THE BASELINE YOU NEED BEFORE YOU ANNOUNCE A CHANGE

A year after a change programme, someone will ask whether it worked. Five measures, taken in the fortnight before the announcement, decide whether that question can be answered.

Emmanuel Duncan · 7 September 2026 · 6 min read

A year after a change programme, someone will ask whether it worked. The answer will be assembled from impressions, from the people still in the room, and from whichever figures happen to be available and happen to point the right way.

That is avoidable, and cheaply. The window is the fortnight before the announcement, and the work is a handful of measures taken while the organisation still looks the way it did.

The reason it rarely happens is that the period before an announcement is the busiest one in the whole programme. Legal review, communication drafts, manager briefings, sequencing of individual conversations. Measurement feels like something that can wait until things calm down. By the time they do, the baseline has gone.

What a Baseline Is For

Two things, and it is worth being clear which one you are buying.

The first is evaluation. At nine or twelve months you want to say something defensible about whether the change delivered what was argued for. Without a before figure, every after figure is unanchored.

The second is earlier and more useful. A baseline gives you a set of tripwires. If time to fill vacancies doubles at month four, you find out at month four rather than at the annual review, and there is still time to act.

The second purpose is the one I would sell internally, because leadership teams respond better to early warning than to retrospective scoring.

Five Measures Worth Taking

These are chosen because they are sensitive to structural change, cheap to collect, and hard to argue with. Five is roughly the limit of what an organisation will actually repeat.

Decision latency. Pick two or three routine approvals that everyone recognises, a purchase requisition above a set value, a leave approval, a recruitment authorisation. Measure the days from submission to decision across the last three months. This is the measure that moves first when a structure is not carrying coordination, and it is the one people notice in their working lives.

Vacancy position. Count of open roles, and average days from approval to offer accepted over the last twelve months. Restructures interrupt hiring in ways nobody plans, and a vacancy backlog is one of the more expensive ways a change quietly fails.

Regretted exits by unit. The last twelve months, classified as regretted or not by the line manager. This is the measure that tells you whether the change cost you people you wanted to keep, and it is worthless if you try to build it afterwards, because the classifications will be made by people who already know how the story ended.

Role clarity. One or two questions, put to the affected population, using wording you commit to repeating exactly. Something plain: whether the person can say who approves their work, and whether they can say what they are accountable for. Keep it short enough that you will run it again.

One operational measure the business already trusts. Cases closed, orders shipped, claims processed, students enrolled, whatever the organisation manages by. Take it from the existing operational report rather than building anything. Its value is that finance and operations already believe it, which means the eventual evaluation is not conducted entirely in HR's own numbers.

Getting Them Before the Announcement Without Leaking It

The obvious objection is that measurement is visible, and a survey landing three weeks before an announcement will be read for what it means.

Four of the five measures are drawn from records that already exist and need no contact with staff at all. Decision latency, vacancies, exits and the operational measure are all pulled quietly from systems.

The role clarity question is the exception. Options, in order of preference. Attach it to something already scheduled, an existing pulse survey or an annual process. Run it far enough ahead that it reads as routine, which usually means before the design work is visible. Or accept a baseline taken in the first fortnight after the announcement and label it honestly as such, since the structure has not yet changed even though the news has.

What I would avoid is a new survey appearing in the quiet weeks before an announcement with no explanation. In an organisation where people are already reading the signs, that is itself a signal.

The Definitions Matter More Than the Numbers

A baseline is only useful if the later measurement is the same measurement.

Write down, for each of the five, exactly what was counted, over what period, from which source, filtered how, and by whom. One page. Store it with the figures, not in someone's email.

The most common way baselines fail has little to do with whether anyone took them. The person who took them leaves. The person repeating the exercise a year later makes slightly different choices, and the comparison quietly becomes meaningless while still looking like a comparison.

Include the query or the spreadsheet itself, not a description of it.

When to Measure Again

Three months, nine months, and eighteen if the change was substantial.

Three months is early enough to act and late enough that the transitional noise has settled. Nine months is where most structural effects have either taken hold or failed to. Eighteen matters only for changes intended to shift how the organisation operates rather than how it is drawn.

Resist the temptation to measure monthly. Measures that are reported too often get managed rather than watched, and the effort of collection erodes the willingness to keep going.

What to Do With a Result You Did Not Want

At some point the three month figures will show something uncomfortable. Decision latency up, vacancies stuck, role clarity down in the units that were most affected.

The value of having taken a baseline is entirely lost if that result is explained away. It is worth agreeing in advance, in writing, what the organisation will do if the measures move in the wrong direction, and who will see them regardless of what they say. That agreement is easier to obtain before the announcement than after, because nobody is yet defending anything.

Where to Start

If a change is coming in the next quarter, block two days now. One to pull the four record based measures and write the definitions page. One to find a home for the role clarity question.

If a change has already been announced and you have no baseline, take the four record based measures anyway, going back twelve months from the current date. The historical records still hold what the organisation looked like before, and that is most of what you needed.

The question a year from now will be asked whether or not you prepared for it. Two days of work decides whether it can be answered.

Emmanuel Duncan is the founder of Aletheia Lunara Consult, a professional education and consulting practice working with organisations across Ghana and beyond on people analytics, workforce research and organisational capability.