What Is DoorDash A/B Testing On Its Checkout And DashPass Upsell?
Lazyweb Research detected 221 distinct experiments at DoorDash (July 2026) — the largest single-company test set in the entire 276-company corpus, roughly 2.6x the next-biggest app. [1] The visible activity concentrates on the home surface (at least 45) with a smaller paywall footprint (at least 5), and the standout dated diffs turn the order-confirmation moment into a concrete-dollar DashPass upsell. [1][2] These are observed before/after variations with inferred rationale, not company-confirmed A/B tests.
Lazyweb Research detected 221 DoorDash experiments (July 2026) — the largest test set of any app tracked, including a checkout upsell showing '$9.00 SAVED with DashPass'.
The finding
Lazyweb Research detected 221 distinct experiments at DoorDash, the highest count of any company in the corpus and roughly 2.6x the second place (Wattpad, 85). [1][3] The detected mix skews to the home surface (at least 45 experiments) with a modest paywall footprint (at least 5). [1] DoorDash is the corpus's clearest example of a high-frequency experimenter iterating across the whole shopping funnel, not just the paywall.
What actually changed
Two dated 2026 diffs show DoorDash weaving monetization into transactional moments. On 2026-01-22, a plain "Submitting Order" modal was replaced by a full-screen DashPass upsell reading "$9.00 SAVED with DashPass deals" with a smaller order card below — inferred rationale: surfacing a concrete dollar amount saved at the moment of commitment increases subscription uptake. [2] On 2026-02-06, the delivery progress bar added a car icon with a checkmark plus a "Your order is complete. Enjoy!" confirmation message — inferred rationale: an explicit completion message improves clarity and satisfaction. [4]
| Detected change | Date | Inferred rationale |
|---|---|---|
| "Submitting Order" modal -> full-screen DashPass upsell ("$9.00 SAVED") | 2026-01-22 | Concrete dollar-saved at commitment lifts subscription uptake [2] |
| Progress bar adds checkmark + "Your order is complete. Enjoy!" | 2026-02-06 | Explicit completion message improves clarity [4] |
How to apply it
DoorDash's checkout upsell is the copyable pattern: attach the subscription pitch to a high-intent transactional moment (order submit) and quantify the benefit in dollars rather than abstract "member perks." If your upsell lives on a settings page far from purchase intent, moving it into the commit step — with a concrete savings number — is the DoorDash-style test. The progress-bar confirmation is a separate clarity change; run it independently. Both are detected variations, not proven winners. [2][4]
Caveats
All 221 experiments are observed variations with LLM-inferred rationale, not company-confirmed A/B tests — no lift is measured. [1] Surface splits are lower bounds because screen category is unlabeled on 1,425 of 4,814 corpus experiments. [cat_null] Only 2 DoorDash experiments were detected in 2026, so this page cites specific dated diffs rather than a 'this year' volume. [1]
The numbers
| Stat | Computed from |
|---|---|
| 221 distinct experiments; at least 45 home, at least 5 paywall; 2 in 2026 | company_total:doordash (value 221; home 45, paywall 5, in-2026 2) |
| Checkout DashPass upsell '$9.00 SAVED', 2026-01-22 | qualitative[] doordash 2026-01-22 entry |
| Wattpad is the second-largest at 85 experiments | company_total:wattpad (value 85) |
| Order-complete confirmation added, 2026-02-06 | qualitative[] doordash 2026-02-06 entry |
| 1,425 of 4,814 experiments have no screen category | screen_category_null_on_experiments (1425/4814) |
Sources & citations
- [1] Lazyweb Research analysis of 221 detected experiments (DoorDash, ~800-app mobile corpus), July 2026. COUNT(DISTINCT experiment_id) on before/after diffs; surface splits from is_paywall + screen_category. ↩
- [2] Lazyweb Research analysis of 221 detected experiments (DoorDash, ~800-app mobile corpus), July 2026. Dated before/after diff, 2026-01-22; rationale is LLM-inferred, not company-confirmed. ↩
- [3] Lazyweb Research analysis of 85 detected experiments (Wattpad, ~800-app mobile corpus), July 2026. COUNT(DISTINCT experiment_id); second-highest company total in the corpus. ↩
- [4] Lazyweb Research analysis of 221 detected experiments (DoorDash, ~800-app mobile corpus), July 2026. Dated before/after diff, 2026-02-06; rationale is LLM-inferred. ↩
- [cat_null] Lazyweb Research analysis of 4,814 detected experiments (276 companies, ~800-app mobile corpus), July 2026. screen_category is NULL on 1,425 experiments, so all surface splits are lower bounds. ↩
Source: Lazyweb Research — proprietary analysis of real, in-market app screens. Cite as Lazyweb Research, 2026-07-07.