Connect Oura or Whoop. Answer ten questions. We rank the AU catalog by fit, never commission.
14 nights, parsed and discarded.
Confidence shown. Low-data audits say so.
32 SKUs across 5 AU brands.
↑ specimen rendered from a real engine response · audit 17c79e54
Not a quiz. An audit — triangulated against the catalog.
Oura or Whoop, 14 nights. Or enter manually. Or skip.
Position, build, budget, feel, temperature, pain points, materials.
Inputs → a spec vector: firmness, cooling, support, pressure, motion — with findings.
Prescription + confidence + 3–5 ranked matches, with reasoning.
We ask the nervous system, not your preferences. Read once, never stored.
| signal | source | your reading | weight |
|---|---|---|---|
| HRV (avg, 14n) | oura · whoop · manual | 47.2 ms | ×1 |
| Resting heart rate | oura · whoop · manual | 58 bpm | ×1 |
| Skin-temp °C | oura · whoop | +0.4 °C | ×3 |
| Sleep efficiency | oura · whoop · manual | 91 % | ×1 |
| Deep / REM share | oura · whoop | 18 / 22 % | ×1 |
| Latency · WASO | oura · whoop · manual | 12 / 25 min | ×2 |
| Quiz answers | 10 questions | 10 / 10 | ×4 |
weights are categorical priors · the engine adapts them per audit · see methodology
"We measured this. Here's what we found. Most mattress quizzes can't say either thing."
Tagged on 8 axes. Same set every audit. No pay-to-play.
scores reflect aggregate-fit across all 14-night audits in the last 30 days · indicative only
14 nights once. The contract:
One read on connect: HRV, RHR, skin temp, stages.
Read, then discarded. No refresh.
Aggregate audit only. No name, no account.
Quizzes treat preferences as conclusions. We read your wearable and treat answers as priors. Matching is mechanical — you see the same answer we saw.
Enter numbers manually, or skip. Confidence drops, and we say so.
Fixed-rate affiliate commission from some brands. Ranking is decoupled and disclosed.
The specimen above is a real engine response, redacted.
Five minutes. No account. Email optional.