What Composa won't tell you
Known limits and failure modes, published on purpose · Last updated 2026-07-10 · Ousios LLC
What the assessment can't see
- It's 12 questions and ~90 seconds. That buys a useful signal, not a portrait. Context — your current role, a rough quarter, a team you don't trust yet — shapes how you answer, and the instrument can't tell the difference.
- It's self-report. You're describing how you work; people are imperfect narrators of themselves. The team debrief exists partly to let colleagues sand down each other's self-portraits.
- The archetype layer is directional, not definitive. The family layer is our reliable signal (~84% recovery in internal analysis); the specific archetype within it is a best estimate — which is why near-ties are shown as near-ties. Full detail on the methodology page.
- Results go stale. A result from a different role or a different season describes a different person. Retake on real change (the Manager's Guide covers cadence).
What the ROI number can't know
- It cannot measure your team. It's a model: time estimates × team size × salary, discounted by a 40% capture factor and per-workflow adoption rates (the full math). The two things that actually decide your outcome — implementation quality and real adoption — are exactly the things a calculator can't know.
- It inherits its assumptions' errors. The per-workflow hour estimates are reconciled against worked scenarios, not measured field data. Until our design-partner results are published, treat every dollar figure as a structured starting point for planning — never as a forecast.
Where the workflows themselves fail
Every one of the 44 recommended workflows ships with its own "pressure test" — the specific way it goes wrong. A few, verbatim in spirit, so you can calibrate the honesty level:
An ADR generated without your actual rejected-alternatives reasoning reads like a press release. The workflow refuses to skip that input — because without it, the output is worse than nothing.
A clean auto-generated report makes a bad week look like a good one. The value is the human exception-review on top — rubber-stamping the draft defeats the workflow.
Under about six people, anonymized pulse surveys are identifiable in practice. Running one anyway erodes exactly the trust it's meant to build. We say this in the product, in the workflow caveat, and here.
It's an internal screening that prioritizes what a qualified human should look at. Treating it as definitive creates false security — high-severity gaps still need a real reviewer.
Some people thrive at volumes that break others. Capacity dashboards flag anomalies; only a conversation confirms what they mean — and monitoring must always run with the team's knowledge.
What we can't prove yet
- No published outcome evidence — yet. We don't have public case studies with measured before/after data. The design-partner program is how we're fixing that, and the evidence page — currently an honest zero — is where results will live, published including what didn't work.
- The validation study is in progress, not done. The instrument's external validation is running now. Results get linked from the methodology page when they're in.
When you shouldn't buy Composa
- Your team is in acute conflict. A working-styles vocabulary won't fix broken trust, and archetype labels can become ammunition. Address the conflict first.
- You want insight into individuals' performance. Wrong tool — and against our terms. Composa never feeds hiring, firing, promotion, or reviews.
- You need a certified psychometric instrument. We're honest about what's validated and what's directional, but if your use case requires a formally certified assessment, this isn't that.
- Nobody will own the follow-through. The value shows up in the workflows your team actually runs at day 30, not in the results screen. No owner, no value — a report can't want things for you.