Understanding Address Confidence Levels
Applies to Address Verification API types that return a match/confidence score, currently Google Address Validation and Azure Maps. Providers that return their own deliverability codes (for example Melissa) are described in their own configuration articles.
What is a confidence level?
When an address is validated, the provider returns a numeric confidence score between 0.0 and 1.0 that reflects how certain it is that the corrected address is a real, deliverable location. To make results consistent and easy to act on, Address Validation for Dynamics 365 converts every provider score into a single, standardized set of confidence tiers. This means an automation or a user can reason about "how good is this result" the same way regardless of which API produced it.
Standardized confidence tiers
The numeric thresholds below are the shared source of truth used across all score-based providers:
| Tier | Score range | What it means |
|---|---|---|
| Very High | 0.90 – 1.00 | Exact, deliverable match. Safe to accept automatically. |
| High | 0.75 – 0.89 | Strong match with minor uncertainty. Generally reliable. |
| Medium | 0.50 – 0.74 | Partial match. Review recommended before relying on it. |
| Low | below 0.50 | Weak or no match. Treat as unverified. |
How each provider labels confidence
Each provider maps the shared tiers to its own label vocabulary. The numeric boundaries are identical; only the emitted text differs:
| Shared tier | Azure Maps label | Google label |
|---|---|---|
| Very High | VERY_HIGH | HIGH |
| High | HIGH | MEDIUM |
| Medium | MEDIUM | LOW |
| Low | LOW | VERY_LOW |
Why is Google shifted one tier lower? Google Address Validation tends to report higher raw scores than Azure Maps for the same quality of match. To keep the practical meaning of each label aligned across providers, Google results are mapped one tier more conservatively. This prevents Google from over-reporting "very high" confidence and keeps the labels comparable when you use more than one provider.
Where confidence appears
Confidence is surfaced in two places:
Single-record validation – the confidence label is returned with the validation result and can be written to a field on the record via your field mapping.
Bulk Address Validation (Preview) – the confidence label is included in the per-record results (and in the downloadable results CSV), so you can sort or filter a run by confidence.
Tips for using confidence
Auto-accept Very High and High results; route Medium to a review queue; flag Low for manual correction.
When comparing providers, rely on the shared tier meaning rather than the raw provider label.
In bulk runs, use the results CSV to triage large volumes by confidence before committing changes.
Note: Bulk Address Validation is currently a Preview feature.