AR Shopping
The number one reason people abandon food orders? They can't tell how big it is.
47% of users surveyed said they abandoned a DoorDash cart because they couldn't tell how big the portion was. That's not a content problem. That's a spatial problem — and AR was the answer.
Setting the Scene
DoorDash had cracked delivery logistics. But at the moment of purchase — the decision-making stage — customers were still flying blind. A photo and a description. No scale. No context. No way to know if this burrito bowl feeds one person or could double as a centerpiece. The gap between 'photo on screen' and 'thing in your hands' was costing real money.
The Problem
High cart abandonment and buyer's remorse aren't random. They're the result of a specific information gap: users can't visualize what they're actually ordering. Especially for new restaurants, unfamiliar cuisines, or anything where portion size, presentation, or composition really matters — uncertainty kills conversion.
What made it hard
- →ARKit only worked on iOS 11.3+, so Android had to wait (8 months, as it turned out)
- →Building photorealistic 3D food models was expensive — every menu item needed its own asset
- →Stakeholders were worried AR would add friction rather than reduce it
- →Older AR-capable devices struggled with performance — had to optimize aggressively
- →The integration had to land within a 6-month window tied to a broader shopping redesign
The key reframe: AR isn't a 'cool feature.' It's a confidence tool. Users didn't need immersive — they needed accurate. One question: 'What will this actually look like?' AR answered it better than any photo ever could.
What the Research Revealed
- →Portion size uncertainty was the #1 cart abandonment reason across every restaurant category we tested
- →Users expressed more confidence ordering from unfamiliar restaurants when AR was available — even before using it
- →Merchant adoption was highest when AR launched from the product card itself, not buried in a menu
- →AR engagement was strongest with new users (first 5 orders) who hadn't yet built restaurant familiarity
- →Realistic 3D models drove confidence; stylized models felt like a gimmick and actually reduced trust
The Approach
I designed a lightweight AR experience integrated directly into the product card flow. Tap 'View in AR,' an ARKit session opens, and a photorealistic 3D model of the meal appears at actual scale in your environment. Intentionally simple — no avatars, no gamification, just accurate representation. For devices without ARKit support, high-quality carousel views and size comparisons stepped in so no user was excluded.
AR entry point on the card
'View in AR' lives as a prominent secondary action on product cards — discoverable but never forced. You can ignore it. Or you can use it and never wonder about portion size again.
Photorealism over stylization
I tested both. Realistic models increased purchase confidence measurably. Stylized models felt gimmicky. Users wanted to see exactly what they'd receive — so that's what we showed them.
Scale indicators
Subtle UI elements (a coin, a hand outline) helped users understand scale at a glance. Testing showed this reduced confusion about sizing by 35%.
Floating placement over floor detection
First version required users to 'find the floor.' They got stuck. Floating the product at eye level resolved 90% of that friction immediately.
Proving It Out
Beta rolled out to 5,000 randomly selected iOS users. Moderated usability tests with 12 first-time AR users. A 2-week A/B test across 50k users. And honest conversations with the top 20 restaurant partners about what they were seeing in their numbers.
- →97% of users completed an AR view without any assistance — the interaction was genuinely intuitive
- →Time spent viewing a product jumped from 8 seconds to 22 seconds when AR was available
- →Cart abandonment dropped from 18% to 13% in the AR-enabled cohort — statistically significant at p<0.01
- →81% of merchants wanted AR for their full catalog after seeing the conversion impact on their numbers
Beyond the Numbers
- →User sentiment shifted from 'I'm unsure about this restaurant' to 'I know exactly what I'm getting'
- →Merchants reported significantly higher conversion when AR was enabled on their menu items
- →New user retention improved — users who experienced AR in their first order came back more
- →Marketing used AR availability as a competitive differentiator in local markets
What I Took Away
AR only earns its place when it solves a specific, real problem in the user's journey. Frame it as confidence, not novelty — and adoption follows. The moment you make it a spectacle, you've lost.
- →Photorealistic 3D asset production is expensive. Partnering with restaurants to supply their own photography for 3D conversion was more scalable than doing it in-house
- →Merchant adoption follows user adoption — had to prove the feature drove orders before restaurants invested time in 3D assets
- →Simplicity in AR UX beats richness. Every additional gesture or UI element increased drop-off. Final version was 40% simpler than the prototype
- →iOS-first was correct for 2023, but earlier cross-platform planning would have accelerated Android parity
What's Coming Next
AR filtering in search and browse (find AR-enabled restaurants earlier in the journey). A self-service tool letting merchants convert their own photography into 3D models. Multi-item AR — visualizing an entire meal together, not just individual dishes. Hyperlocal analysis of AR usage patterns by geography and cuisine type.