How CAI works

Methodology

Clinical Asset Appraisal is recorded asking-price intelligence for used clinical equipment — not an appraisal, not medical or regulatory advice, and not sold-price data. Every number on this site is derived from real, attributable listings across 9 active sources. This page explains exactly what we collect, how we classify it, and how we compute the peer-group medians shown on every device page.

Last updated: August 3, 2026 · Tracked devices: 191 · With public market data: 182

1. Asking prices AND realized sold prices — kept separate

We track two distinct things and never blend them into one number. The Library Median is the median of publicly listed asking prices (dealer marketplaces, surplus portals) — a useful ceiling and starting point, not an appraisal, and not a price any seller is bound to accept. Separately, we surface realized sold prices where we have them: auction closes(GovDeals / HiBid final winning bids), eBay Terapeak 90-day sold data, andcommunity-reported sales (Report-a-Sale). Each number is labeled with its basis and how many comps back it. Asking medians typically run well above realized prices — that gap is exactly why we show both: the asking ceiling, and what units actually clear at (auction closes are a liquidation floor, not a dealer price). A discount is always computed against one anchor and labeled as that anchor.

2. Peer-group classification

Every listing is classified into one of these mutually exclusive buckets before any pricing math happens:

Complete System — fully assembled, ready-to-deploy unit.
Major Module — significant operational sub-system (e.g. a probe, a detector, a vaporizer).
Part / Accessory — component, cable, battery, board, accessory.
For Parts / Repair — non-functional, salvage, "as-is" listings.
Consumable — single-use accessories.
Service Manual / Software — non-physical assets.
Under Review — listing whose classification confidence is too low to publish.

We never mix these buckets in a peer median. A complete-system median is computed only from complete-system listings. Mixing them is the single biggest reason third-party used-equipment price estimates fail.

3. Why median, not mean

Used clinical pricing is full of outliers. A single $82,400 DOTmed dealer asking price next to a $10 GovDeals lot for the same model destroys the mean. The mediansits at the middle of the price distribution and ignores both tails — that's exactly what you want for "what does this thing actually sell for". We show p25, median, p75, min, and max publicly. The mean is shown only as a small secondary number with a tooltip explaining the trade-off.

4. Confidence thresholds

A listing is published in peer-group comps only when classification_confidence ≥ 0.65.
A listing appears in the Deal Feed only when both device_match_confidence ≥ 0.75 AND classification_confidence ≥ 0.65.
Active Listings on a device page require classification_confidence ≥ 0.75 — weaker matches go to admin review, not public display.
Per-bucket comp groups with fewer than 5 listings are labeled Thin and treated as directional only.

5. Source coverage grades

Each device gets a coverage grade based on its observed listings and source diversity:

STRONG — 50+ listings observed and 3+ sources contributing.
GOOD — 20+ listings, 2+ sources.
THIN — 5+ listings.
VERY THIN — fewer than 5 listings.
NO DATA — nothing observed yet (an expansion target).

The current source list (active vs planned) is visible on /stats.

6. Stable deals (how we know a deal has held)

We snapshot the firm below-typical deals once a day into a small history table. A device earns a ★ Stable deal badge when below-typical firm deals show up for it on a sufficient number of those daily snapshots within the last 7-day window.

The window adapts honestly: we only count snapshot days that actually exist. As history accrues, the window matures to the full ≥ 5 of the last 7 days criterion.
The badge reflects observed listing persistence — it does not mean any specific listing is still available. Sellers may relist, end early, or change price; every linked listing should be verified at the source.
If a device has no demonstrated persistence in the current window, it gets no badge. The signal's whole value is that it's true.
The history table is device-level, so the badge is per-device, not per-listing.

7. Known limitations

We don't see final sold prices — observed listing prices only.
Some sources require login or proprietary APIs — those are marked as "planned" on /stats.
Auction prices skew lower than dealer prices and are not directly comparable. We surface both, separately, in the per-source table on every device page.
Recall and adverse-event data are matched at the device family level unless we have an exact manufacturer + model match. Don't treat "X matched recalls" as a regulatory determination.
Verification (URL or screenshot) is decision support — it doesn't replace physically inspecting the unit, requesting service records, or confirming regulatory status.

8. Sources & regulatory data

Listings are aggregated every 24h from eBay · GovDeals · DOTmed · HiBid · LabX · PublicSurplus, classified, scored against peer-group medians, and source-linked. Coverage is FDA-aware: openFDA recalls + MAUDE adverse events are checked per device (matched at the device-family level — see limitations).

9. Legal posture

The Equipment Library is recorded market data, not an appraisal, not medical or regulatory advice, and independent of — not endorsed, sponsored, or authorized by — any marketplace or manufacturer. We earn no commission, referral fee, or other consideration from any equipment transaction. We are paid to value equipment, never to trade it. Buyers are responsible for verifying condition, service history, regulatory status, and seller claims independently. We respect robots.txt and the terms of service of each source; we cache responses to avoid wasteful re-fetching. We do not collect, store, or process PHI/PII through any user-facing tool. See the Corrections page to report inaccurate data.