AI-selected upgrade amounts in donor emails

A new AI model for upgrade suggestions showed no measurable impact.

What we tested

Recurring donors are one of the most valuable sources of sustainable fundraising revenue. That's why we're continuously looking for ways to help existing supporters deepen their commitment and increase their impact over time.

One opportunity comes through recurring donation emails, where supporters can not only confirm their latest gift but also increase their monthly contribution through AI-suggested upgrade amounts.

We wanted to understand whether more personalized upgrade recommendations could encourage more supporters to increase their recurring gifts. To test this, we replaced the AI model with a newer AI model designed to generate more relevant upgrade amounts for each donor.

Hypothesis

If we generate the suggested upgrade amounts in installment emails with a new AI model, the share of recurring donors who upgrade their recurring plan will rise by about 11%, because more relevant suggestions make stepping up feel like an easier, more natural choice.

Test setup

Duration	64 days
Timing	November 2024–January 2025
Audience	~415,000 recurring donors with a USD recurring plan under $200 who received an installment email
Group split	• 50% control (suggested amounts from the existing model) • 50% test (suggested amounts from the new AI model)
Key metrics	• Upgrade conversion rate (share of donors who increased their recurring plan) • New donation conversion rate • Average donation amount (ARPPU) • Average revenue per user (ARPU)

Key results

The new model performed similarly to the existing approach across all major fundraising metrics.

No statistically significant change in upgrade conversion
No statistically significant change in new donation conversion
No statistically significant change in ARPU or ARPPU
No statistically significant change in recurring plan downgrades or cancellations

One notable behavioral difference emerged: supporters were more likely to select the first suggested upgrade amount when presented with recommendations from the new model.

Conclusion

The hypothesis was not confirmed.

The new AI model did not increase recurring donation upgrades or improve fundraising outcomes compared to the existing AI model.

While donor behavior shifted slightly toward selecting the first suggested upgrade amount, the overall impact on recurring revenue was neutral.

As a result, we will continue using the current AI model and pause further development of the new version. Future testing will focus on broader measures of recurring revenue growth and supporter value.

What this means for you

Growing recurring revenue isn't just about acquiring new monthly donors. It's also about helping existing supporters increase their impact over time.

That's why we continuously explore new ways to strengthen recurring giving, but only ship the changes that produce measurable fundraising results.

With Fundraise Up, your recurring giving program is constantly being refined, tested, and optimized to help you build stronger donor relationships and more sustainable revenue.

→ Explore how recurring giving tools can strengthen your donor relationships.