Please note that this methodology was used starting with the Q1 2026 report, you can read more about why we implemented these changes here
Panel Methodology
FEP estimates trends in individual giving from a panel of US nonprofits from a dataset contributed by several donor-management and online-fundraising software providers. Because the organizations in the data are not a random sample of the sector, and because the mix of organizations shifts over time for reasons unrelated to giving (platforms gaining and losing clients, organizations migrating between systems, new adopters uploading years of historical records at once), the panel is defined to isolate organizations whose reporting is most likely to reflect real giving activity rather than these platform-driven movements.
An organization is included in a given quarter’s panel if it meets all three of the following criteria:
The panel is rebuilt every quarter. The 24-month reporting-consistency requirement identifies organizations that are actively using their systems, removes organizations that have stopped reporting, and, because the panel is resampled each quarter, prevents shifts in panel composition from accumulating over time. The 90-day upload-lag requirement restricts the panel to organizations whose data arrives promptly and predictably, which is what allows the late-data adjustment described below to work. The $25 million ceiling limits the influence of a small number of very large organizations, whose individual campaigns and system changes can move headline figures without reflecting a broader trend.
Weighting
FEP weights the panel by cause area, benchmarked against 2023 IRS Form 990 filings. Weighting adjusts the panel so that its composition better reflects the wider population of nonprofits. Where a group is over- or under-represented in the panel relative to the sector, each organization in that group is counted at more or less than its face value when sector-wide figures are calculated. Weighting by cause area corrects for one known imbalance. It does not make the panel fully representative of the sector.
Late Data Estimation
Donations are frequently recorded some time 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.
This creates an asymmetry in year-over-year comparisons. At the time a report is written, the prior (baseline) year has had a full additional year in which its late transactions could arrive, while the current (reporting) year has not. Left unadjusted, the comparison understates growth.
To correct for this, FEP applies an equal reporting window to both years of every comparison: the same upload cutoff date is applied to the baseline year as to the reporting year, so both are measured under identical conditions.
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.
We can break down the top-line estimations of error into two components to model it in more detail. The first is data drift: fundraising data that belongs to a particular time period but is reported later. This is the largest and most important component to capture. The second is methodology error: variability as we tweak our methods. The data in the FEP report is processed to account for outliers and bad data. This preprocessing introduces an element of variance which will always exist, even if we do no preprocessing. If we just used raw data in our analysis, we would have very high methodology error due to noise and outliers.
For more information about the methodology or any past changes, please consult our methodology improvements page here. or request more information about the methodology via email at [email protected].