How to Run a Five-Year Grantmaking Retrospective

Most foundations can describe their strategy in a sentence and their last grant round in detail. Far fewer can say what five years of grants actually did — where the money landed, which relationships deepened, which ones quietly ended, and whether the pattern matches what the board approved.

A five-year retrospective answers that. It is a different exercise from routine portfolio analysis, which looks at the current shape of grantmaking as an annual discipline. A retrospective looks backward across a longer window, usually because something prompts it: a strategy refresh, a leadership change, an anniversary, a board that has started asking harder questions.

The analysis is rarely the hard part. The hard part is the record. Five years of grants live across a grants system that was probably replaced at least once, spreadsheets maintained by people who have left, and PDF reports in formats that changed three times. Program areas were renamed. A grantee merged. Two records say "Riverside Youth Center" and "Riverside Youth Ctr." and nobody has ever reconciled them.

That reconstruction work is the project. Plan for it, and the analysis becomes straightforward.

What a retrospective asks that an annual review does not

Annual portfolio analysis asks what the current shape is. A retrospective asks what changed, and whether the change was decided or drifted.

That difference matters because most of the interesting findings in a long window are about movement rather than position. A single year shows concentration. Five years show whether concentration is increasing. A single year shows a cause mix. Five years show a program area that grew 40% without anyone proposing that it should.

Six questions structure most retrospectives well:

1. How did our grantmaking change? Volume, average grant size, cause mix, and program area over time.
2. Where did the money actually go? Geographic distribution of grants, not of grantee head offices.
3. How did conditions in those places change? The context the funding sat inside, over the same period.
4. What did other funders do over the same period? The external comparison that turns your numbers into a signal.
5. What happened to organizations we backed repeatedly? Grantee trajectories after your grants, stated as association rather than cause.
6. What are the three or four conclusions worth putting to the board? The reduction step, which is where most reviews fail by producing forty charts instead.

Questions two, four and five are answerable from public data. Questions one and six need your own records and your own judgment. That split determines how the project is staffed.

Why the historical record is the hard part

Three specific problems account for most of the effort, and none of them are analytical.

Grantee identity. The same organization appears under different names, before and after a rebrand, with and without "Inc.", occasionally under a fiscal sponsor. Any analysis of repeat funding or relationship length is wrong until this is resolved. Attaching an EIN to every historical grant is the single highest-value cleanup step, because the EIN is what lets your internal record join to any external dataset.

Changed taxonomy. Program areas get renamed and restructured. A grant coded "Community Health" in 2021 might have been "Health & Wellbeing" in 2019 and split across two areas in 2023. A retrospective that reports on today's categories will show apparent shifts that are really just recoding. The fix is a mapping table from historical codes to current ones, built deliberately and documented, so that the board can see which changes are real.

Unstructured reporting. Grantee reports from five years ago are PDFs and Word documents in formats that changed repeatedly, often with no consistent fields. This is where the themes, outcomes and challenges live — the qualitative half of any review worth reading — and it is the reason retrospectives get quoted as expensive consulting projects.

None of this is glamorous. All of it determines whether the findings survive scrutiny.

What the public record can and cannot tell you

Public IRS data is excellent for the comparison layer and limited for your own history.

For your own grants, your internal records are authoritative. They have the grant purpose as written, the program area as coded, the payment schedule, and the reports. Form 990-PF shows what you disclosed, not what you knew.

For everyone else's grants, the public record is the only option, and it has a boundary worth understanding. The Taxpayer First Act, enacted July 1, 2019, made electronic filing mandatory for Form 990 and Form 990-PF — but for tax years ending July 31, 2020 and later (Baker Tilly). Before that, e-filing was voluntary. Filings from earlier years do exist in machine-readable form, but the sample skews toward larger organizations and those that chose to file electronically.

The practical consequence: a claim like "philanthropy in our region shifted toward housing between 2016 and 2019" drawn from pre-mandate e-file data is describing the funders who happened to e-file, not the field. State the limitation next to the number, or restrict the comparison to years where coverage is complete. There is also a lag of roughly 12 to 24 months between a fiscal year ending and its filing becoming publicly available, which trims the recent end of any window.

The division is clean enough to plan around: bring your own history, use the public record as the benchmark.

