Mohammed Teto · Authority Explorer
← All tutorials

Review an executive's identity before changing a profile

Business question

An international leader appears under different names and roles across a biography, a French profile, a directory and an event page. Which differences need correction, and which are approved alternatives? This workflow supports the identity foundation of an authority review.

Teaching scenario

Entirely fictional: Alex Rowan is Cedar Notes' Research Director. Four invented profile observations are supplied. The owner-approved reference explicitly permits the French role “Directeur de recherche” and the organization variant “Cedar Notes Publishing.” These approvals are inputs, not discoveries made by the tool.

The directory observation uses “A. Rowan” and “Chief Scientist.” The event observation has a name but no role, organization or description. None of the placeholder example-domain URLs was researched as evidence.

Run the exact example

From examples/fictional-workflow/ in a fresh copy of this repository:

python3 tools/entity-consistency-auditor/audit.py --identity inputs/identity.json --profiles inputs/profiles.csv --as-of 2026-11-01 --output-dir reports/identity

Open reports/identity/report.html. The same run writes JSON and CSV outputs. The CLI runs offline on the supplied inputs.

Read the actual output

These are field counts across four profiles, not counts of people, an accuracy percentage or an authority score. Case and whitespace normalization does not automatically approve translations, abbreviations or legal-name changes.

Make the review decision

Adapt across markets

The sample aliases apply to all observations in that run. When a translated role is approved only for one market, use separate approved records and runs. Preserve original scripts, accents, organization suffixes and dates unless an authorized alternative is documented. An exact match to an incorrect reference is still not truth verification.

Next step

Use the organization-evidence tutorial to connect material claims to actual reviewed sources. Define the observation protocol before asking whether any correction changed search or AI representation. No ranking or citation result is measured here.

Interactive identity view · Original tool and releases · Full sample inputs