Address validation has a blind spot shaped exactly like an apartment building. The primary address, number and street, validates cleanly against postal reference data such as the USPS addressing standards in Publication 28, which define how secondary unit designators, APT, STE, UNIT, and the rest, are supposed to be recorded. The unit, the part that decides whether the parcel reaches a person or a lobby, is where validation gets vague. Shoppers put the apartment in the second address line, in the first line after a comma, in the company field, or leave it off and trust the driver. Systems then normalise the address against reference data that may not carry unit level detail for that building, return a confident 'valid' on an address that is validly incomplete, and the failure surfaces days later as a failed delivery, a reshipment, and a support ticket in which everyone involved did their part correctly. In dense housing markets this is not an edge case. It is a standing share of orders, failing at the last ten metres for want of one field being checked as carefully as the postcode.
The reference data itself has a unit problem
Postal validation is only as complete as the underlying address database, and unit level coverage is the thinnest part of it. A building can exist in the data with its street address and postcode while its internal unit list is absent or stale, especially for new developments, converted buildings, and large complexes where units were renumbered. Validation against such data can do two bad things: accept an address with no unit as fully verified, training the checkout to ask for nothing more, or flag legitimate unit formats as unrecognised and nudge the shopper into deleting the very detail the delivery needs. Either way the green tick in the checkout is reporting confidence the data has not earned, and the order inherits the risk.
Related reading on this site: Home and Disposable Email Checks Catch Less Than Their Dashboards Claim cover the neighbouring problems in more depth.
Checkout design can catch what reference data misses
The cheapest fix is a question, asked at the right moment. When an address validates to a known multi unit building and no unit was supplied, ask for one, specifically, before payment, not as a generic review your address warning that shoppers have learned to click through. Keep the unit in its own structured field rather than letting it drift through free text lines, because downstream systems, labels, carrier integrations, address book saves, each parse free text differently and the unit is the piece that falls off in transit between them. And when a shopper's saved address and their typed address differ only in the unit, believe the typed one and update the saved one. People move flats inside the same building more often than systems assume, and the saved address book is a museum of former homes.
Failed delivery data is the audit trail
Carriers record why deliveries fail, and 'insufficient address' returns cluster by building and by postcode in a pattern any store can pull. A monthly look at failed deliveries by address turns the validation argument from opinion into a list: these twenty buildings generate a third of our address failures, so these postcodes get the extra unit prompt and the address book cleanup email. Validation vendors sell certainty. Your own failure data sells a map of exactly where certainty is not arriving, which is more useful.
Frequently asked questions
Why does an address validate but the delivery still fail?
Because validation often confirms the primary street address against reference data that lacks complete unit level detail. An address can be formally valid while missing the apartment or suite number the carrier needs to complete the delivery.
Where should apartment and suite numbers be entered?
In the dedicated secondary address field, using the standard designators set out in USPS Publication 28, such as APT or STE followed by the unit number. Keeping the unit in a structured field prevents it being dropped or misparsed as the address moves between checkout, labels, and carrier systems.
What should a checkout do when no unit is given for a multi unit building?
Ask for the unit specifically before payment when the validated address is a known multi unit property. A targeted prompt at that moment catches most omissions, while generic address warnings are routinely ignored.
How can a store find its worst addresses for delivery failure?
By reviewing carrier failed delivery reasons, especially insufficient address returns, grouped by building and postcode. The failures cluster, and the clusters identify exactly where extra validation prompts and address book cleanup will pay off.