report a free-license AI cohort on one page. Print two denominators: everyone who accepted a license on day 1, and everyone who attended the first session. Report certificate completion and work-sample results as two separate lines against both denominators. List the misses by reason. State your small-number privacy rule. That report is shorter than most grant reports, and a funder can check every number in it.
This is general information about reporting practice, not legal, financial or grant-compliance advice. Your funder's agreement and, for WIOA-funded work, your state's guidance control what you must report.
Outcome, time and roles
Outcome: one page a funder can audit line by line, plus a roster file you keep and never send.
Time: about an hour, if you kept a roster during the cohort. Half a day if you didn't.
Roles:
- Program manager: owns the report and signs it.
- License administrator: exports the license and progress data.
- Facilitator: supplies attendance and work-sample scores.
Prerequisites:
- A roster with one row per invited learner.
- The date each license was accepted.
- First-session attendance.
- Certificate status for each learner.
- Rubric scores, if your cohort scores work samples.
Use this template if, do not use it if
Use it if:
- Your learners got a free license, such as Massachusetts residents' no-cost access to the Google AI Professional Certificate through the MassTech AI Hub.
- Your funder asks for completion, credential or skills outcomes.
- You ran a cohort with a defined start date.
Do not use it if:
- Your funder supplies a mandatory form. Use theirs, and attach this page as a supplement.
- You cannot tell which learners attended session 1. Fix your attendance process first.
Why standard reports mislead on free licenses
Free licenses break the usual denominator. When a course costs nothing, many people sign up who never start. Which group you divide by changes the answer more than anything the program did.
Jobs for the Future reported one community-college rollout of the Google IT Support certificate: 4,497 students enrolled, 2,902 "completed at least one course," and 1,325 "completed all five courses required to earn the certificate." Same finishers, two honest rates: 29.5% of enrollees, or 45.7% of those who completed at least one course.
Survivor math is a third, misleading rate. It divides finishers by the people still attending at the end. If most of the people left in the room finish, that rate approaches 100%, whatever happened to everyone else. It is an easy way for a cohort report to flatter itself. No funder can catch it unless the report prints the starting counts.
Money rides on this. In Allegheny County, Pennsylvania, Partner4Work pays training providers half when a participant starts and the remaining half only "upon each participant's completion of the training program or obtainment of an industry-recognized credential," according to a Results for America case study. Providers "are not paid the second payment in the event the participant drops out." Under that model, the report is the invoice.
The template, line by line
Copy this block into your document. Fill in every line. Write "0" rather than leaving a line blank.
COHORT REPORT - [Program name] - [Cohort ID]
Dates: [first session] to [last session] Report date: [date]
Course: [certificate name] Funder: [name] Grant/contract: [number]
A. WHO WE STARTED WITH
A1 Invited (licenses sent) .................... [n]
A2 Day-1 enrollees (license accepted) ......... [n] <- Denominator 1
A3 Session-1 starters (attended session 1) .... [n] <- Denominator 2
B. COMPLETION A - CERTIFICATE EARNED
B1 Certificates earned by [cutoff date] ....... [n]
B2 Rate vs day-1 enrollees (B1 / A2) .......... [%]
B3 Rate vs session-1 starters (B1 / A3) ....... [%]
C. COMPLETION B - WORK SAMPLE PASSED (rubric)
C1 Samples passing ([rule, e.g. 8/12, no zeros]) [n]
C2 Rate vs day-1 enrollees (C1 / A2) .......... [%]
C3 Rate vs session-1 starters (C1 / A3) ....... [%]
C4 Second-scorer agreement on pass/fail ....... [x of y]
D. MISSES (A2 minus learners with B1 or C1)
D1 Never started .............................. [n]
D2 Started, stopped (by reason category) ...... [n each]
D3 Still active, expected finish date ......... [n]
E. NOTES
E1 Small-number rule: counts under [5] shown as "<5"
E2 What we changed mid-cohort: [one line]
E3 Contact for audit: [name, email]Line notes:
- A1 vs A2. Invites sent are not enrollees. A learner who never accepts the license never had a chance to finish.
- A3. This is the denominator for "people we actually served." Print it next to A2, never instead of it.
- B and C stay separate. A certificate shows course completion. A passing work sample shows a learner can apply the skill to a real task. Never add them, and never report "completers" as the larger of the two.
- C4. Report agreement only if two people scored. If one person scored, write "single scorer."
- D. Misses are part of the result. Reasons should be categories ("work schedule," "device," "health/family," "unknown"), not stories.
A worked sample (fictional)
Illustrative example, with invented round numbers:
A1 Invited ............................ 30
A2 Day-1 enrollees .................... 25
A3 Session-1 starters ................. 20
B1 Certificates earned ................ 15
B2 vs day-1 enrollees ................. 60%
B3 vs session-1 starters .............. 75%
C1 Work samples passed ................ 10
C2 vs day-1 enrollees ................. 40%
C3 vs session-1 starters .............. 50%
C4 Pass/fail agreement ................ 9 of 10
D1 Never started ...................... 5
D2 Started, stopped ................... 5 (each reason category "<5")
D3 Still active ....................... 0Suppose 16 learners attended the last session. Survivor math would report 15 of 16, or 94%. The honest page shows 60% and 75%, plus the five who never started. The funder can see both, and can see why they differ.
ed.cafe has not run a cohort yet, so none of these numbers are real. The Finish Ledger method explains the two-denominator rule this template applies.
The small-number privacy rule
Small cohorts can identify people. "1 learner stopped for health reasons" in a group of 12 neighbors is a disclosure.
- Pick a threshold and state it on the page, for example "counts under 5 shown as '<5'."
- Apply it to every reason category in section D, and to any demographic breakdown.
- Check for subtraction. If a total and all but one category are shown, the hidden one can be calculated. Suppress a second cell.
- Keep names off the page. The roster stays with you. Offer the funder an audit visit to see it, if your agreements allow.
Adapting to your funder's format
Funders use their own forms. We have not reviewed Boston's Neighborhood Jobs Trust or Office of Workforce Development forms, or every foundation's template. Map this page onto theirs:
- Find their "enrolled" field. Put A2 there, and note A3 in the comments.
- Find their "completed" or "credential" field. Put B1. Add a note: "Rate vs day-1 enrollees: [B2]."
- Find any "skills gained" field. Put C1, and name the rubric.
- Attach this page as a supplement, even if they didn't ask.
For WIOA-funded cohorts, whether this certificate counts toward credential attainment is a separate and unsettled question. Read our TEGL 10-16 analysis before you put B1 in a credential field.
Five common mistakes and fixes
1. Reporting invites as enrollment. It deflates your rate and hides who never accepted. Fix: separate A1 and A2.
2. Reporting only the flattering denominator. Fix: both A2 and A3, every time, side by side.
3. Merging certificate and work-sample results. "22 completers" could mean anything. Fix: two lines, two counts.
4. Leaving out the misses. A report without section D invites the funder to assume the worst, or not to notice. Fix: list them by category, with the privacy rule applied.
5. Changing the cutoff date after the fact. Learners who finish after the report date are real, but they belong in the next report. Fix: state the cutoff in B1 and report late finishers as an update.
Health check before you send
- Do B1 plus D1, D2 and D3 account for every learner in A2, allowing for overlap with C1?
- Can a stranger recompute every percentage from the counts on the page?
- Is every count under your threshold suppressed?
- Did the program manager sign it?
The free monthly ed.cafe Staff Room session, where staff can bring a draft report for a second pair of eyes, starts after ed.cafe's first cohort report; founding seats are open now.
