River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

Nonprofit Impact Report Template

Three documents and three sheets that compute cost per outcome on four denominators, because the same ledger gives a different answer on each.

Free download  ·  No account needed

Every impact report template hands you a slot for a hero statistic and a section called impact by the numbers. None of them asks the question a sophisticated funder asks first, which is cost per outcome. That figure is not a single number. It is a ladder of denominators, because a programme touches more people than it enrols, enrols more than it completes, and completes more than it can show an outcome for.

So this pack computes the figure on every rung and prints them side by side. In the worked example a literacy programme runs from $226.10 per person reached to $2,927.44 per outcome achieved, a factor of 12.9, on one year of one ledger with no arithmetic error anywhere. The rungs also disagree about which programme is cheapest. Adult Literacy leads on cost per person reached, Housing Stability leads on cost per outcome achieved, and two of the three change position between those columns.

The second undeclared choice is what to do with the participants nobody measured. 341 adults completed and 268 hold both a baseline and an exit score, so counting only measured successes gives $2,927.44 per outcome and extrapolating the responder rate gives $2,296.29. Last year gets recounted on this year's definitions before any trend is claimed, because in the worked example the published comparison reads as a 67.2% collapse and the programme grew on every definition applied to both years. The whole exercise starts from the indicators you can actually collect.

The same year, four defensible costs per outcome

Cost per outcome on every rung, every rate carrying its denominator, and last year recounted on this year's definitions.

Cost per Outcome

Illustrative, for a fictional organisation called Harlowe Bridge Initiative. One fiscal year, three programmes, one ledger. Every column below divides the same numerator by a different population.

ProgrammeFull costPer reachedPer enrolledPer completerPer outcomeSpan
Adult Literacy$500,591.65$226.10$828.79$1,468.01$2,927.4412.9x
Workforce Placement$469,486.92$280.46$1,476.37$2,193.86$3,848.2513.7x
Housing Stability$362,321.43$273.24$810.56$1,199.74$2,569.669.4x
Cheapest on this rungn/aLiteracyHousingHousingHousing2 of 3 move
What is in the numeratorAmountTreatment
Direct programme cost$1,096,600.00Charged straight to a programme
Management and general$235,800.00Allocated on direct programme cost, a basis the note states
Fundraising$203,600.00Not allocated. 13.3% of spending sits outside every figure above
Total functional expenses$1,536,000.00Ties to the statement, which is the point of this block
The counting choice, pricedOutcomesCost per outcome
Literacy, measured successes only171$2,927.44
Literacy, responder rate applied to all 341 completers218$2,296.29
Swing from that one undeclared assumption47$631.15, or 27.5%

Two things are load bearing here. The numerator never shrinks as you climb: the people who did not reach the outcome still consumed staff time, materials and space, so taking their cost out would produce a figure that pays a programme for losing them. And the ranking is not one list. Literacy is the cheapest programme per person reached and the second cheapest per outcome achieved, so a board handed one column has had a portfolio decision made for it. Workforce Placement is dearest on every rung, and that is the finding rather than an embarrassment: it enrols 19.0% of the people it reaches because it screens for work readiness, which is also why its verified employment rate is the number it is. Whether $3,848.25 per person placed in work of 20 hours or more is worth it is a question a board can answer. Whether workforce development is efficient is not.

Outcome Summary by Programme

The participation funnel, then every rate with the population it divided by on the same line. Coverage is measured over completers, because an enrolment that never started has nothing to measure.

ProgrammeReachedEnrolledCompletedMeasuredCoverageAchievedOf measuredOf completers
Adult Literacy2,21460434126878.6%17163.8%50.1%
Workforce Placement1,67431821418988.3%12264.6%57.0%
Housing Stability1,32644730219865.6%14171.2%46.7%
Sum of the three5,2141,36985765576.4%43466.3%50.6%
Distinct individuals4,491723 people are enrolled in two programmes, so the sum above overstates individuals by 16.1%
The same 341 literacy completers, three defensible ratesNumeratorDenominatorRate
Among those measured17126863.8%
All completers, unmeasured counted as no outcome17134150.1%
All enrolled, same rule17160428.3%
Measurement gaps, which are not randomPeopleWhy
Literacy completers with no exit score7351 of the 73 stopped attending in the final third, where the exit assessment sits
Workforce completers with no 90-day check25Unreachable by phone. Employment status unknown rather than absent
Housing completers with no 12-month check104Moved out of county, and the court records check does not follow them

Every column in the funnel has been published by somebody as people served, which is why the definition of each one travels with it. The three rates in the middle table come from the same 341 people and the same 171 successes, and the only thing that changes is what went underneath the line. An impact report that prints 63.8% is not lying. It is quoting the highest of three true numbers without saying which population produced it, and 63.8% against 50.1% is the difference between a strong programme and an ordinary one in a funder's notes. The bottom table is what decides whether the gap between them can be closed by extrapolation, and here it cannot: the 73 are the learners who faded in the last month of the course, so they are the group least likely to have gained two levels.

Trend Against Prior Year

Adult Literacy, this year against last. Restating a prior year means recounting its raw participant records on the current rule, not adjusting the figure it published.

