If you’re reading this, you probably already know giving cash directly to someone in poverty works, and are donating based on the research and evidence. You might also believe that you decide where to give based solely on a nonprofit’s mission or effectiveness. However, most of us can be nudged to give based on less-than-logical factors, like the design of a website or the subject line of an email.

Similar to the randomized controlled trials we use to measure the impact of our cash transfer programs, our fundraising team often runs tests to see which tactics will raise the most money. The goal of these tests is to maximize the dollars we deliver to people in poverty at the lowest cost, and without using poverty porn to guilt you into giving. By sharing our learnings publicly, we want to help other organizations maximize their fundraising dollars to do more good.

1. A better web design and donation form made people 34% more likely to donate

GiveDirectly only started a fundraising team a few years after it was founded, and that very small team only had the resources to maintain a pretty bare-bones website. Because many early donors were drawn to us due to the evidence behind cash transfers, we thought this website conveyed that we were more focused on efficiently delivering your dollars than slick marketing.

As we grew to deliver funds from major institutions and governments, our scrappy website started to hurt more than it helped our image. If you worked at an aid agency, you might worry that the tech-savvy organization you just funded looks like they run on GeoCities.

Once we committed to updating our homepage, we used it as an opportunity to run a three-arm test to find out:

  1. Does a polished homepage with more recipient voices raise more money than a simple homepage?
  2. Does embedding the donation form raise more money than a simple “donate” button?

The test found that:

  • A website visitor was 34% more likely to donate if they saw the new design with embedded donation form instead of the old one (with conversion increasing from 1.98% to 2.67%, which is strong for the nonprofit industry).
    • 20% of this uplift was driven by the embedded donation form
    • 14% was driven by the new design
  • These changes will raise up to $710,000 more per year for people in poverty at our current website traffic level

Because of these results, we’ve made other webpages (like this one about extreme poverty) more robust and are already seeing more donations there as well.

2. Larger match multipliers don’t drive more donations, but contingent matches can

“A generous donor has offered to 8X your donation if you rush your gift before the clock strikes midnight!” says nearly every nonprofit email you receive. If you’re a more skeptical donor, this might raise the question: do match offers actually increase donations?1

The short answer is yes: Existing evidence shows that matches do raise more money, and we’ve seen similar results with our own tests — offering a match raised 7-8x more money than asking for donations without a match. On Giving Tuesday in 2022, the pace of donations dropped by half after the $150,000 match fund was used up around 2:30pm:

So we knew that offering a match is better than no match at all — but given that we don’t have unlimited match funds, how do we maximize dollars for recipients with the funds we have? We wanted to answer two specific questions to help figure out the optimal match structure:

  1. Which match multiplier works the best? (e.g. 1.5x? 2x? 3x?) 
  2. Do contingent matches work better than normal multiplier ones? (e.g. “if we raise $50,000 today, it unlocks another $10,000”)

A bigger match multiplier doesn’t raise more money, and a really small multiplier can still encourage people to give

Donors who seed matches typically don’t have stipulations for how to structure the match. So we created an email experiment testing various match sizes (1.5x, 2x, and 3x) compared to a control group asked to give without a match, and we found that:

  • Donors offered any match were 2x as likely to click and 3-5x as likely to donate
  • Offering a higher match multiplier didn’t lead to statistically significant increases in donation rate or average donation amount

Given every match multiplier performed similarly, these results suggest we should just offer the lowest possible match multiplier in order to preserve match funds for as many donors as possible:

But how low can you go and still have an effect?

We ran another small test comparing 1.1x, the smallest plausible multiplier, against the common 2x match (i.e. your $10 donation becomes either $11 or $20). 

  • The 1.1x match convinced 25% fewer donors to give compared to the 2x match, and the average donation was 19% lower
  • This means that the total amount raised was 40% lower for the 1.1x match 
  • However, the 1.1x match used 94% less match funds than the 2x match

Because of this test, our strategy with matches is now:

  • Use a 1.5x multiplier for most matches, and never go above a 2x match multiplier2
  • If there are future campaigns where we expect to be constrained by match funds rather than audience size, we should consider a multiplier below 1.5x to stretch them further

A match contingent on a group taking a joint action can be more motivating than a standard match

“A donor will give $25,000 if 50 people make their first donation today” is an example of a contingent match. Rather than multiplying your donation by some factor, these matches unlock a lump sum donation once a certain threshold is reached. This structure is well-suited when optimizing for a behavior (like upping a monthly donation) as opposed to simply immediate dollars raised.

To learn which match structure works better, we ran a test offering monthly donors one of two matches:

  1. 2x multiplier: Increase your monthly donation in the next 5 days and have the first month doubled
  2. Contingent: If 50 people increase their monthly donation in the next 5 days, a donor will give $25,000

We found that: 

  • Monthly donors were 3.7x more likely to increase their donation when offered the contingent match compared to the multiplier match (2.6% vs. 0.7%)
  • The contingent match email also had a 71% higher clickthrough rate on the email
  • However, the average amount donors increased their monthly donation was 35% smaller for the contingent group ($20 vs. $31). 
  • Overall, the contingent match still raised 2.3x more money than the multiplier match, because the increase in likelihood to give outweighs the decrease in average donation increase.

However, these matches need to be more carefully structured than a standard 2x match: in this experiment, only 34 monthly donors increased their monthly donation (below the 50 needed), which means we did not unlock the contingent donation. We tried to set a threshold that was aggressive but achievable, but our estimates for how many people would respond to the contingent match were too high in this case.

3. Pledging to give drives more donations, but it’s difficult to get people to pledge

Every year, anywhere from 14% to 75% of the donations we receive from individuals come in December (before the U.S. tax deadline). This volatility makes it difficult to forecast how much money we can deliver to people in poverty the following year. So in summer 2021, we tested sending emails to donors to ask how much they planned to give at the end of that year:

Those who made a pledge received a reminder email that December about the amount they’d pledged to give.

Compared to donors who didn’t receive the pledge email:

  • People who pledged to give (pledgers) were 2.5x more likely to give (75% vs. 30%)
  • On average, pledgers gave 93% of the total they’d originally pledged, showing that pledges could be a good way to predict future donations 
  • People who received the pledge email gave 19% more
  • However, it’s possible that the self-selecting donors that chose to respond to the pledge email were more likely to give anyways, and were not specifically influenced by the email itself

However, this effort did not ultimately raise much more money, because <6% of people who received the pledge email made a pledge. This is partly due to the difficulty of getting engagement over email, as only 15 – 30% of donors open any given email we send. We tried it again in 2022 to see if we could increase the pledge rate, but saw similarly low engagement. Given the effort required to run this pledge process, we decided to not repeat it in future years.

Do you have an idea for a future fundraising test, or want to help fund our next experiment? Let us know here, or apply for one of our job openings.

Footnotes

  1. This might also raise another question: are these match offers “real”? i.e., wouldn’t the people seeding match funds donate regardless? In an ideal world, every dollar donated to a match fund wouldn’t have been donated otherwise, and that’s certainly what we aim for when we decide which donors we ask to fund a match. However, a 100% counterfactual match is rarely feasible since many match donors would give regardless. Ultimately, even though not all match funds might be incremental, we’ll continue to offer them because they still conclusively drive more donations for people in poverty. Many other nonprofits may not be as rigorous with their matches: many are always running a match, and may even have fine print implying that not every donation will be matched.
  2. We still sometimes use a 2x multiplier for major fundraisers, because it’s what donors expect for events like Giving Tuesday and “double your donation” is the most explainable type of match. We’re also working on adding the ability to run 1.5x matches to our fundraiser feature.