Getting geography right, not just present

Most grant datasets place a grant at the grantee's registered address. For a place-based funder, that is frequently the wrong answer.

A nonprofit headquartered in a city center may deliver most of its services in three towns forty miles away. Mapped by head office, that funding appears as city investment. Mapped by delivery, it tells a different story — often the story the board actually wants, because it shows reach into places the foundation says it prioritizes.

This is not a minor refinement. In a review whose whole purpose is "did our dollars go where we said they would," the difference between registered address and service area can invert the conclusion. If you hold service-area information for your grantees — many funders do, in application forms or program officer notes — it should be used in preference to any address-derived geography. If you do not, the honest presentation is to label the map as grantee locations rather than service locations and say so on the slide.

Comparing conditions over time without overclaiming

A real retrospective compares two moving things: your grantmaking, and the conditions it sat inside. Comparing five years of grants against today's poverty map is a common shortcut and a misleading one.

Public indicators are published in vintages. American Community Survey five-year estimates, for instance, are released for each period, so 2016-era conditions can be compared with recent ones — at county level straightforwardly, because county boundaries are stable. Below that, care is required: the 2020 Census redefined census tract boundaries, so tract-level figures from before and after do not line up without a crosswalk. Several health and housing indicator sets also have methodology changes that break comparability across the same window.

Two rules keep this defensible:

• Match the vintage to the year. 2017 grants belong beside 2017-era conditions.
• Never state the foundation caused the change. Funding and outcomes can be shown side by side, and the association can be described. Attribution requires a counterfactual that grant records cannot supply. A board paper that says "we invested here and conditions improved" without that caveat is the finding most likely to be challenged, and rightly.

The external layer: what the portfolio looks like from outside

The comparison layer is what turns internal numbers into meaning. Three views do most of the work.

Co-funders. For each grantee, who else was funding them across the period? This surfaces the organizations that depend heavily on you, the ones where you are one of twenty funders, and the co-funder networks you sit inside without having decided to.

Peer funders. What did comparable foundations in your footprint fund over the same years? Benchmarking against peers answers the question a board always asks — is this pattern normal? — and it is only answerable from public filings.

Insist on a specific peer set. A benchmark against "foundations nationally" is not usable: it averages across asset classes, regions and program areas that have nothing to do with your situation, and any board member can say so in one sentence. Agree a named group of roughly 10 to 20 genuinely comparable foundations, selected on four axes — assets and annual grantmaking level, geography, program areas, and foundation type — and agree it before the analysis runs rather than after the numbers arrive. A peer set chosen after the fact invites the suspicion that it was chosen to flatter.

Grantee trajectories. What happened to the organizations you backed repeatedly? Their subsequent filings show revenue paths, growth, contraction, and reliance on government funding. Presented as association rather than attribution, this is often the most striking section of a retrospective: the organizations you funded early, before they were established, are visible in a way no internal system records.

One derived figure is worth calculating specifically. The share of each year's grants going to organizations you had never funded before is a direct measure of whether a portfolio is renewing or recycling. If that share stays very low across five years, the process is reproducing an existing network regardless of how carefully each individual decision was made.

Structuring the review as three conversations

A retrospective that is commissioned, executed and then presented once tends to arrive at a board meeting as a finished object nobody can interrogate. Three conversations, placed deliberately, produce a better result — whether the work is done internally, by a consultancy, or by a data provider.

Before the analysis: taxonomy and hypotheses. Agree what the categories actually mean. "Education", "systems change", "underserved" and "capacity building" all sound self-evident and are understood differently by any two people in the room. This is also the moment to say out loud what the board is genuinely wrestling with, so the review answers the live question rather than a generic one.

Before the report is finalized: challenge the preliminary findings. This is the most valuable of the three and the one most often skipped. It is the meeting where someone says "that looks like a geographic shift, but our strategy changed in 2023," or "those ten grants were one initiative and shouldn't be counted as ten independent decisions." Data-derived findings either survive that conversation or get corrected by it, and a finding that has survived it is much harder to dismiss later.

With the board: implications, not charts. The final session should be facilitated rather than presented — "given these patterns, what might we want to preserve, investigate, or change?" The analysis is the input to that discussion, not its conclusion.

Reducing it to findings a board can act on

The failure mode of retrospectives is volume. Forty charts, no argument, and a board that thanks the team and changes nothing.

