Week 8: The 990 Lab

This week’s big question

What can a real Form 990 tell you that an AI summary can’t, and what does a second filing tell you about the first? Lab week. Monday you read your organization’s filing against a second one and build A3; Wednesday you check a chatbot’s summary of a comparable against its filing and start B2. Module 4 · builds LO5 (financial-document literacy) · thread: data/evidence.

Readings

Session 1 · Mon Oct 12: The 990 Lab

A Rail A lab: you build the A3 990 comparison, your organization’s Form 990 read against a second organization’s from a different subsector. First, Gregory and Howard and the Overhead Myth letter on why an overhead share needs care: it reflects reporting choices as well as spending.

After the reading, we compare a food bank and a theater from their own 2023 Forms 990, which cover different months: Hoosier Hills Food Bank’s (January to December 2023) and Constellation Stage & Screen’s (July 2023 to June 2024). Discuss these questions with a neighbour and write brief notes on sheet 1:

1. Money in, money out: where does each organization’s money come from, and what does it spend the most on?
2. The overhead share: the food bank reports 6 percent of its expenses for management and fundraising; the theater, 30. Is the food bank more efficient? Use Gregory and Howard and one line from the filings.
3. Your filing: find Part IX, line 25 on your organization’s 990. What share goes to management and fundraising, and what would you check before calling it high or low? If your filing cannot show it, say why.

After eight minutes, we compare your answers with our reading of the two filings and with a model contrast sentence. An overhead share describes how one year’s costs were reported, not how well the programs work.

Then the lab. On sheet 2 you fill A3’s four rows for your own pair (size and revenue mix, where the money goes, people, and the failure theory each money pattern suggests), with each filing’s tax year copied from page 1, item A. On sheet 3 you write the sharpest contrast, with a number and its line. A neighbour checks every figure for a page, both tax years, and the contrast for an efficiency claim the filings cannot support. Volunteers read their contrasts aloud at the end.

Keep all three worksheets: sheets 2 and 3 are the draft of Sunday’s upload. Hand in an individual exit slip with your name, both organizations and tax years, and one contrast supported by a number and its filing location. Bring your organization’s 990 with the PDF open and a candidate comparison organization.

Session 2 · Wed Oct 14: Funding-model sketch studio

A Rail B studio, and the start of the B2 Funding lab. Each teammate claims a different real comparable: an organization doing your startup’s kind of work that files a full Form 990 or 990-EZ. You ask a chatbot to summarize its finances, freeze the answer unedited, and check it against the filing with the 990 Lab method from Week 3. Your team then sketches an illustrative first-year funding mix for its new nonprofit from the lines you verified. Three questions run through the session:

1. The claim: what exactly does it say, and where on the filing could you check it?
2. The verdict: correct, wrong, or unverifiable? Give the part, line, and page, and for a wrong claim, the filing’s figure.
3. Year one: which verified line supports each share, and what would a new organization keep or change, and why? If a source on the comparable’s filing does not fit your startup, say why you leave it out.

We run all three on one real case first: Community Kitchen of Monroe County. With a neighbour, you check four sample claims about it against its 2024 Form 990. Then we set its own account of how it began in 1983 beside that filing, and I show a year-one sketch for a hypothetical meal startup. A filing shows what a comparable raised, not what a new organization can raise.

On sheet 1 you put the first two questions to your own comparable: the prompt, and four verified rows, revenue claims first because the sketch needs them; a teammate then checks every row for a verdict and a location. On sheet 2 your team puts the third question to its startup: each share with its comparable line and what changes, the assumption behind the least-sure share, and the funder who could stop saying yes. Keep both worksheets and share the team sketch with every teammate: B2 is built from them, with rows five to eight due in the template by Sun Oct 25. Hand in an individual exit slip with your name, your comparable, and one AI claim’s verdict and exact filing location. Bring a candidate comparable with its EIN (your own or a teammate’s site-visit organization counts if it fits your concept) and your team’s photo of Week 7’s sheet 1, where the sketch starts.

What’s due / prep

• Rail A · A3, 990 comparison (15 pts, credit) due Sun Oct 18: your org and a comparison org from a different subsector, side by side on four rows, every figure cited to part, line, and page, both tax years copied from page 1, item A, plus the sharpest-contrast sentence. Built in Monday’s session; the filing rules are on the 990 Lab how-to. Visit still ahead? It has to happen before A4 is due on Sun Nov 15. A3 does not wait for it: both filings are public.
• Monday’s comparison org: no candidate yet? Any of these works if its subsector differs from yours: Bloomington Developmental Learning Center (EIN 35-1784455), WFHB (31-0935271), Boys & Girls Club of Bloomington (35-0997525), or Stone Belt Arc (35-1059827).
• Wednesday’s studio starts the B2 Funding lab (10 pts, credit, due Sun Oct 25): each teammate claims a different real comparable org to verify, and the team drafts its funding-model sketch.

Looking ahead

• The B2 Funding lab is due Sun Oct 25. On Mon Oct 19, bring your organization’s 990 and your A3 table to the subsector tour and NTEE introduction. The TA-led NTEE hunt and B2 clinic on Wed Oct 21 check your comparable and evidence. Bring its downloaded filing and a working draft with at least eight claims numbered, even if some rows still need checking. Keep a usable comparable; replace it only if it fails the assignment’s tests or cannot support eight checkable claims.