A clearer picture of giving: the Q1 2026 report is generated using the first major update to FEP methodology in five years
The Fundraising Effectiveness Project (FEP), a joint initiative of GivingTuesday and the Association of Fundraising Professionals (AFP), is making the first major update to its methodology since 2021. This note explains what is changing, why, and what it means for the numbers you rely on.
The 2021 methodology served the sector well, and the trends it reported have proven directionally reliable. But our dataset has grown, our understanding of its quirks has deepened, and your feedback has pointed us toward places we could do better. The changes below are refinements rather than reversals, designed to make FEP’s estimates more accurate, transparent, and consistent.
What is FEP, and what is it for?
FEP estimates trends in individual giving using a large sample of US nonprofits, tracking changes in dollars, donors, and retention. Each metric is expressed as a year-over-year, year-to-date comparison: for example, the Q2 2024 report compares performance over January–June 2023 to performance over January–June 2024.
We draw on the largest donation-transaction datasets available. Several donor-management and online-fundraising software providers contribute records, each one representing a single gift made to a nonprofit on their platform. That scale and timeliness allows FEP to offer an early read on giving trends months before tax filings or survey-based estimates become available.
FEP does not try to size the giving market. Other approaches, such as Giving USA’s, estimate the total dollar value of US giving and how much it grows each year, and are built to be more representative of the sector for that purpose. FEP’s role is different and complementary: to track trends during and throughout the year, to follow donor growth and retention alongside dollars, and to break trends down by donor characteristics such as gift size, frequency, and lifecycle. This combination is what FEP is built to provide, and what our methodology has to support.
The challenge of measuring giving from transaction data
FEP shares a challenge common to any statistic built from administrative records rather than a designed survey: the organizations in the data are not a random sample of the sector, and the makeup of that sample shifts over time for reasons unrelated to giving trends.
Platforms gain and lose clients. Organizations migrate from one system to another. New adopters come on board and upload years of historical records all at once. Left unaddressed, these platform-driven movements can look like market signals, making it appear that giving rose or fell when all that really changed was who was in the dataset that quarter.
Our methodology has always existed to separate the genuine signal from these artifacts. We do this in three ways: we define a panel of eligible organizations whose activity is most likely to reflect real giving rather than platform churn; we apply a late-data adjustment to account for transactions that keep arriving after a reporting period closes; and we weight the panel to correct for known imbalances between it and the wider sector. Each of these areas is being improved.
What is changing, at a glance:
Previous Methodology
Locked at the start of each year; could include organizations that stop reporting during the time period being analyzed.
Three criteria:
In each of the three previous years, the organization reported at least 25 donations worth at least $5,000 in total.
In each of the three previous years, the organization’s dollar growth was no more than 300% and no less than -67% each year.
In each of the three previous years, the organization raised under $25 million.
New Methodology
Rebuilt every quarter, excludes organizations that stop reporting during the time period being analyzed.
Three criteria:
The organization has reported at least one transaction in each of the previous 24 months up until the end of the reporting period.
For each of those 24 months, the median lag between gift date and entry date is under 90 days.
The organization reported under $25 million in total fundraising in each of the baseline and reporting years.
Previous Methodology
A fixed percentage point “uplift” added to the headline figures each quarter.
New Methodology
An equal reporting window is applied to both years of every year-over-year comparison, so they are measured on the same footing
Previous Methodology
By organization size (fundraising volume) and cause area.
New Methodology
By cause area only.
Previous Methodology
Aggregate growth within fundraising-size bands
New Methodology
The typical (median) growth within each band, plus a new view by donor-base size
A panel built on consistent reporting
Previously, we built the panel once a year and defined a “stable” organization based on its outcomes, keeping those whose year-on-year donation totals stayed within set bounds for three years. After observing the effects of this across multiple reporting cycles, we have learned that this approach has two related drawbacks.
First, it lets events from several years earlier shape the panel used for a current report. The panel for the 2025 reports, for example, was screened on each organization’s results for 2022, 2023 and 2024 (and implicitly 2021, to calculate the growth rate in 2022). A single unusual year could push an organization’s growth outside the permitted range (no more than +300%, no less than −67%) and keep it out of the panel for several reporting years, long after that event had stopped reflecting how giving was moving today.
