Datasets

What to send the QC Analytical Results API: precision (duplicate pairs) and accuracy (CRMs and blanks) datasets, their columns, the formats, and the limits.

A dataset is your QC results, sent once and asked about many times. There are two kinds, matching the app’s two sheets:

Kind	What it holds	Charts
`precision`	Duplicate pairs: Sample 1 and Sample 2 of each pair, one row per analyte per pair	The eleven precision charts
`accuracy`	CRM and blank results, one row per analyte per result	The seven CRM and blank charts

A chart asked of the wrong kind answers 422 wrong_dataset_kind.

Columns

Column names are matched ignoring case and spaces around them. Extra columns are ignored. These are the ones the charts use, as the app’s export template names them.

Precision (duplicate pairs)

Column	What	Required
`LABCODE`	Laboratory code, e.g. `ALS`	yes
`DESPATCHNO`	Despatch (batch) number	yes
`LABJOBNO`	Laboratory job number	yes
`ASSAYNAME`	Assay name: element, method and unit, e.g. `Au_AA25_ppm` (or `ELEMENT`, `METHOD` and `UNIT` columns, which the API joins the same way)	yes
`CHECKSTAGE_CK`	The check sample’s stage, e.g. `Pulverising Stage`, `Crushing Stage`, `Field Prep Stage`	yes
`SENDDATE`	Despatch send date	yes
`RETURNDATE`	Laboratory return date	yes
`ASSAYVALUE_OR`	Sample 1’s value (the first of the pair)	yes
`ASSAYVALUE_CK`	Sample 2’s value (the second of the pair)	yes
`LOWERDETECTION`	The method’s lower detection limit for this analyte	yes
`ORIG_SAMPLE`	Sample 1’s id	yes
`CHECK_SAMPLE`	Sample 2’s id	yes
`UPPERDETECTION`	Upper detection limit (for `apply_upper_detection`)	no
`ANALYSISSUITE_CK`	Analysis suite (for the `suite` parameter)	no

Accuracy (CRMs and blanks)

Column	What	Required
`LABCODE`	Laboratory code	yes
`DESPATCHNO`	Despatch (batch) number	yes
`LABJOBNO`	Laboratory job number	yes
`ASSAYNAME`	Assay name, as above	yes
`STANDARDID`	The CRM or blank’s id, e.g. `ALS-26`	yes
`STANDARDTYPE`	Its type, e.g. `COMPANY_STD`, `LAB_STD`, `COMPANY_BLANK`; the blanks chart takes a blank type	yes
`CHECKID`	The result’s own id (one per CRM result): the default x axis	yes
`SENDDATE`	Despatch send date	yes
`RETURNDATE`	Laboratory return date	yes
`ASSAYVALUE`	The result	yes
`LOWERDETECTION`	Lower detection limit	yes
`STANDARDVALUE`	The certified value (for `use_certified_values`); may be blank	yes (the column)
`STANDARDDEVIATION`	The certified standard deviation; may be blank	yes (the column)
`UPPERDETECTION`	Upper detection limit	no
`ANALYSISSUITE`	Analysis suite	no

Values

Send every value as text, as it is in your database. <0.01, -0.01 and the like are read as below detection, and below-detection results are never plotted or counted, except on the blanks chart, where they are kept at half the detection limit (see How this is calculated).
Dates in any usual form. Day-first and month-first are decided once per dataset from the dates themselves, so keep to one form.
NULL (as some database exports write it) is read as blank.

Formats

JSON: {"kind": "precision", "rows": [{...}, {...}]}, one object per row, Content-Type: application/json.

CSV: the file as it is, Content-Type: text/csv, with ?kind=precision or ?kind=accuracy on the address. The first line is the column names.

Either may be gzipped: add Content-Encoding: gzip. A 50 MB CSV is a few MB gzipped and goes in a second or two.

What comes back

{
  "dataset_id": "ds_bbd34394c3dd10abfeebe84d",
  "kind": "accuracy",
  "rows": 39919,
  "created_at": "2026-10-22T06:10:00Z",
  "expires_at": "2026-10-22T07:10:00Z",
  "facets": {
    "labs": ["ALS"],
    "analytes": ["Ag_MEICP41s_ppm", "..."],
    "suites": ["Expl-Core", "..."],
    "crm_types": ["COMPANY_BLANK", "COMPANY_STD", "LAB_STD"],
    "crms": ["A0029", "ALS-26", "G398-4c", "..."],
    "despatches": 174,
    "dates": {"return": {"from": "2022-11-16", "to": "2025-11-12"}, "send": {"from": "2022-10-29", "to": "2025-10-19"}}
  }
}

rows is the rows kept after tidying. The facets are the exact values the charts take: labs and analytes for every chart; stages (precision) for stage; crm_types and crms (accuracy) for crm_type and crms; suites for suite. dates is the span of each date column, for date_from and date_to.

GET /v1/datasets/{id} gives the same again, and DELETE /v1/datasets/{id} drops it now.

Limits

Rows in one dataset	500,000
Rows across all your datasets at once	1,000,000 (the least recently used dataset is dropped to make room)
One request	50 MB
A dataset is kept	one hour after its last use (each call with it starts the hour again)
The same file again	the same dataset: the id you already have
Calls	300 a minute per key

A dataset that has gone answers 410 dataset_expired: send it again. A client that sends its data at the start of each run, and again on a 410, needs nothing else.

Privacy

A dataset is held in memory on the API server that received it (in London or Frankfurt), only for the hour, and is never written to disk. If that server restarts, the dataset is gone and your next call about it gets 410 dataset_expired: send it again. The log of each call holds its request id, your key’s name, what was asked (chart and parameters), its status and how long it took: never the data, never the key. See the Privacy notice.