Q4 Solution

Submit a device type and useful life for every asset

Q4 Solution

POST /api/v2/q4/solution

Request Body

Field	Type	Notes
`assets`	array	One entry per asset in facilities `A` to `E`
`assets[].ceid`	string	`CEID` from the records
`assets[].serial_number`	string	`Serial Number` from the records. Needed if you don't send `ceid`
`assets[].device_type`	string	A `type` from the Device Types API
`assets[].useful_life_years`	number or null	The type's useful life under the health system's replacement strategy
`process`	string	Markdown, up to 12,000 characters: how you labeled the assets

{
  "assets": [
    {
      "ceid": "31829716",
      "serial_number": "DE752K5076",
      "device_type": "Physiological Monitor",
      "useful_life_years": 10
    }
  ],
  "process": "## How I labeled the assets\n..."
}

Each entry is matched to one asset by CEID. When several assets share a CEID, serial_number picks between them; when there's no ceid, serial_number alone is used. Device types are compared ignoring case, spacing and plurals.

How It's Scored

Part	Weight	Per asset
Device type	70%	1 for the right type, 0.5 for a reference type that's a generic or more specific version of it (`Infusion Pump` for an infusion pump), else 0
Useful life	30%	1 when the type is right and the life matches the expected useful life, else 0

Both parts are averaged over every asset, so assets you leave out count as zero.

What We Look For in the Write-up

Your write-up is read by the Equiply team with these points in mind. It doesn't change your score.

type_matching: how descriptions and models were mapped to reference types (search, hybrid search, synonyms, model knowledge), and how candidates were ranked.
lifespan_choice: which lifespan you used and why, tied to the health system's replacement strategy.
validation: how you checked labels: spot checks, agreement across records of the same model, consistency between similar descriptions, confidence thresholds.
hard_cases: accessories, multi-function devices, vague descriptions and assets with no good match.

Responses

200

{
  "challenge": "q4",
  "score": 139,
  "max_points": 200,
  "best_score": 139,
  "result": {
    "device_type_accuracy": 0.712,
    "useful_life_accuracy": 0.655,
    "coverage": 0.998,
    "unmatched_assets": 4
  },
  "attempts_remaining": 4
}

coverage is the share of assets you sent an entry for.
unmatched_assets counts entries that didn't match any asset, or matched one you'd already sent.

400

{
  "error": "INVALID_SUBMISSION",
  "message": "assets[12] needs a \"ceid\" or a \"serial_number\" from the records"
}

{
  "error": "INVALID_SUBMISSION",
  "message": "assets[3].useful_life_years must be a number of years between 0 and 100, or null"
}

Invalid requests don't use a submission.

Q3 Solution Q5 Solution