Second, past fundraising activity was not a foolproof filter for continued, consistent CRM usage. We observed that, among organizations who met the previous panel’s growth criteria, a minority still churned or stopped reporting during the year being measured. These organizations would then disproportionately impact the topline growth figures
The new approach defines stability by behavior instead: an organization that uploads transactions reliably, every month, is a more reliable signal of current fundraising activity. To qualify for a given quarter’s panel, an organization now needs to have reported at least one transaction in every one of the preceding 24 months, with most of each month’s gifts uploaded promptly, and to fall under an annual $25M ceiling.
The consistent and prompt reporting criteria do four things. First, it favors organizations actively using their donor-management systems, so the data reflects real activity rather than sporadic or backfilled record-keeping. Second, it removes “dropouts” (organizations that stop reporting), which are largely a platform artifact rather than a real market event. Third, by rebuilding the panel every quarter, we prevent panel composition “decay” from impacting our metrics. Finally, the 24-month window is the minimum needed for a full Q4 year-over-year comparison, so it keeps the restriction placed on the sample as light as possible. Together, testing showed these conditions produce a more consistent distribution of organization-level growth rates, with fewer extremes in either direction, than the old sampling method.
We cap the panel at organizations raising under $25M a year for the following reasons. Proportionally speaking, there are relatively few organizations with very large fundraising volumes in the data, so the sector above this threshold is not well represented, and because there are so few of them, they can disproportionately drive the topline results, particularly for dollars. That concentration is a problem for two reasons. A single large organization’s idiosyncratic year, having an unusually large campaign, for example, can move the headline figure without reflecting any broader trend. And it compounds the platform-driven effects noted earlier: if one of these organizations changes the tools it uses or imports a batch of historical transactions, that platform activity can be mistaken for a real shift in the market. The cap does mean we set aside some legitimate activity at the top end, and as the sample grows we may be able to raise it.
The following diagram provides a visual representation of how the panel criteria are applied under the old and new methodology for the 2026 reporting year. Under the old methodology, the panel is defined based on outcomes in 2023, 2024, and 2025 (represented by the light blue blocks). This panel is then set in stone, and used to measure quarterly, year-to-date changes in 2026 compared to 2025. Under the new methodology, the panel is resampled each quarter based on consistent reporting across a 24-month period (represented by the 24-month green blocks, which roll into the end of each reporting quarter).
A more thorough solution to late data
There are many reasons donations can be recorded after they are made. Some of the platforms we draw on are donor-management (CRM) systems, and a donation only enters the system once it has been keyed in. Offline donations such as checks, cash, or donations taken at an event are often entered in batches days or weeks later, once they have been reconciled against bank deposits. Donations taken on a separate payment platform may only reach the CRM when the two systems next sync. Because each is dated to when the gift was actually made, it surfaces in the data as a late arrival.
Our data providers upload a new batch of transactions every month. Each upload contains transactions entered into their systems during the previous month, so an upload on 1 May 2026 would primarily contain transactions with an entry date in April 2026. Most of these will be gifts made in April, but will also include a tail of transactions entered late: gifts made in earlier months but only recorded in the system in April.
This creates an asymmetry in year-over-year comparisons. Transactions from the prior year have had an additional year’s worth of uploads in which to appear in the data, meaning the baseline period is more complete than the reporting period at the time a report is written. Left unadjusted, a year-over-year comparison understates growth.
The old method handled this with a fixed percentage uplift each quarter, plus an error range derived from historical adjustments – a reasonable estimate drawn from past patterns, but a blunt one. It was applied only to the headline totals, which meant the topline and the breakdowns beneath it (by gift size, donor type, and so on) no longer added up to each other.
The new method is simpler. We apply the same upload cutoff date to both years of a comparison, so each is measured under identical conditions: an equal reporting window. Because this is applied to individual transactions, every breakdown is adjusted automatically and consistently. The headline figures and the numbers underneath now always reconcile, and because the correction is built into the data itself, we no longer manually adjust the figures or apply the error bar ranges.
We can implement this approach now thanks to the work of our data providers, who now supply an explicit ‘entry date’ for each transaction. Previously this had to be inferred from the date transactions were uploaded, which could be unreliable. With the entry date recorded directly, we can adjust for late data reliably going forward.
The following diagram represents this visually. Every dot in this diagram is a Q1 donation, made in January, February or March in both the baseline (2025) and reporting (2026) years. Its position represents the month the donation was actually entered into the system, not when it was given. Most are recorded promptly (January through April), but a thinning tail keeps arriving across the rest of the year as donations are logged late. The equal reporting window fixes this by applying the same May 1 cutoff to both years: only gifts entered before the cutoff are counted, so last year’s late uploads (grey) are set aside and the two years are measured on identical terms.
