Your checkout funnel shows a significant drop-off at the address step, but the field works fine in your QA environment. Nobody filed a bug. No error logs. The autocomplete just stops helping at some point, and buyers bail.
This is one of the more frustrating problems to debug because autocomplete failure is usually invisible in standard analytics. A drop-off at step 3 shows up in your funnel report, but nothing tells you why the buyer stopped there.
What actually goes wrong with address autocomplete
There are a few distinct failure modes, and they look identical in your funnel data.
Failed or missing suggestions. The autocomplete API returns nothing—or returns suggestions for the wrong country—when the buyer types a partial address. On mobile, where there's no keyboard shortcut to paste, this means starting over. Many don't.
Wrong geocoding. The suggestion looks right but maps to the wrong unit, building, or zip code. The buyer selects it, doesn't catch the error, and either gets a failed delivery later or spots the mismatch during order review and abandons rather than fix it manually.
Mobile keyboard friction. Address fields often trigger the wrong keyboard type on iOS and Android. If the number pad pops up for a street address field, the buyer has to switch keyboards before typing. It's a small thing, but on a slow mobile connection with autocomplete already half-broken, it's enough to lose them.
API rate limits or latency spikes. If you're using the Google Places API with a capped quota, suggestions stop appearing at peak traffic. This looks like a random, intermittent drop-off—and it often only surfaces during paid campaign pushes when your traffic is highest and you can least afford it.
Any of these can gut your address step completion rate. Baymard Institute research suggests ecommerce checkout flows can recover up to 35% more conversions by fixing friction at individual steps—address fields included.
Why standard analytics won't catch this
Event tracking tells you a buyer landed on step 3 and didn't reach step 4. Heatmaps might show rage clicks near the address field. But neither tells you what the autocomplete widget did when the buyer started typing.
You need to watch the session.
What session replay shows you at the address step
With session replay, you can filter sessions by funnel step (address entry → checkout abandonment) and watch what actually happened. The patterns you'll spot:
- The autocomplete dropdown appeared but showed suggestions for the wrong region
- The buyer typed a full street address, got no suggestions, deleted everything, typed again, and gave up
- The dropdown showed the right street but didn't populate city and zip—so the buyer had to fill those manually before abandoning
- On mobile, the buyer tapped the address field, the keyboard covered the dropdown entirely, and they couldn't see any suggestions at all
In CloseTrace, the practical filter to run is: sessions where the address field was focused AND the checkout step was not completed, segmented to mobile only. That separates the abandoners who actually tried to use the field from those who left before engaging with it at all. Watching 10–15 of those sessions back-to-back usually surfaces the dominant failure mode within an hour.
The geocoding problem is sneakier than the others
One caveat worth naming: some autocomplete issues only surface post-purchase, not at checkout. If the autocomplete suggests a plausible but wrong address and the buyer accepts it, they complete the checkout. The problem shows up later as a failed delivery or a customer service ticket.
Session replay won't catch that failure on its own—you'd need to cross-reference it with your fulfillment data. But replay does catch the cases where buyers noticed the wrong address during order review and abandoned, which is the more common version of the problem and the more recoverable one.
Also worth flagging: if your autocomplete library is client-side only without backend address validation, you can't fully trust that suggested addresses are deliverable. That's a separate fix from the UX issues session replay surfaces, but it warrants a separate audit.
Where to start
If your funnel shows unusual drop-off at the address step—especially on mobile or during paid traffic spikes—watch the sessions before you change anything.
Run this in CloseTrace: address field interaction + no checkout completion, mobile only, last 14 days. Watch 10 sessions. You'll likely see the same failure pattern repeat, which gives you a concrete fix to prioritize—wrong region settings, missing city/zip population from the suggestion, or an API quota that's throttling suggestions at peak hours.
If you're also losing buyers who started filling earlier steps but never reached the address field, lead recovery can capture those partial entries before they go cold. The post on d2c checkout funnels breaks down where the other drop-offs tend to hide in the steps around address entry.
Fix the dominant failure mode first, then recheck the funnel numbers.