Practical: Reconciling Your Entities in the Knowledge Graph Reconciler, Step by Step

This practical lesson turns the Reconciler walkthrough from Lesson 3.2 into a written procedure with screenshots of the real tool, so you can reconcile your own site at your own pace and finish with a Connection score.

What this lesson covers

• The six Reconciler steps, with what you should see at each
• Choosing a match-score floor from your own results
• Reading verdicts and type conflicts
• The reconciliation export and the Scorecard
• Troubleshooting 403s, all-unmatched runs and rate limits
• Handling ambiguous entities without forcing them

The procedure starts with the export from the Extractor practical, because its type and MID columns are what make conflict and over-linking detection possible. Each step shows the real tool on the demo corpus: the loaded file with its distinct-entity count, the floor and cap settings, the three verdict tiles, the table with Google’s preferred label and types, the conflicts filter and the export. A short “what you should see” check after each step tells you whether your run is behaving.

The lesson then shows how to set the floor from your own data by running once low and reading where genuine matches stop, and how to work the table from the weakest matches up, where the wrong ones hide. It ends with the exported file loaded into the Entity Audit Scorecard to score Connection, the brand-first scoring rule, and a short decision guide for ambiguous entities: drop them if they are too generic to matter, fix the wording if they matter, and leave them out of the score until then.

The troubleshooting section covers the three failures most people meet on their first run — a 403 because the Knowledge Graph Search API was never enabled, an all-unmatched result because the key is restricted to the Natural Language API only, and a 429 from looking up too quickly — with the fix for each and a reminder that completed lookups are kept.

This practical continues Module 3, Knowledge Graph Auditing & Entity Mapping, in The AI-powered Entity Audit, and leads into the unreconciled entity report. It suits learners running the Reconciler on their own or a client’s site for the first time.