How to set realistic, measurable grant evaluation KPIs

The target you set in the proposal follows you for years

Evaluation measures are often written in the last days before a deadline, after the narrative is done and the budget is settled. Someone picks numbers that sound ambitious, a reviewer scores the evaluation section, and the application is submitted. If it is funded, those numbers become commitments. Federal agencies are directed by 2 CFR 200.301 to measure recipient performance against the goals and milestones in the award, and progress reports under 200.329 compare actual accomplishments with the objectives you set. Foundations read your reports against your proposal too.

An overpromised target produces two bad outcomes. Either the organization spends the award period explaining why it missed, or staff feel pressure to count generously. Neither builds trust. Realistic, measurable targets protect the program, the staff and the relationship with the funder.

Separate outputs from outcomes

Outputs count what the program delivers: workshops held, participants enrolled, home visits completed, screenings performed. Outcomes describe what changed for participants as a result: credentials earned, blood pressure controlled, children reading at grade level, families stably housed six months later. Most proposals need both, and reviewers notice when an evaluation section lists only outputs and calls them outcomes.

Outputs are easier to measure and largely within your control. Outcomes are harder and depend on factors you do not control. That is exactly why funders want them. A strong evaluation plan has a small number of output measures that show the program is being delivered as designed, and a smaller number of outcome measures that show whether it is working.

Start with a baseline you can defend

A target without a baseline is a guess. Before setting any outcome target, find out where you are today. For an existing program, pull the last two or three years of data and calculate the actual rate: how many participants completed, how many reached the outcome, how many you lost to follow-up. For a new program, use the closest comparable data you have, such as a pilot cohort, a similar program at a partner organization, or published results for the model you are adopting, and say clearly which one you used.

Then set the target relative to the baseline. If 61 percent of last year's participants earned a credential, a target of 68 percent with added tutoring support is a plausible improvement you can explain. A target of 90 percent is a claim a reviewer with field experience will discount. Explain how you derived each target in a sentence. Reviewers reward that transparency.

Write each measure so anyone can calculate it

A measure is only useful if two different people would calculate the same number from the same data. 'Improved family stability' cannot be calculated. 'Percentage of enrolled families with no change of address due to eviction or doubling up in the 6 months after program exit, verified by follow-up call or case record' can.

Precise definitions also settle arguments before they start. If the denominator is 'all enrolled participants', people who leave in the first week count against you. If it is 'participants who complete intake and attend at least three sessions', they do not. Either choice can be reasonable, but it must be stated in the proposal, applied the same way every reporting period and explained in the report.

• Numerator: who counts as achieving the outcome
• Denominator: who is included in the population measured
• Timeframe: when the outcome is measured relative to enrollment or exit
• Data source: the system, form or record the number comes from
• Frequency: how often it is collected and reported
• Owner: the staff member responsible for the data

Check that you can actually collect the data

The most common evaluation failure we see is not a missed target. It is a measure the organization never had the means to collect. Six-month follow-up sounds simple until you try to reach participants who have changed phone numbers. Employment wage data may require a data-sharing agreement with a state workforce agency. Clinical measures need to be captured consistently in the electronic health record, not in free-text notes.

For each measure, confirm who will collect it, in which system, and how much staff time it takes. If the data requires a partner, get a letter or an agreement before you submit. If follow-up is difficult, budget for it: incentives for participants, staff time for outreach, or an external evaluator. A federal budget can include evaluation costs, and reviewers are more confident in plans that pay for themselves. When the NOFO prescribes required performance measures, such as program-specific measures an agency reports under the GPRA Modernization Act, build your local measures around those rather than in competition with them.

Fewer measures, reported honestly

Three to five outcome measures and a similar number of output measures are enough for most programs. Every additional measure adds collection work and reporting burden for the life of the award. Choose the ones that test the most important links in your theory of change.

When you report, report what happened, including misses, with a short explanation and what you are changing. Funders who see an honest account of a shortfall and a credible correction are more likely to renew than funders who receive vague narratives that avoid the numbers. ELIXIR CONSUMERS links each KPI in a proposal to its data source in the vault and carries it into the funder report template, so the measure you promised is the measure you report. The discipline behind that is simple: baseline first, precise definitions, data you can collect, and honesty when the numbers come in.