# Fundraising ROI: Does an Online Campaign Raise New Money, or Move It?

A geo-randomized field experiment measured what a Facebook fundraising campaign added to giving, not what the ad platform credited to it. This guide explains the result and a simpler check any organization can run.

## Key takeaways
- In a field experiment by Maja Adena and Anselm Hager, run in November 2017 and published in Management Science in 2025, almost 8,000 German postal codes were randomly assigned to see a charity's Facebook fundraising videos or no campaign at all.
- The campaign raised donation revenue and frequency to the charity during the campaign and for five weeks after, and it was profitable for the fundraiser, but some of the new giving was drawn away from similar charities.
- The authors also found that click data may be an inappropriate stand-in for donations, which is a warning about any return figure built from clicks and platform-credited conversions.
- M+R Benchmarks 2026 reports that nonprofits spent $0.10 on digital ads for every dollar of 2025 online revenue, with paid search returning $2.48 per dollar spent and display ads $1.11, figures based on revenue credited to ads, not on randomized comparisons.
- An organization can check its own campaigns without withholding asks from anyone, by comparing all-channel revenue against a matched earlier period or by starting a campaign in some regions a few weeks before others.

Real fundraising ROI is the extra money a campaign brings in compared with what would have arrived anyway, minus what it cost. That is a harder number to find than the one an ad dashboard reports. The best evidence on it comes from a nationwide field experiment by Maja Adena and Anselm Hager, run in November 2017 and published in [Management Science](https://doi.org/10.1287/mnsc.2020.00596) in 2025. It is one of the few public field tests of this question that counted every donation rather than clicks. They randomly assigned almost 8,000 German postal codes either to see a large international children's charity's Facebook fundraising videos or to see no campaign, then counted every donation to the charity from each postal code, through every channel. The campaign raised giving during its run and for five weeks after, and the authors' best estimate is that it paid for itself, though the uncertainty range includes a loss. It also drew some giving away from similar charities. A return figure that ignores both of those facts is measuring something else.

## How did the Facebook fundraising experiment work?

The design is what makes the result trustworthy. According to the authors' [working paper](https://bibliothek.wzb.eu/pdf/2024/ii20-302r2.pdf) (WZB discussion paper, second revision February 2024), the researchers assigned 94% of Germany's 8,181 postal codes either to a 14-day Facebook video campaign or to a pure control group that saw no campaign. The campaign ran in November 2017, a normal season for fundraising appeals in Germany. They then studied the charity's complete donation records by postal code for 12 weeks, covering the month before, the campaign itself, and the weeks after through December 31.

Counting all donations by place, rather than clicks on ads, solves two problems at once. A donor who saw the ad and then mailed a check is counted. And a donor who clicked the ad but would have given in December anyway is not counted as new money, because the comparison with the control postal codes removes giving that would have happened regardless. The published abstract says the geo-randomized design "circumvented many difficulties inherent in studies based on click-through data, especially substitution and measurement issues."

## What did the campaign actually add?

The published paper reports three findings: the videos "increased donation revenue and frequency" to the charity "during the campaign and in the subsequent five weeks," the campaign "was profitable for the fundraiser," and the effects were similar whichever video was shown and however impressions were assigned. The working paper gives the point estimates. Donations rose by an estimated €17.65 per million inhabitants per day, against a control-group average of €129.3. Each €1 spent on the campaign returned an estimated €1.45 in additional donations over the campaign and the 38 days after it, and an estimated €2.53 over the long run once the authors applied a lifetime multiplier of 1.75. That multiplier is an assumption, built from typical retention of about 30% after a donor's first year and about 60% in later years, that each donation eventually brings further gifts worth 75% of it. The authors call these lower-bound estimates for a largely untargeted campaign, and the figures come from the 2024 working paper. The published 2025 version may report revised estimates.

**€1.45** — in additional donations for each €1 of Facebook ad spend during and five weeks after the campaign, measured against randomized control postal codes (working-paper point estimate) ([Adena & Hager, WZB Discussion Paper SP II 2020-302r2](https://bibliothek.wzb.eu/pdf/2024/ii20-302r2.pdf))

The uncertainty around those estimates is wide, and the authors say so. The campaign cost €33,700 in direct costs. The point estimate of additional donations was €47,726, but the 90% confidence interval, the range the authors' method says the true figure falls in 90% of the time, ran from €4,892 to €90,571. The low end of that range would not have covered the cost. The working paper's totals divide to about €1.42 per €1 rather than the €1.45 it reports, a small inconsistency within the paper that does not change the conclusion. The honest summary is that the best estimate was profitable, the true figure could have been well below or well above it, and one campaign in one country is a strong clue rather than a law.