Weighting on what we can measure well
To adjust for known imbalances in the sample, FEP weights its panel. Weighting is a way of correcting for known differences between our panel and the wider population of nonprofits. If, say, arts organizations make up 5% of nonprofits nationally but 10% of our panel, we give each arts organization in the panel half the weight, counting its dollars and donors at 50%, when we calculate sector-wide figures, so the totals reflect the sector’s makeup rather than our sample’s. Previously we weighted on two characteristics: cause area and organization size (measured by fundraising volume). We are removing the size-based weighting but keeping cause area weighting.
The problem with weighting on fundraising dollar volume is that it is a poor proxy for an organization’s actual size, and is entangled with the very thing we are trying to measure. A small charity that lands one large gift can jump into a higher “size” band for a single year before falling back, so the band can reflect a transient event rather than a durable trait of the organization. And because a prior year’s dollar volume is mechanically linked to that year’s growth rate, smaller-banded organizations tend to show faster growth through arithmetic alone (a mix of small-base effects and regression to the mean). That produces an apparent “smaller organizations grow faster” pattern that is largely a measurement artifact. Weighting on this characteristic therefore risks adding noise rather than correcting bias.
We tested this directly using IRS data, running 1,000 simulations per year from 2020 to 2023. Fundraising-based weighting did not reliably improve estimates: it made the average estimate worse in three of the four years, and worse in 43% of individual simulations overall.
We are keeping cause-area weighting because it satisfies the two conditions that volume weighting does not: it is a stable characteristic of an organization that is not mechanically tied to growth, and IRS Form 990 filings give us a credible external benchmark to weight against. We are also clear about its limits. Correcting for cause-area imbalances does not make the panel fully representative, other characteristics we don’t observe in the data may still be unevenly distributed across it. For transparency, we will start reporting the unweighted topline growth metrics in the report, in addition to the weighted metrics, to give readers more insight into how this part of our methodology affects our calculations.
The following metrics compare the unweighted calculation of our topline metrics, to the metrics produced when weighting by cause area:
Dollars
Unweighted 4.25% vs weighted 4.32%
Donors
Unweighted -0.65% vs weighted -0.82%
Retention
Unweighted -0.0768 p.p. vs weighted -0.0495 p.p.
Redefining “size” as prior-year fundraising volume
Related to the problems described above, we are making two changes to how we report metrics by organization “size.” First, we will stop calling these “size” categories and instead label them for what they are: prior-year fundraising volume. Second, we will report the median growth within each band rather than the aggregate. The median answers a more straightforward question: among organizations that raised a given amount last year, what was the typical growth this year? It is also far less prone to the small-base and regression-to-the-mean effects described above, which could make smaller bands appear to grow faster than larger ones, mechanically, year after year. We are also adding a complementary view by donor-base size, grouping organizations by how many unique donors they had last year, to give another lens on how organizations of different scales are faring.
Looking forward
These changes are not an end in themselves. The previous methodology served the sector for five years and our aim is for this one to be just as durable, and to be a stable foundation that we can confidently build on.
One direction we are actively exploring for a future iteration of this work is benchmarking. We know many readers use the FEP report to make sense of their own performance, asking not just “how did the sector do?” but “how did we do relative to organizations like us?” The report, though, is mainly built to measure broader trends, with some high-level reporting by cross-sections such as size and cause area. As a benchmarking instrument, that makes it fairly blunt.
With a dedicated benchmarking tool, rather than reading headline figures, an organization could select the characteristics that best describe it, such as cause area, prior-year fundraising volume, and donor-base size, and see a cross-section built around organizations that look like theirs, reported against the metrics most relevant to them. A Human Services organization that raised $250,000 last year from around 500 donors, for instance, could compare itself against that specific slice rather than the sector as a whole. Further down the road, we may be able to incorporate characteristics that sit outside the current data altogether, such as an organization’s total revenue or staff size, or geographic region.
The methodology described here will make work like this possible. Taken together, a panel built on consistent reporting, a transaction-level late-data adjustment that extends cleanly to any breakdown, and a weighting approach we can stand behind will, together, give us a firmer footing to grow the project and develop more tools that serve the sector.
Request more information about the methodology via email at [email protected]