Verifying addresses where streets have no numbers
Where streets have no numbers, stop matching address strings. Capture a location (a map pin with the applicant's own directions and details), then corroborate it at the level of the area against independent records, and grade how strongly it is supported.
Charles Archibong, Co-founder
· 5 min read
Key takeaways
- Capture a location, not a string: a map pin plus the applicant's own directions and details.
- Compare at the level of the area; street-level string matching turns formatting into false mismatches.
- Never require a house number from someone who does not have one.
- Grade the evidence: a pin the applicant stood at is stronger than one a record agrees with, which is stronger than a claim.
When there is no formal street address, capture a location instead of a string, and corroborate that location at the level of the area rather than the street. In practice: the applicant drops a pin on a map where they live, adds what only they can tell you (the building, the gate, directions), and you check whether independent records place them in the same area. Then grade how strongly the address is supported rather than calling it simply verified or not.
Many real addresses in Lagos, Nairobi or Kumasi read like "third house after the yellow kiosk, off the main road, blue gate". No string comparison will match that against a utility bill or a government record, however clever the fuzzy matching. The problem is the method, not the data quality.
Why does string matching fail here?
Traditional address verification assumes both sides of the comparison describe the same place in roughly the same format: number, street, city, postcode. Where streets are unnamed or unnumbered, the applicant, the utility company and the government office each write the address differently. Compare them character by character and you produce mismatches out of formatting, and your reviewers learn to ignore the result.
Two things follow. You need a representation of the place that does not depend on how anyone spells it, and you need to compare records at a level of detail they can actually agree on.
What should you capture instead?
A pin on a map
Ask the applicant to place a pin where they live. A satellite view helps: "find your roof" is answerable in a dense compound where "find your street on a road map" often is not. The pin gives you coordinates that are the same however the address is written.
A shareable code derived from the coordinates makes the location usable by people. Plus Codes (opens in a new tab), an open-source addressing system, turn latitude and longitude into a short code that works "for people or places that don't have" a street address, and can be typed into a maps app.
What only the applicant knows
A map can suggest the street and the area. It cannot know the plot number, the flat, the building name or which gate to use. Let the applicant add those, and label them as claims.
Do not make fields mandatory that some people cannot fill. Someone in an unnumbered compound cannot answer a required house number, and a required field they cannot complete is a dead end, not better data.
Directions and an entrance photo
Free-text directions and a photo of the entrance have little value for matching and a great deal for anyone who later has to find the place: a delivery rider, a field agent, a recovery team.
How do you corroborate a location?
Compare the pin with independent evidence at the level of the locality (the town, district or area):
Evidence | What it tells you | Typical agreement check |
|---|---|---|
The government record's address (from the identity check) | Where the person told the state they lived, at some point | Does it name the pin's area? |
A proof-of-address document | Where a bill or statement was sent | Does it name the pin's area? |
A device location fix taken when the applicant confirms | Where the phone was at that moment | Is it within a tolerance of the pin? |
Your own history with the customer | Where you have delivered to or found them before | Is it the same area? |
Treat disagreement with care. Government records are often years old, and people move. A pin that agrees with a recent bill but not with an old government record is a common, benign pattern after a move. Route it to review rather than decline.
How strong is the evidence? Grade it
A useful way to express the result is a small ladder of evidence rather than pass or fail:
Attested: the person demonstrably stood at the pin. Their phone's location fix when they confirmed was close to it.
Corroborated: an independent record, such as the government record or a proof-of-address document, names the pin's area.
Collected: the applicant's claim, with nothing yet backing it. Still valuable for deliveries, field visits and later checks.
Your decision rules then say what each grade is good enough for. A savings account might accept a collected address. A credit line might need corroborated or better.
What about many people at the same pin?
Several customers sharing one location is normal: families, compounds, flats in one building. A sudden cluster of new applicants at one pin in a short period is different, and is a common sign of organised account opening. Watch the recent count, not the all-time count, and use it as a reason for review, never as an automatic block and never as evidence that two people are the same person.
An illustrative flow
A microfinance lender in Kano onboards traders. The applicant searches for their area, drops a pin on a satellite view over their compound, adds "Gidan Malam Sani, green gate behind the mosque" as directions, and confirms. Their phone's location fix at confirmation is 40 metres from the pin. Their government record names the same local area.
The address is attested and corroborated, with no house number anywhere in it. When a loan officer visits months later, the pin, the code, the directions and the entrance photo get them to the gate.
How Myaza Trust handles this
Address Intelligence is a workflow step that captures a map pin on a satellite map, an address search, the applicant's own building, unit and directions, and optionally an entrance photo. Every pin gets a Plus Code. The platform corroborates the pin at locality level against the government record's address and the proof-of-address document, and returns a tier: attested when the device's own location fix at confirmation was within 250 metres of the pin (widened to the fix's reported accuracy, capped at 1 km), corroborated when an independent record names the pin's area, or no tier for a collected claim.
The result never changes the verification's own pass or fail status. It feeds decision rules through fields such as address.tier, address.govRecordMatch and address.entitiesAtSamePin30d, and a shared pin is a review signal, never an identity link. Location permission is never required to finish the flow. The derived street line never invents a house number. Where enabled, a multi-day presence check can confirm the address over the following days; ask us whether it is available for your account. The Address Intelligence documentation covers the configuration.
What to do next
Replace free-text address matching with a pin plus the applicant's own details.
Make only the fields everyone can answer mandatory.
Compare records at area level, and expect old government addresses to disagree.
Decide which evidence grade each product needs.
Route recent clusters at one location to review, never to automatic decline.
Keep directions and entrance photos for the people who will need to find the address later.
Sources
Charles Archibong
Co-founder
Charles Archibong co-founded Myaza Trust. He writes about identity verification, financial technology, and the practical work of building trusted digital services.
