peopleanalyst

use cases · R&D / Engineering leadership

The platform team is bleeding, and everyone's blaming pay

Your platform team is leaving faster than anyone — and you're sure it's pay. The data says otherwise.

For who

Engineering VPs and founders watching infra/platform attrition climb

What it finds

That the binding constraint is Support — recognition/visibility, not comp; leavers are at or above market.

What you get

A constraint you can act on before the comp reset — aim the budget at the condition that's actually binding.

Binding constraint

supportCapability and pay are fine; the binding constraint is Support — platform engineers have no visibility into who uses what they ship (the prosocial-impact / relatedness gap). The work is invisible relative to product teams. A comp reset burns budget and doesn't move attrition.

The situation

A ~200-person scaleup is losing platform/infra engineers faster than any other team. Exit interviews are vague ('better opportunity'). Leadership's working hypothesis is compensation — people are leaving for higher offers — and they're about to spend on retention bonuses and a comp-band reset.

How the walkthrough goes

  1. 01customer-situation

    Your platform team is leaving faster than anyone — and you're sure it's pay.

    200-person scaleup. Infra/platform attrition climbing. Exit interviews vague. A retention-bonus pool and a comp-band reset are already on the finance agenda.

  2. 02problem-cost

    You're about to spend big on a hunch.

    A comp reset is expensive and hard to walk back. If pay isn't the cause, the budget's gone and the team still leaves.

  3. 03insight

    Most teams blame comp. The data says the binding constraint is Support.

    Platform engineers have no line of sight to who uses what they ship — the recognition/relatedness gap that sustains the work. Pay isn't what separates the leavers from the stayers.

  4. 04desired-outcome

    Keep the team — without the wasted spend.

    Aim the budget at the condition that's actually binding (visibility/recognition), not the comp band.

  5. 05product-path

    Performix finds the constraint; AnyComp rules out pay.

    Protected feedback + CAMS names the single binding condition; the comp engine confirms your leavers are already at or above market.

  6. 06proof

    Conditions predict who leaves. Pay doesn't.

    In the data, the Support/visibility items separate leavers from stayers; comp percentile doesn't.

  7. 07risk-reversal

    Honest by construction.

    Protected feedback (anonymity as a primitive) + a hard minimum-group-size gate. No individual is exposed.

  8. 08next-step

    Run the diagnostic on the team that's actually leaving.

    One read, before the comp reset. If it is pay, you'll know; if it isn't, you just saved the budget.

Grounded in the research

Walkthrough data is composite and clearly labeled — shaped from the research to show the real shape of the finding, not a named client.

Reduce regretted platform-team attrition by fixing the binding condition (visibility/recognition) instead of comp — the decision-error avoided is a mis-targeted retention spend. Delta: share of attrition variance explained by conditions vs. by pay.