Smart pre-conversion upsell

An updated AI model for pre-conversion upsell increased average revenue per paying donor by 8.7%.

What we tested

Pre-conversion upsell is a feature in Checkout that appears after a donor selects a one-time gift, but before they choose a payment method. It invites them to upgrade their gift to a recurring plan right in the moment of giving.

When AI is enabled, it decides whether to show this upsell and selects the optimal amounts to display. For major gifts, the upsell is skipped to keep the experience friction-free.

To ensure this feature stays high-performing, we continuously train and test new AI models. In this experiment, we put a new model head-to-head with our current one across the Checkout Modal and Campaign Pages, using live traffic from organizations with this feature enabled.

Goal

Measure the performance of a new AI model that determines:

Which amounts to suggest during the upsell
When to pre-select the Fee Coverage checkbox

Hypothesis

The updated AI model would lead to higher average revenue per user and improved fee coverage rates, without significantly reducing conversion to donation (no more than 1.5%).

Test setup

Duration	• 26 days on Checkout Modal • 41 days on Campaign Pages
Timing	May-July, 2025
Audience	A total of 319,787 donor sessions: • 230,597 visitors in Checkout Modal • 89,190 visitors in Campaign Pages
Group split	Visitors were randomly assigned to one of four groups (each with 25% of traffic): • Control: Existing logic for upsell display and Fee Coverage • Test 1: New AI logic for showing the upsell screen + suggested amounts + updated Fee Coverage logic • Test 2: New AI logic for suggested amounts only + updated Fee Coverage logic • Test 3: New AI logic for upsell screen display only + updated Fee Coverage logic
Key metrics	• Donation conversion rate: Percentage of visitors who completed a donation • Recurring conversion rate: Percentage of visitors who chose a recurring gift • Average Revenue Per User (ARPU): Average donation revenue across all visitors, including those who didn’t donate • Average Revenue Per Paying User (ARPPU): Average revenue from visitors who donated, capturing the value of each donor • Fee coverage rate: Percentage of donors who opted to cover transaction costs These metrics help us evaluate both the reach and depth of donor impact, measuring not just how many people gave, but how much they gave and whether the upgrades improved financial outcomes.

Key results

Segment insights

Strongest gains were observed for donations between $25–$100
- +4.2% in ARPPU for one-time donors
- +8.7% in ARPPU for recurring donations (with LTV)
ARPPU lift in recurring giving was particularly strong in Canada (+30.13%)
ARPU with LTV lift by countries and devices:
- +4.61% in Great Britain
- +5.82% in Australia
- +2.90% in Canada
- +3.39% on Android devices

All key metrics passed the threshold for statistical significance (p-value < 0.05), confirming that these improvements are reliable.

Campaign Pages for Test 2 group

No statistically significant uplift observed in any tracked metrics
Small positive effects in:
- EU one-time donors: +2.3%
- Mac OS users for recurring donations: +0.6–5.6%

Note: Test Groups 1 and 3 were disabled on June 11 after a drop in recurring conversion was observed.

Conclusion

Campaign Pages

The experiment didn’t produce statistically significant gains.
No changes will be rolled out at this time.

What this means for you

Every optimization we ship is backed by rigorous testing, so you don’t have to guess what’s working. This update brings smarter, AI-powered suggestions to the pre-donation upsell screen, helping your donors give more, more often, with less friction.

You benefit from: