Postcode lookup at checkout: the address autocomplete errors nobody logs

A postcode in the UK can cover a run of addresses on both sides of a street, and in dense areas a single code can serve a tower with hundreds of flats. The lookup resolves the code to a list, the shopper taps the entry that looks like theirs on a small screen, and the form fills with a neighbouring number or the building name without the flat. Every field is valid. The parcel still goes to the concierge of the wrong block or the house four doors down.

Carriers then mark the order delivered, correctly, to the address on the label. The shopper report says never arrived, support sees a delivered scan, and the store reships or refunds while learning nothing, because the error lives in the gap between the address the shopper meant and the address the lookup recorded. Logging which fields the lookup filled, and whether the shopper edited them, turns that gap into data.

How a correct postcode delivers to the wrong door

Flats, suites, and units are the data lookups handle worst, because reference datasets record buildings more reliably than the subdivisions inside them. The pattern to defend against is specific: the shopper types Flat 4, selects their building from the lookup, and the selection rewrites the address lines without the flat. The fix is a rule developers can implement in a day: lookup selections may fill empty fields and may suggest changes, but they may not silently delete a unit identifier the shopper already entered.

The same rule protects business deliveries, where the company name and unit on an industrial estate are the parts a driver actually needs. If your failed delivery reports cluster on business parks and apartment blocks, you do not have a carrier problem. You have a lookup overwrite problem with a carrier symptom.

Unit numbers are the first casualty

Every checkout with a lookup needs a manual entry path that is one tap away, labelled plainly, and does not punish the shopper for using it. Some addresses are new builds missing from the dataset, some are houses with names and no numbers, and some shoppers simply know their address is unusual. Burying manual entry behind a link styled as an error recovery makes those shoppers feel they did something wrong.

Keep the manual path fast and unembarrassed

Keep reading: International phone fields at checkout: format errors you caused yourself, Auto applied discounts at checkout: the test is what happens to the code box.

Verify the resolved address like it matters, because it does

Watch the manual entry rate in your analytics as a health metric, not a failure. A sudden rise after a theme or app change usually means the lookup broke or narrowed its coverage. A steady small rate is normal and healthy: it is the sound of the edge cases being served instead of forced through a dataset that does not contain them.

Before payment, the confirmation the shopper needs is one clean rendering of the full delivery address, unit included, with an edit link that returns them to the form with their entries intact. Stores skip this because the address was just entered, which misses the point: the shopper entered a postcode and a tap. The rendered address is the first time they see what the store thinks they said.

Measure the change on failed delivery rate and on address related support contacts, not on checkout completion, which barely moves. Autocomplete earns its place on entry speed. The verification step earns its place on the van not coming back with your parcel still on it. Between the two, the second is the one with the invoice attached.

Frequently asked questions

Why do orders go to a neighbouring address when the postcode was correct?

Because a postcode covers a range of addresses, and the lookup step asks the shopper to pick their exact address from a list. A wrong pick, or a selection that drops a flat or unit number, produces a valid looking address that is not theirs. Showing the full resolved address for confirmation catches most of these before dispatch.

Should address lookup overwrite fields the shopper already typed?

No. A lookup selection should fill empty fields and suggest the rest, but never silently delete information the shopper entered, especially flat, unit, or suite identifiers, which reference datasets handle least reliably. Overwriting is the mechanism behind most lookup related delivery failures.

Is manual address entry still necessary with autocomplete?

Yes. New builds, named houses, and unusual addresses are routinely missing from lookup datasets, and those shoppers need a fast, clearly labelled manual path. Track the manual entry rate: a sudden spike after a site change usually signals a lookup problem.

How should I measure whether my address lookup is causing problems?

Look at failed delivery rate, address related support contacts, and how often shoppers edit fields after a lookup selection. Completion rate barely responds to lookup quality, so it is the wrong metric for this particular question.