vicious cycle at the engagement point

    Diagnose the loop: reactive staffing → no input on the questions → low-impact work → executives conclude research isn't impactful → sidelining and layoffs. The short circuit is constant engagement: put the researcher in the room from project kickoff.

    Pro tip A researcher present at framing can shape the one question that yields maximum business impact, then answer it fast and get out.

  2. 2

    Give the researcher the team's OKRs, not their own

    PMs, engineers, designers, and the researcher share one set of success metrics. Either you're doing it together or you're not.

    Watch out Separate research OKRs institutionalise the service relationship and let the team disown the insight.

  3. 3

    Measure influence, not output

    The metric of success is that the decision-making meeting cannot happen without the researcher present. That indicates trust, relationships, and active participation rather than input-into-someone-else's-process.

    Pro tip Antin applies the same test to PMs: you want to be the PM people want on their team, not the one they route around.

    Watch out Output metrics (studies shipped, reports written) actively reward middle-range busywork.

  4. 4

    Concentrate researchers and let the pain do your recruiting

    Rather than stretching one researcher across a whole product space, pair them tightly with the partners who will actually involve them, protect their time, and let their impact show. Neighbouring teams will see it and want it.

    Pro tip Antin's headcount strategy: never argue abstractly for research; get the partners who wish they had a researcher to make the argument for you.

    Watch out This deliberately creates pain for uncovered teams. That is the mechanism, not a side-effect — but it needs leadership air cover.

  5. 5

    De-silo insights into one machine

    UX research, market/competitive research, data science, customer-service feedback, and NPS/voice-of-customer programs should feed a single integrated insights function rather than throwing findings 'over the transom' at product teams from every direction.

    Watch out Product teams drowning in uncoordinated insight streams will discount all of them equally.

In the wild

The integrated insights function at Airbnb

On joining, Antin's listening tour found product people overwhelmed by insights arriving from data scientists, product specialists, customer service, and voice-of-customer programs with no way to reconcile them. He built an integrated insights function that folded in UX research, market/competitor work, smart data-science integration, and the NPS program.

One insights machine feeding product decisions instead of a stream of disconnected findings coming over the side.

Headcount via the pain of loss

Instead of pitching leadership on why research matters, Antin paired researchers with the partners most willing to engage them, protected their time, and let neighbouring teams watch the impact.

Partners who wished they had a researcher made the headcount argument for him.

Common mistakes

Hiring researchers before you know how to integrate them

The ZIRP era hired researchers into reactive service roles with great intentions and no integration model — setting them up to fail, then concluding research was low-value when it did.

Spreading one researcher across everything

Thin coverage guarantees shallow relationships, which guarantees reactive middle-range work, which restarts the vicious cycle. Concentration plus visible pain beats coverage.

Is it for you?

Best for

Design/research leaders and heads of product restructuring how insights work gets commissioned and staffed

Not ideal for

Companies with one researcher and a genuine need for broad triage coverage across many products with no leadership appetite for uneven staffing

From the transcript

I never asked for research teams to have their own separate okrs

31:30

my metric for success is when they won't that have that meeting without you that's my metric for Success

32:00

I would much rather create pain in that situation than spread someone too thinly

57:30

as long as research is a service discipline I think we're going to be stuck in this spot

15:30

creating that integrated lean process where a researcher is arm inarm from the beginning is the most important advice I'd give

51:00

From the episode

The UX research reckoning is here

Judd Antin (Airbnb, Meta)