Chapter 8 of 13 · Funded partner road · About 15 minutes

Capture each funder's taste, and your own

Two funders can read the same accurate report and come away with opposite impressions. The difference is rarely the numbers; it is whether the report was written the way that funder reads.

On this page: What taste is · Funder taste file · Your taste file · Where it lives · Read it as the funder · Never the numbers · Try it · Questions

What is funder taste in reporting?

In short: funder taste is how a specific funder reads: what they look at first, the words they use, the proof they trust and what makes them doubt. Your own taste is how your organization speaks. Both are judgment, not decoration.

The idea comes from marketing. In Loop, HubSpot's Kipp Bodnar and Kieran Flanagan describe a Taste Profile with two halves, customer taste and brand taste, that tells AI how to reach people. This chapter adapts that idea to reporting. The customer is the funder who reads your report. The brand is your organization's own voice and judgment.

It matters more now that AI writes most first drafts. Without context, every draft comes out in the same generic voice and reads right to nobody. A taste file tells the AI which number goes first, how to name a miss and which word will make this program officer wince.

Keep one distinction clear. The requirements map holds what a funder requires. The taste file holds how they prefer to read it. When the two disagree, the agreement wins.

What goes in a funder taste file?

In short: five parts, a few lines each, and every line written from something the funder actually said or published, not from a guess.

The richest source is the questions a funder sent back on your last reports. Each one marks a place where the report did not read the way they needed. Copy them in word for word, with the date.

What goes in your own taste file?

In short: five parts that keep every report sounding like your organization: why you exist, how the program works, what you learned, what participants say and the words you never use.

Write it once and reuse it in every funder's project, so your reports do not sound like four different organizations. Draw on your theory of change, team debriefs, surveys and the claims your leadership has decided not to make.

The "what we learned" part is the one teams skip, and the one funders value most. A report that names a miss and what changed reads as a partner who is paying attention.

Where should the taste files live?

In short: in one AI project per funder, such as a Claude Project or a ChatGPT Project, beside the agreement, the data dictionary and past reports with the funder's feedback.

One project per funder. Every draft for that funder starts here.

Sopact's founder works this way. He keeps a Claude project for each prospect, holding call transcripts, their published material and a short note on how they think, and uses it whenever he writes back to them or builds an article for them.

Keep participant records out of the project. The taste file describes a funder and a voice; the numbers come from your records at drafting time, through a source you have chosen.

Build a funder taste file for [FUNDER NAME] from the material attached:
- the onboarding call transcript: [ATTACH]
- the questions and comments they sent on our past reports: [ATTACH]
- their published strategy and grantee reporting guidance: [ATTACH OR LINK]
- recent emails from their program officer: [PASTE]

Write five short sections:
1. How they read (when, how many reports, what they look at first)
2. Their words (terms they use, terms they avoid)
3. Proof that moves them (evidence they trust or ask for)
4. What loses trust (what made them doubt, question or follow up)
5. Format and rhythm (length, schedule, template, codes)

Rules: every line must cite the document and quote the words it comes from. If a section has no evidence, write "not known yet" and suggest one question we could ask. Do not guess from what funders in general prefer. Keep the whole file under one page.

How do you read a draft as the funder first?

In short: before you send, ask the AI to read the draft as that funder's program officer and say bluntly what lands, what is missing, which sentence loses them and what they would ask back.

This is a rehearsal, not a verdict. It catches the report that opens with a story when this reader wants the table. It will not catch a wrong number; that check comes in the next chapter.

You are the program officer at [FUNDER NAME]. Use the funder taste file in this project to read the attached draft report the way you would: [DESCRIBE, e.g. one of 40 reports the week before the board].

Be blunt. Tell me:
1. What lands: the parts you would repeat to your board.
2. What is missing that you expected to see.
3. The one sentence where you stopped trusting or stopped reading, and why.
4. Every word or claim that does not sound like us, using our taste file.
5. The questions you would send back after reading it.

Do not rewrite or change any number. If a number looks wrong, undefined or unsupported, list it as a question instead.

Then close the loop. When the real questions and reactions arrive, add them to the funder's taste file with the date. After two or three cycles, the rehearsal gets close to the real thing.

Can taste change the numbers?

In short: never. Taste changes words, emphasis and format. The counts, definitions and periods come from the agreement and the dictionary, and they are the same in every report.

If a funder's taste seems to ask for a different number, such as a longer placement window, that is not taste. It belongs in the requirements map, reported in their definition with a note on how it differs.

The method also works upward. A portfolio manager can keep a taste file for their own board or donors before writing the portfolio report. And a funder can share a short "how we read reports" note with partners at onboarding, which saves everyone a year of guessing.

Sopact has helped teams such as Open Play Foundation capture this kind of funder and organization context. A context layer in Sopact that applies definitions and taste as the AI Assistant writes is coming soon.

An honest limit: a taste file is a set of notes about people, and people change. A new program officer may read differently, so treat the file as a draft to confirm, not a rule. And the AI's reading as the funder is an approximation; one real question from the funder outweighs it.