Data Certification Quickstart¶

Use this page when you need a quick, defensible answer to one question: “is this dataset good enough to use?”

By the end, you should have a contract, a certification report, and a shareable audit dashboard.

Read before: Getting Started if you have not produced a run or dataset yet.

Read next: Data Certification Full Guide when you need the full protocol details.

Use a virtual environment

Run all commands from an active .venv.

What Certification Produces¶

contract validation results
compliance score
deterministic dataset and contract fingerprints
trust grade: draft, research_grade, or clinical_grade

Fastest Working Path¶

1. Create a contract¶

iints data certify-template --output-path data_contract.yaml

Edit the schema, units, and value ranges so they match your dataset.

2. Validate the dataset¶

iints data certify data_contract.yaml data/my_cgm.csv --output-json results/certification.json

For pipelines, use a strict gate:

iints data certify data_contract.yaml data/my_cgm.csv \
  --min-mdmp-grade research_grade \
  --fail-on-noncompliant \
  --output-json results/certification.json

3. Generate a dashboard¶

iints data certify-visualizer results/certification.json --output-html results/mdmp_dashboard.html

The dashboard is a single-file HTML artifact you can share offline.

Optional: Create Synthetic Mirror Data¶

iints data synthetic-mirror data/my_cgm.csv data_contract.yaml \
  --output-csv data/synthetic_mirror.csv \
  --output-json results/synthetic_mirror_report.json

Use this when you need schema-realistic development data without distributing raw sensitive rows.

Grade Meaning¶

Grade	Interpretation
`draft`	useful for iteration, not ready for rigorous research claims
`research_grade`	acceptable for research workflows
`clinical_grade`	strongest validation level currently exposed by the SDK

Optional Python Gate¶

import pandas as pd
from iints import mdmp_gate

@mdmp_gate("contracts/clinical_mdmp_contract.yaml", min_grade="research_grade")
def process(df: pd.DataFrame) -> int:
    return len(df)

This blocks or warns when incoming data does not meet the required quality level.

Where To Go Next¶

If you want to...	Continue with
understand every rule and fingerprint	Data Certification Full Guide
use certified data in a benchmark	Scientific Workflow
inspect the science behind realism claims	Evidence Base