ComparisonDefinition appliedPriorThis yearChangeComparable
As publishedAttended one session, against enrolled1,842604-67.2%No
Restated on reachedAttended at least one session, both years1,8422,214+20.2%Yes
Restated on enrolledIntake completed and a tutor assigned, both years561604+7.7%Yes
Restated on completed24 or more of 32 sessions, both years298341+14.4%Yes
Outcomes and costPriorThis yearChange
Full programme cost$451,800.00$500,591.65+10.8%
Measurement coverage71.5%78.6%+7.1 pts
Success rate among the measured64.8%63.8%-1.0 pts
Counted outcomes138171+23.9%
Cost per counted outcome$3,273.91$2,927.44-10.6%
Where the 33 extra outcomes came fromOutcomesKind of result
More people completed+19.9Programme result
More of them were measured+15.7Data collection result
The success rate among them fell-2.6Programme result, working the other way
Total+33.0Ties to the reported change of 33

The published comparison at the top reads as a collapse and the programme grew on all three restated definitions, so the entire apparent decline is a definition that tightened between two reports while neither report said so. The bottom table is the other half of the same discipline. Cost per counted outcome improved 10.6%, which is real arithmetic, and 15.7 of the 33 extra outcomes came from measuring more of the same learners rather than teaching them better. Coverage rose 7.1 points over a year in which the success rate among measured learners fell from 64.8% to 63.8%. An organisation that fixes only its data collection publishes a better cost per outcome, and a funder reading one column cannot tell that apart from a better programme.

What is in the pack

01

Cost per Outcome on four rungs

Full cost against people reached, enrolled, completed and shown to have achieved, with the between-programme ranking printed on every rung.

02

The numerator nobody strips

Participants who did not reach the outcome stay in the cost, because taking them out produces a figure that rewards a programme for losing people.

03

Outcome Summary by Programme

Every rate carries its numerator, its denominator, the instrument and the threshold on the same line, so no figure travels without them.

04

The unmeasured group, characterised

Counted and extrapolated side by side with the reason they differ, so the published figure is an interval rather than whichever endpoint reads better.

05

Trend Against Prior Year

Last year recounted on this year's definitions, then the change decomposed so better data collection cannot pass as a better programme.

06

Methodology Note, read before the narrative

Counts, definitions, the allocation basis and the limitation register, with every figure in the report tracing back to a line in it.

How it works

  1. 1

    Send the outcome data at participant level

    Not the summary counts. A summary has already collapsed the four populations the funnel needs, and it cannot be decomposed back into them. Where the participant records do not exist yet, that is a data collection design problem rather than a reporting one.

  2. 2

    Build the funnel before any rate

    Reached, enrolled, completed, measured and achieved, per programme, with the rule behind each. A funnel that goes back up is a definition problem, and it stops there.

  3. 3

    Allocate the cost and name the basis

    Management and general spread on a stated basis, fundraising held out and priced as a share of spending, so a reader knows what sits outside every figure.

  4. 4

    Restate last year, then decompose

    Recount the prior year on current definitions, then split the change into more completers, better coverage and the success rate among those measured. The figures you already filed have to survive it.

Frequently asked questions

Which cost per outcome should we publish?

All four, with the rung named on each. If one figure has to lead, lead with cost per outcome achieved and put the other three underneath it. A single figure with no denominator stated is not wrong, it is unfalsifiable, and a board cannot make a portfolio decision on one.

Should the cost of participants who did not succeed come out?

No, and the practitioner guidance on this is explicit. A cost per outcome incorporates the cost of delivering services to people who did not reach the intended outcome as well as those who did, which is what makes it cost per outcome rather than cost per success. Strip it and the figure rewards attrition.

Do funders actually ask for cost per outcome?

Federal awards ask in writing. A recipient must relate financial data to the performance goals of the award and supply unit cost data where the terms require it, and performance reports compare costs to units of accomplishment. Private funders ask the same question less formally, usually in the second meeting.

Can we extrapolate over the participants we never measured?

Yes, if it is labelled, the counted figure sits next to it, and the unmeasured group has been characterised first. In the worked example 51 of 73 unmeasured completers left in the final third of the course, so they are the group least likely to have succeeded and extrapolation flatters the result.

Our numbers went down. Do we publish that?

Check whether they went down first. A definition that tightened between two reports produces a fall nobody caused: 1,842 against 604 reads as a 67.2% collapse, and the same programme grew on reached, enrolled and completed. Restate the prior year from its records before you explain a decline.

Where does overhead belong in a cost per outcome?

In the numerator, on a basis you write down. The worked example allocates management and general on direct programme cost and holds fundraising out entirely, which leaves $203,600 of real spending, 13.3% of the total, outside every figure. Your functional expense statement is where the pools come from.

Is this the same thing as an evaluation?

No, and the report should say so. There is no comparison group here, so nothing in it establishes that the programme caused the change. A proper evaluation is the separate exercise built for that question. This report's honest limitation is what makes the figures reusable in the next grant proposal.

Find out which rung your number is on

Send the programme data and the expense detail. The first thing back is cost per outcome on four denominators and the spread between them.

Build my report