Does Your Evaluation Need a Logic Model?

3 min read

Yes, but probably not the kind you're picturing. Your evaluation needs a one-page logic model that makes your program's causal claims specific enough to test. It doesn't need the laminated poster of boxes and arrows that got made for a grant application and hasn't been looked at since. Without the first kind, an evaluation has no way to decide what to measure or what a result means.

What a logic model actually is

A logic model is your program's argument, written as a chain: the resources you put in, the activities you run, the outputs those activities produce, and the outcomes you expect to change as a result, in the short, medium, and long term. A theory of change is the same idea with the why spelled out: the reasons you believe each link leads to the next. In practice, a good logic model carries a bit of both.

The value isn't the diagram. It's that writing the chain down forces you to say which link you're claiming, and each link is a claim someone could check. "Tutoring sessions lead to better reading scores" is a hope. "Twice-weekly sessions build decoding fluency within ten weeks, which shows up in spring reading benchmarks" is something an evaluation can actually test.

What an evaluator uses it for

Choosing the primary outcome. The logic model tells you which outcome sits closest to what the program directly does, and which ones are too far downstream to move in a one-year grant cycle.
Separating outputs from outcomes. Once the chain is on paper, it's obvious that "120 families enrolled" is an output box, not the result. Reports that blur the two usually never had a logic model to keep them apart.
Diagnosing a null result. If the long-term outcome didn't move, the model tells you where to look. Did the activities happen as planned (implementation)? Did the short-term outcome change but fail to carry through (theory)? Those are very different findings, and funders treat them very differently.
Spotting what needs a comparison. Each outcome box is a place where you'll need to show change beyond what would have happened anyway. The model makes it clear which boxes carry the claim and deserve a real comparison, and which only need to be monitored.

Signs your logic model won't help

Every arrow points everywhere. If each activity connects to every outcome, the model isn't making a claim. Cut it down to the links you'd actually defend.
The long-term outcome is a mission statement. "Thriving communities" can't be measured. Keep it as the goal, and make sure the intermediate outcomes are concrete enough to count.
No timing. Say when each outcome should show up. An outcome expected in year three shouldn't be judged in month six.
No assumptions. Every arrow rests on one ("parents can attend evening sessions," "the referral partner keeps sending clients"). Write the important ones down. When a program underperforms, a broken assumption is often the reason.
It was written after the data came in. A model built to fit the results you got proves nothing. Date it, and revise it openly when the program changes rather than quietly.

How much effort to put in

For most programs, one page and a couple of working sessions with the people who run it is enough. Start from the outcome you want to claim and work backward: what has to change first for that to happen, and what does the program do to cause that? A draft built this way in an afternoon beats a polished template filled in from the top down, because it's organized around the claim you'll eventually need to prove.

If you have a logic model that nobody uses, or you need one before an evaluation starts, turning it into a measurable evaluation plan is exactly the work we do with nonprofits.