We launched voice payments in a Riyadh banking app in June 2025. Completion rates jumped 63% in four weeks.
This article contains real completion rate data from three verticals: banking, delivery, and e-commerce. Three cities: Riyadh, Dubai, Cairo. Broken down by flow type and user segment. This is production data, not pilots or demos. Some voice features failed. This is analysis, not promotion.
Banking: Money Transfers in Riyadh
Baseline completion rates for typed money transfers sat at 41% before voice integration. Drop-off happened at two points: beneficiary account entry (22% abandoned) and confirmation screen (37% abandoned). Users spent an average of 38 seconds on account number entry alone.
We integrated voice for beneficiary selection and amount specification on repeat transfers. The flow remained typed for new beneficiaries and complex scenarios. Voice activated automatically for users with three or more previous transfers to the same account.
After four weeks, completion rates for voice-enabled repeat transfers hit 67%. That's a 63% increase from the 41% baseline. First-time transfer completion rates stayed flat at 39%, confirming voice worked only for established patterns. Users aged 35-54 showed the highest voice adoption at 71%, while 18-24 users adopted voice at just 23%.
What failed: Voice editing. When users needed to correct an amount or change a beneficiary mid-flow, 89% switched back to keyboard. Voice worked for execution, not modification.
Food Delivery: Voice Ordering in Dubai and Cairo
We tested voice ordering across two cities with different implementations. Dubai focused on repeat orders (same restaurant, similar items). Cairo tested voice for both repeat and new orders. The Voice User Interface market growing at 20%+ CAGR validated our timing, but we needed MENA-specific data.
Dubai completion rates for voice-enabled repeat orders: 78% versus 52% for typed repeat orders. That's a 50% increase. New orders stayed typed, maintaining 44% completion. Users over 45 showed 82% voice adoption for repeat orders. Users under 30 adopted voice at 31%.
Cairo tested voice for new restaurant discovery and ordering. Completion rates dropped to 28% versus 44% for typed new orders. Voice search worked poorly for browsing menus and comparing options. Users wanted to see photos and prices, not hear descriptions. We killed the new-order voice feature after three weeks.
City differences emerged in dialect handling. Dubai's predominantly Khaleeji dialect showed 91% accuracy. Cairo's Egyptian Arabic required dialect-specific training and reached 87% accuracy. Cross-dialect models failed at 62% accuracy.
E-Commerce: Search and Checkout Across Verticals
Voice search for product discovery converted to cart at 12% versus 34% for typed search. Voice failed for browsing. Users wanted visual comparison, not sequential audio results. We removed voice search after two months of testing.
Voice-enabled checkout for repeat purchases (saved address, saved payment) completed at 69% versus 48% for typed checkout. That's a 44% increase. First-time checkout with voice dropped to 31% versus 48% typed. Voice required established trust and simple, low-variance flows.
Flow complexity mattered. Simple replenishment orders (same product, same quantity) showed 73% voice completion. Complex first purchases with multiple SKUs dropped to 29% voice completion. Users aged 40-55 adopted voice checkout at 68%. Users 18-29 adopted at 19%.
What failed universally: Returns, product comparison, browsing. Voice added friction to any flow requiring decision-making or visual evaluation.
Patterns Across All Three Verticals
Success factors held constant: High-frequency actions (repeat transfers, reorders, repurchases), low variance flows (same beneficiary, same restaurant, same product), and established user trust (minimum three previous transactions). Amazon Alexa and Google Assistant hold over 45% market share in voice user interface industry, proving platform maturity that we leveraged in our implementations.
Universal failures: Browsing (voice can't replace visual product comparison), editing (corrections require keyboard), public space usage (67% of users disabled voice in public locations), and complex decision-making (comparing options needs visual reference). Voice assistant users in US expected to grow from 142 million in 2022 to 153.5 million in 2025, showing established behavior patterns we saw replicated in MENA contexts.
User segments: 35-54 age group adopted voice at 3x the rate of 18-29 users across all three verticals. High-frequency users (top 20% by transaction volume) adopted voice at 4x the rate of occasional users. Engineering effort ranged from 6 weeks (banking repeat transfers) to 12 weeks (e-commerce multi-flow integration).
Mobile App Voice Integration ROI Summary
Voice UI conversion rates improved 44-63% for high-frequency, low-variance transactional flows. Voice failed for browsing, editing, and first-time complex transactions. The pattern held across banking, delivery, and e-commerce in Riyadh, Dubai, and Cairo. Voice works for specific transactional patterns, not as a universal interface replacement.