How to write evaluation recommendations that get actioned

Most evaluation recommendations are written so that no one has to do anything. They name a problem, point at a direction, and leave the reader to work out who acts, what changes, and by when. So nothing happens. The recommendation is noted, the report is filed, and next year's evaluation finds the same issue.

This is not an analysis problem. It is a writing problem, and it belongs to whoever drafts the recommendation.

Here is a recommendation I have read, in one form or another, in more reports than I can count: "Consideration should be given to improving the accessibility of digital services." It is accurate. It is also useless. It names no owner, no action, and no date. A busy executive reads it, agrees with it, and moves on, because there is nothing in the sentence to act on.

A recommendation is not a finding. A finding describes what is true. A recommendation tells a named person what to do about it. When both are written in the same hedged, passive register, the recommendation inherits the finding's voice, and the reader can no longer tell an instruction from an observation.

The Commonwealth Evaluation Policy expects evaluations to be used, not just produced. Use is won or lost in the recommendation, because it is the only part of the report that asks anyone to change anything.

A recommendation needs an owner, an action, and a date

Every recommendation should name three things. The role accountable for acting, not "the agency" but the position that owns the decision. The specific action, in a verb the reader can picture doing. And the timeframe, so it can be tracked, and so it cannot be quietly deferred forever.

Miss any one of the three and the recommendation stops being actionable. No owner, and everyone assumes someone else. No action, and agreement changes nothing. No date, and it slides to the bottom of a list that never clears.

"Consideration should be given" is where accountability goes to die

Passive voice hides the actor. In a finding, that is fine. In a recommendation, it is fatal, because the actor is the whole point. "Consideration should be given" tells no one to do anything.

Rewrite the earlier example and the difference is obvious. "The Digital Services branch head should fund a WCAG 2.2 audit of the top 20 transaction pages by the end of Q3, and report the results to the executive board." Same finding underneath. Now someone can act, and someone can be asked why they have not.

Write the recommendation as a test, not a wish

Before you write a recommendation, answer one question: who reads this and has the authority to act on it? If you cannot name the role, the recommendation is not ready, and no amount of careful wording will fix that.

This is the same discipline content design applies to any high-stakes page. Decide the reader and the action first, then write. A recommendation is content with a job, and the job is a decision.

Someone has to own the recommendation layer

In most evaluation teams, the method has an owner and the recommendations do not. They are drafted last, often by whoever is closest to the deadline, and reviewed for accuracy rather than for whether anyone can act on them.

This is a governance gap, not a talent gap. Name a person accountable for how recommendations are written, the same way you name someone accountable for the sampling or the analysis. Give them a simple standard: every recommendation names an owner, an action, and a date, and passes the test above before the report goes out.

What good looks like

When this is working, you can read the recommendations on their own, with no other part of the report, and know exactly who has to do what by when. There is no "consideration should be given". Each recommendation could be lifted straight into an action log. And one person owned the standard they were written to.

So open your last evaluation and read only the recommendations. Could a new starter tell, from the words alone, who acts, what they do, and by when? If not, the analysis was never the thing holding it back.

Photo by epicantus on Unsplash.