## Where did some of the new money come from?

The experiment also measured giving to other charities. The authors found "some crowding out of donations to other similar charities or projects," meaning that part of what the campaign raised for the charity would otherwise have gone to organizations working on similar causes. For the charity running the ads, that money is still new revenue. For the sector as a whole, part of it was moved rather than created. The working paper adds that this suggests donors may not treat their giving budgets as flexibly as some earlier research assumed.

> Click data may be an inappropriate proxy for donations. — Adena and Hager, Management Science (2025), abstract

The finding about clicks may matter most for everyday practice. Clicks and video views are what an ad platform can see, and they are what most return figures are built from. The experiment's measure, total donations by place compared with a control group, gave a different and more reliable picture. For more on how credit for a gift gets assigned across channels, see our guide to [channel attribution](https://donorinsights.com/articles/channel-attribution).

## What does fundraising ROI in ad benchmarks actually measure?

Return on ad spend, or ROAS, is the revenue credited to ads for each dollar spent on them. [M+R Benchmarks 2026](https://mrbenchmarks.com/advertising/) reports that nonprofits increased digital ad spending 18% in 2025 and spent "$0.10 in digital ads for every dollar of online revenue." Paid search returned $2.48 per dollar spent, display ads $1.11, and digital video $0.56. On TikTok, it took $590 of spending to generate one donation, a ROAS of $0.04.

Return on ad spend by channel for nonprofits in 2025, as revenue credited to ads per $1 spent (M+R Benchmarks 2026, Advertising)

| Channel | Revenue credited per $1 of ad spend |
| --- | --- |
| Paid search | $2.48 |
| Display | $1.11 |
| Digital video | $0.56 |
| TikTok | $0.04 ($590 per donation) |

These benchmarks are useful for comparing channels with each other. They are not the same measurement as the experiment. Return-on-ad-spend figures are built from revenue that tracking systems credit to ads, which, depending on the tracking setup, can include view-through revenue, defined in M+R's [glossary](https://mrbenchmarks.com/glossary/) as "revenue from donors who made a donation (typically within 30 days) of seeing, but not clicking on, an ad." Credited revenue can include gifts that would have arrived anyway, and it misses gifts that arrived through the mail or by phone. The experiment's authors conclude that click data may be an inappropriate proxy for donations, so the credited figure and the added figure should not be assumed to match.

## How can you check whether your own campaign raised new money?

Few organizations can randomize ads across thousands of postal codes. Two simpler comparisons get closer to the truth than a platform dashboard, and neither withholds an ask from any donor. The first compares all-channel revenue in the campaign weeks with a matched earlier period, adjusted for the growth you saw in months with no campaign. The second starts the campaign in some regions a few weeks before the others, so every region is asked, and compares the early regions with the late ones during the gap. Both are strategy contrasts rather than proof of cause. They point in a direction, and a result that repeats across campaigns deserves more trust than a single run.

Here is the first method for a fictional organization, Stonebridge Animal Rescue, which spent $6,000 on a November social-media campaign. The figures are illustrative.

Illustrative matched-period check for a fictional organization, Stonebridge Animal Rescue (hypothetical figures)

| Measure | Amount |
| --- | --- |
| Ad spend | $6,000 |
| Revenue the ad platform credited to the campaign | $15,000 (a reported return of $2.50 per $1) |
| All-channel revenue, campaign weeks plus five weeks after, this year | $62,000 |
| Same weeks last year, no campaign | $50,000 |
| Growth in the months with no campaign, this year against last | 8% |
| Expected revenue in the campaign window without a campaign ($50,000 plus 8%) | $54,000 |
| Estimated added revenue ($62,000 minus $54,000) | $8,000 (about $1.33 per $1 spent) |

In the example, the platform credited the campaign with $15,000, but the whole file grew by about $8,000 more than the trend predicted. Both numbers can be true. Some donors who clicked the ad would have given in December anyway, and the platform counted them. The matched comparison counts only the growth above the trend. The campaign still earned more than it cost, just less than the dashboard said, which is the same shape of result the experiment found. For what it costs to bring in each new donor, see our [donor acquisition cost](https://donorinsights.com/articles/donor-acquisition-cost) guide.

1. Measure revenue from every channel, not only online gifts credited to the ad.
2. Keep counting for five weeks after the campaign ends, the window over which the experiment found effects continued.
3. Compare against a matched period, adjusted for your growth in months with no campaign, or stagger the start date across regions so every donor is still asked.
4. Report the estimated added revenue next to the platform's figure, and treat the gap as information.
5. Repeat the comparison across several campaigns before drawing firm conclusions about a channel.

