Cutting Checkout Drop-off by 40%

Chaldal came to us with a funnel that looked fine in aggregate and terrible in every cohort cut. Checkout completion had drifted from 61% to 46% over two quarters, and the usual suspects — page speed, payment failures — were clean.

The drop-off wasn’t one step, it was the address form

Session recordings showed the real pattern once we watched enough of them back to back: users weren’t abandoning at payment, they were abandoning while re-entering delivery addresses that autocomplete kept getting wrong for non-standard Dhaka addresses.

We shipped a saved-address model with manual pin-drop as a first-class option instead of a fallback, and rewrote the autocomplete matching to weight landmark text over street-name text — landmarks are how most people actually describe where they live here.

Results held through Ramadan, the real test

Checkout completion recovered to 64% within three weeks and, more importantly, held through the Ramadan order spike — the period that had originally masked the problem as "seasonal."