What we validate against
Google Maps, Deutsche Post and more, not just format rules.
A frequent pre-sales question: "Is the validation just rule-based, or do you check against real address data?" The answer: both, in two stages.
Stage 1: rules (in-checkout)
The in-checkout validation is rule-based by design: format checks like "is a house number present", special characters, packing station formats. Rules are fast enough to run live in the checkout without slowing anything down.
Stage 2: reference databases (post-checkout)
After the order, the address components (street, house number, postal code, city, country; never names or emails) are verified against official reference data:
| Source | Coverage | Strength |
|---|---|---|
| Deutsche Post AddressFactory | Germany | The authoritative source for German deliverability, processed within Germany. Basis for DHL routability (Leitcodierung). |
| Google Maps Address Validation | Worldwide | Catches typos in street and city, wrong postal codes, and non-existent addresses. |
| Our own address database | All countries | Learned validations that speed up repeat lookups. |
The strongest available signal wins, and corrections only happen at high confidence. If the sources disagree or nothing matches unambiguously, the customer is asked instead. We never guess.
On the Pro plan, messy input for Germany, Austria and Switzerland is additionally restructured by an AI step before it reaches the databases. See AI-assisted address parsing (Pro).
Rural addresses in New Zealand, Australia and beyond
Rural addresses in New Zealand and Australia go through the same Google Maps Address Validation as any other address, there is no separate rural process. What differs is the map data itself: sparsely populated rural areas are mapped in less detail than cities, so a rural delivery (RD) number or a property-only address can be harder for Google to match with confidence than a suburban street address.
When Google's data does not confirm a rural address clearly, AddressHero does not fall back to a best guess. The order is flagged for review instead, so the customer confirms or corrects it rather than receiving a package addressed from a low-confidence guess. This is the same "never guess" rule described above, it just triggers more often in rural regions because the underlying map data is thinner there.
If your store ships a lot of rural NZ or Australian orders and sees more flags than expected, that reflects the map coverage for that region, not a misconfiguration.
What this catches
Spelling mistakes in street or city
Typos in the postal code
Missing or non-existent house numbers (existence where reference data allows it; whether a house number is CORRECT for that customer can never be verified, only whether it exists)
Mismatches between ZIP, city and country
Privacy
Only address components are sent to the validation providers, no names, no emails. Details in GDPR compliance.
Related articles
Was this helpful?