The assignment: prepare a funder-ready update that can be checked
A fictional workshop planned for forty students enrolls thirty-six. Costs exceed the original budget, and only some participants complete both assessments. A program manager needs an update explaining delivery, funding needs and what the results actually show. Your task is to produce a budget bridge, an outcome-coverage table and a short narrative.
The files contain budget quantities and rates, funding status, and anonymous synthetic participant IDs with completion and score fields. No actual student or beneficiary data is used. The expected score range is zero through one hundred, and improvement means a post-score strictly higher than the pre-score for a person with both measurements.
CDC’s program-evaluation guidance distinguishes program resources, activities, outputs and intended outcomes. Use that distinction here: enrollment and completion describe delivery; assessed skill change is a different measure. The case has no randomized comparison group, so it cannot establish that the workshop caused a score change.
Data and worked inputs
Sources[1]
Explain the cost overrun one operating input at a time
The planned budget is $9,400. Actual cost is $10,020, an increase of $620. Facilitator hours rise from eighty to ninety at $50, adding $500. Materials for thirty-six rather than forty participants save $180. Transport adds $200 and evaluation adds $100. Venue cost is unchanged.
| Item | Amount |
|---|---|
| Planned cost | $9,400 |
| Facilitator hours | +$500 |
| Materials volume | −$180 |
| Transport | +$200 |
| Evaluation | +$100 |
| Actual cost | $10,020 |
| Committed funding | $9,000 |
| Unfunded actual cost | $1,020 |
The bridge is $9,400 + $500 − $180 + $200 + $100 = $10,020. This explanation is more useful than saying the program was over budget by 6.6%. It identifies the changes a manager can discuss and shows that lower enrollment did not automatically lower total cost.
Only $9,000 of funding is committed. The additional $2,000 in the file is prospective, so it does not close the $1,020 actual-cost gap. It is a potential financing option, not confirmed cash. The project also does not establish when committed funds are received; a funding-gap analysis is distinct from a dated cash forecast.
Match assessment pairs by participant ID
Twenty-seven of the thirty-six enrolled participants complete the program, giving 75% completion. Thirty have a pre-score; twenty-four have both pre and post. The paired group covers two-thirds of enrolled participants. Do not pair rows by position or compare the mean of all thirty pre-scores with the mean of twenty-four post-scores.
For each paired participant, calculate post minus pre, then average those within-person changes. The mean gain is thirteen points in the paired group. Twenty of the twenty-four improve; four have no change. The mean and the share improving are different summaries, and neither describes unmeasured outcomes automatically.
Missing post-scores stay missing. Replacing them with zero would invent a severe decline; dropping their participants from the enrollment denominator would hide coverage. Report the paired result and its coverage together. In this case, missingness is itself an evaluation finding because one-third of enrollees lack a complete pair.
Show what the missing outcomes could change
At least twenty of the thirty-six enrollees are known to improve under the stated binary definition. If none of the twelve without paired measurements improved, the whole-cohort share would be 20/36, or 55.56%. If all twelve improved, it would be 32/36, or 88.89%. Those are logical missing-data bounds, not sampling confidence intervals.
The bounds concern the fraction whose score increased. They do not bound the whole-cohort mean score change, because the missing changes could have different magnitudes. They also do not establish causality: even complete before-and-after data would leave other explanations without a valid comparison design.
This is a useful example of an informative limit. “We cannot know anything” is too weak, while “participants improved by thirteen points” is too broad. A defensible statement is that the twenty-four paired participants improved by an average of thirteen points and that whole-cohort coverage and attribution remain unresolved.
Choose a denominator that matches the question
Actual cost per enrollee is $10,020/36, about $278.33. Cost per completion is $10,020/27, about $371.11. Neither is cost per person whose outcome was caused by the program. You cannot compute that causal quantity from this case.
Do not compare $278.33 with another program’s cost per graduate without noticing the denominator difference. Also ask whether the costs cover the same period and activities. A lower number can reflect omitted support rather than greater efficiency.
Use the unit costs to ask an operating question: which costs changed with participation, and which were committed regardless of enrollment? Then connect that question to next cycle’s plan. Reducing evaluation spending might lower cost while making results harder to assess; a recommendation should acknowledge that tradeoff rather than optimize the cost ratio alone.
Write the update a program director should be comfortable sending
The example below is intentionally concrete about both the achievement and the gap. It avoids turning a partial outcome dataset into a claim about every participant.
Original fictional program-and-funding memo
Prepare a workbook that makes its exclusions visible
Keep raw participant records, paired calculations and cohort totals in separate views. Validate unique IDs, boolean completion values and scores within the declared scale. An out-of-range score is an error, not a reason to cap it at one hundred without checking.
In the budget model, calculate cost from quantities and rates rather than hardcoding the total. Keep prospective funding separate from committed funds. Add an optional next-cycle scenario, but label assumptions about enrollment and staffing rather than presenting them as measured facts.
The final student package should contain the budget bridge, outcome table, missing-data interpretation and memo. The reference checker can confirm arithmetic on the supplied data. It cannot evaluate whether an actual program should be funded or whether your proposed data collection is appropriate for a real population.
Data and worked inputs
Use the analysis for development or evaluation applications
For a fundraising-support role, emphasize the reconciled budget, funding status and honest reporting narrative. For an evaluation role, emphasize pairing, coverage and what cannot be inferred. The same project demonstrates different skills depending on the responsibility you are discussing.
An accurate description is: “Built a synthetic nonprofit budget and outcome review, separating a $1,020 funding gap from incomplete assessment coverage and distinguishing paired gains from causal impact.” Do not claim an awarded grant, a real beneficiary result or a program you evaluated professionally.
A good interview follow-up is how you would improve the next evaluation without simply collecting more fields. Explain which specific missing evidence matters, how it would change the decision and what burden its collection creates. That demonstrates care for the people involved as well as the numbers.