> **How current are these numbers?**
>
> The experiment ran in November 2017 in Germany and was published in 2025, so treat its effect sizes as directional for a U.S. organization today, and trust its design lesson more than its exact figures. The point estimates quoted here come from the February 2024 working-paper revision. The published abstract reports the same conclusions. The M+R figures describe 2025 advertising at its 180 participating organizations.

The comparison in step three needs revenue from every channel, by week and by region, in one place. Donor Insights reads your contacts and gifts, so a campaign's window can be set against a matched period and your growth trend without exporting spreadsheets from three systems. The [methodology](https://donorinsights.com/methodology) starts from your own giving records, and the [platform](https://donorinsights.com/platform) shows whether the donors a campaign brought in are still giving a year later, which is where the long-run return in the experiment came from.

## FAQ

**What is a good fundraising ROI for online ads?**

M+R Benchmarks 2026 reports that paid search returned $2.48 per dollar spent in 2025 and display ads $1.11, as revenue credited to ads. Those figures compare channels with each other. The more important question is how much of the credited revenue was new, which requires comparing all-channel revenue against what would have arrived without the campaign.

**What is the difference between ROAS and incremental return?**

ROAS, or return on ad spend, is the revenue a tracking system credits to ads divided by what the ads cost. Incremental return is the extra revenue the campaign caused, compared with what would have come in anyway. Credited revenue can include gifts donors would have made regardless, and it can miss gifts made by mail or phone.

**Did the Facebook fundraising experiment find that ads work?**

Yes, with caveats. The randomized campaign raised donation revenue and frequency during the campaign and for five weeks after, and it was profitable at the point estimate. The confidence interval was wide, and some of the new giving came from similar charities rather than being new to the sector.

**How can a small nonprofit measure whether ads raised new money?**

Compare total revenue from every channel during the campaign and five weeks after with the same weeks in an earlier year, adjusted for your growth in months without a campaign. Or start the campaign in some regions a few weeks before others. Neither method withholds an ask from anyone, and both give a direction rather than proof.

**Why are click-through rates a poor measure of fundraising success?**

Clicks are what an ad platform can see, but donations are the outcome that matters, and the two can diverge. The Management Science experiment concluded that click data may be an inappropriate proxy for donations and recommended designs that measure giving directly.

## Sources
- [Adena, M., & Hager, A. (2025). Does Online Fundraising Increase Charitable Giving? A Nationwide Field Experiment on Facebook. Management Science, 71(4), 3216-3231. DOI 10.1287/mnsc.2020.00596.](https://doi.org/10.1287/mnsc.2020.00596)
- [Adena, M., & Hager, A. (2024). Does online fundraising increase charitable giving? A nationwide field experiment on Facebook. WZB Discussion Paper SP II 2020-302r2, second revision February 2024 (design details and point estimates).](https://bibliothek.wzb.eu/pdf/2024/ii20-302r2.pdf)
- [M+R (2026). M+R Benchmarks 2026: Advertising. 2025 data.](https://mrbenchmarks.com/advertising/)
- [M+R (2026). M+R Benchmarks 2026: Glossary (view-through revenue).](https://mrbenchmarks.com/glossary/)

## Disclaimer

**Educational purpose.** This article is published for general informational and educational purposes only. It is not investment, financial, funding, donor, tax, legal, accounting, or fundraising advice, and reading it creates no advisory, fiduciary, or client relationship. Consult your own qualified professionals before making decisions.

**Not an offer.** Nothing here is an offer, solicitation, or recommendation to buy, sell, donate to, or fund any organization or security. Pray, Inc. DonorInsights.com is not a registered investment adviser, broker-dealer, law firm, accounting firm, or fundraising counsel, and publishes only impersonal commentary of general and regular circulation.

**Public data, no guarantee.** Analyses of named organizations rely on public sources such as IRS Form 990 filings, an organization's own published materials, and reputable press. These sources are believed reliable but are not audited or independently verified by us. The content may contain errors or omissions and is provided "as is" with no warranties of accuracy, completeness, or currency.

**Opinions and estimates.** Donor-level and financial figures for named organizations are our inferences and opinions, including forward-looking projections that are not guarantees of future results.

**No affiliation.** Named organizations are independent and are not clients, affiliates, sponsors, or endorsers of DonorInsights.com. Third-party names and trademarks belong to their owners and are used only for identification and commentary. Links to third-party sites are not endorsements.

**No reliance.** We accept no liability for any action taken based on this content.

**Corrections.** This analysis reflects public sources available as of publication. If you are an organization discussed here, or any reader, and you believe a figure or statement is inaccurate, email hello@donorinsights.com and we will review it and correct any error promptly.

Your use of this site is governed by our [Terms of Use](https://donorinsights.com/terms).

Source: DonorInsights.com — https://donorinsights.com/articles/online-fundraising-roi