Aim for three or four findings. Each one needs three properties:

• Specific. "Our top five grantees received 48% of grant dollars over five years, up from 31%" rather than "concentration has increased."
• Traceable. Every figure should drill through to the underlying grant records and filings. A finding that cannot be traced back to its evidence will be argued about rather than acted on.
• Framed as a question. "Is this the concentration we intended?" invites a decision. "We are too concentrated" invites a defense.

Include the decisions to change nothing. A board that consciously decides to remain concentrated has made a strategy; one that never examined it has drifted.

Where software fits, and where it does not

The analysis layer is well served. Once grant records are clean and EIN-matched, most of the six questions are queries, and the external comparison layer is available from public filings.

Plinth's US database covers approximately 205,000 grantmaking foundations and 17.9 million grants built from IRS Form 990 and 990-PF filings, which supplies the benchmark half of a retrospective directly: co-funders for each grantee, peer funders in your footprint, grantee financial trajectories across multiple filings, and geographic distribution against public need indicators. Portfolio insights covers the internal half, reading grantee reporting alongside grant records so that recurring themes across many reports surface without anyone reading forty documents individually.

Two limits are worth stating plainly, because they bear on how a retrospective should be scoped. Nothing here rates or ranks grantees, and nothing attributes outcomes to the funder — the analysis is a lens on your own decisions, not a verdict on the organizations you funded.

And the reconstruction stage described earlier is genuinely work, not a feature. Bulk import handles structured records — grant lists, spreadsheets, exports from a previous system — reliably. Five years of inconsistent PDF and Word reports is a different matter, and it is handled as a piece of work done alongside the foundation rather than an upload. This is why Plinth runs retrospectives as scoped one-off projects: bring a program area, the grant history and the reports in whatever state they exist, and the reconciliation, extraction and comparison are worked through together, ending in findings that trace back to original documents. For foundations wanting to test the approach before committing to a full review, a single focus area and one year of reports is enough to see what the method produces.

Frequently asked questions

How long does a five-year retrospective take?

The analysis is fast once the data is clean. Budget the majority of the timeline for reconstructing and reconciling the historical record — grantee identity, program taxonomy mapping, and extracting content from old reports.

Why five years rather than ten?

Five years is long enough for patterns to separate from noise and short enough that current strategy is still recognizable. Ten years reaches back past the point where most foundations' program structures — and the public e-filing record — are consistent enough to compare cleanly.

Can we do this from public 990-PF data alone?

Not for your own history. Form 990-PF shows recipients and amounts as disclosed, without grant purpose as you wrote it, program coding, or any reporting. Public data is the right source for the comparison layer, not for your own record.

What is the single most valuable cleanup step?

Attaching an EIN to every historical grant. It resolves duplicate grantee names and is what allows internal records to join to any external dataset.

How do we handle program areas that were renamed?

Build an explicit mapping from historical codes to current ones and document it. Without that, real strategic shifts and simple recoding look identical in the output.

Should the review compare us to peer foundations?

Yes, and it is one of the few questions only public data can answer. Peer comparison establishes whether a pattern is distinctive or standard for funders of similar size and focus in the same region.

What should a five-year retrospective cost, and what should be included?

Cost varies with portfolio size and, more than anything, with the state of the historical records. Judge a proposal on scope rather than headline price: the full period rather than a sample, manual review of classifications and anomalies, an agreed peer group for benchmarking, access to the categorized underlying data and not just the report, at least one revision cycle before findings are final, and no requirement to buy an ongoing subscription alongside it.

What is the most revealing question to ask a provider?

Ask who specifically will do the analysis and roughly how much senior time is included. The answer distinguishes a genuine analytical engagement from software output with a report attached, and it is difficult to answer vaguely.

Can a retrospective show that our funding caused improvements?

No. Grant records support association, not attribution — establishing cause requires a counterfactual that filings cannot provide. Present funding and conditions side by side, and describe the relationship without claiming it.

Recommended next pages

• Foundation Portfolio Analysis — the annual discipline this project version builds on
• How to Benchmark Your Foundation — the peer comparison layer in detail
• Co-Funder Networks — who else funds your grantees
• What Your Form 990-PF Reveals — the public version of your own record
• How to Research a Foundation — the same data, used to look outward

Last updated: August 2026