Process changes people actually adopt

A correct process change is not the same thing as a change that survives month six — the rollout is a social object, not a document.

The idea

Most process improvements fail while being technically right. The new way is faster, safer, cheaper — and eight weeks later people are quietly doing the old thing again.

That happens because a change is judged by two different questions. Yours is is this better? Theirs is was I part of this, can I see why it was decided, and will anyone still care next month? A rollout that only answers the first question buys compliance, and compliance has a shelf life.

So plan the change the way you would plan a relationship: who is consulted before the pen moves, what rule gets published alongside the decision, who re-explains it in the language of the line, and who owns it after you stop watching.

Rollout designer — six months, four groups

Place each stage in a week, give it an owner, and watch four groups with different starting scepticism respond. The shaded band is the spread between the most and least adopting group; the solid line is the mean.

Same plan, different constraints. Switch tabs to rerun it.

band across four groups mean adoption night shift

Without a before-measure the audit can only prove compliance, never improvement — so its power to hold the gain is weaker.

How it works

Six moves, in an order that matters. The first three build legitimacy, the fourth and fifth build capability, the sixth is the only one that survives you.

1. List who must be consulted before anything is written. Four kinds of people: those who do the work, those who inspect or sign it, those who absorb the mess when it fails at 3am, and whoever will own it in month seven. If someone will be blamed for the outcome, they are consulted before, not informed after.
2. Consult before the design freezes — and prove it. The test for real consultation is simple: name one thing in the draft that changed because of what you heard. If nothing could have changed, you ran an approval tour. Say out loud which parts are fixed (usually the goal) and which are genuinely open (usually scope, sequence, exceptions, timing, measures). People forgive a narrow window; they do not forgive a fake one.
3. Publish the criteria, not just the outcome. "We chose option B" invites a rematch. "We chose option B because it was the only option that held handover time under ten minutes without a second signature, and here is what we traded away" ends the argument for most people — including the people who lost it. Visible rules narrow the gap between the enthusiasts and the sceptics.
4. Pilot on one shift, in real conditions. Evidence from a peer outranks assertion from a memo. Piloting on the most sceptical shift is expensive and worth it: if it works there, the argument is over. In a regulated setting, the pilot comes after the change is approved and staff are trained — see the second tab.
5. Translate at the frontline. An SOP written by people who have never worked the line is accurate and unusable. Have someone who has worked that shift re-explain it in their own words, with the three edge cases they already know about. That translation step is where the ceiling of adoption is set.
6. Sustain: , owner, measure. Almost every "will it stick?" question is really about month six. Answer it with an audit cadence (weekly for a month, then monthly), a named local owner, and one measure taken before and after. Without those three, gains do not collapse — they erode, which is worse, because nobody can say when.

The arithmetic the simulator is running

Legitimacy is built additively, then capability caps it, then reversion pressure eats it. Here is the night shift — the group that starts most sceptical and loses the decision — under the "designed to stick" plan.

starting trust (has resisted the last two changes) .... 0.10
+ consulted before the design froze ................... 0.28   (0.16 base x 1.6 sensitivity x 1.1 own lead ran it)
+ criteria published, and they lost the argument ...... 0.26   (0.10 everyone + 0.16 "I can see the rule")
+ they ran the pilot themselves ....................... 0.22   (0.10 own shift + 0.12 peer evidence)
+ SOP re-explained by their own shift lead ............ 0.06
                                                        ----
legitimacy .......................................... = 0.92

adoption target = capability x (0.25 + 0.65 x legitimacy) - reversion pressure
                = 1.00       x (0.25 + 0.65 x 0.92)      - 0.02
                = 0.83

reversion pressure at month 6, raw .................... 0.20
  x audit cadence running          (x0.35) ............ 0.070
  x local owner already handed over (x0.40) ........... 0.028
  x consultation was genuine        (x0.72) ........... 0.020

same workforce, no consultation / no criteria / no audit:
  legitimacy 0.05, capability 0.68, pressure 0.20
  early "directive glow" +0.20 fading over 9 weeks -> peak ~0.44 in month 2
  month 6 target = 0.68 x (0.25 + 0.03) - 0.20 = 0.00

The numbers are a teaching model, not a measurement. The shape is the point: skipping consultation is genuinely faster at first, and that early speed is exactly what hides the reversion until month four.

When to use it

The full six-stage rollout is for changes that touch how people work when nobody is watching. Scale it down honestly for anything smaller.

Situation	What to do, and the trade-off
Small, reversible change on one team	Consult the doers, skip the pilot, keep the measure and one check-in. Full ceremony on a small change reads as bureaucracy and spends credibility you will need later.
Cross-shift or cross-function change	Run all six stages. The cost is roughly six to ten weeks of front-loading before anything visibly changes — that is the trade.
A shift that has resisted previous changes	Consult them first, not last, and ask what broke last time rather than what they think of this. Give them the pilot and let the result be theirs. Cost: you may have to publicly concede something from the last rollout.
A change imposed from above you think is wrong	Argue upward with evidence and a proposed alternative, on the record. Then find the widest genuinely open space — timing, sequence, exceptions, what you measure — and consult on that. Be honest that the decision was made above you; do not pretend the whole thing was open, and do not join the grumbling. Disagreeing downward buys one evening of goodwill and costs you the ability to hold any standard afterwards. Escalate again only with data from the rollout.
Regulated setting (clinical, financial, safety-critical)	Sequence is constrained: approve and document the change, train the affected staff, then pilot. An unapproved trial run is a deviation, not an experiment, and carries real compliance exposure. Consultation is still free and still first — being regulated constrains what you may do, not who you may ask.

Watch out for

Consultation performed after the document exists. The meeting where you present a finished SOP and ask "any concerns?" is an approval tour. People read it instantly, and it costs more trust than not asking at all, because now you have also asked them to pretend. Test yourself: if no answer you could receive would change a line of the draft, do not call it consultation — call it a briefing, and be straight about why.
Publishing the outcome without the criteria. "After reviewing the options we've decided to standardise on the new handover form" tells the losing group only that they lost. Publish the rule that decided it and what you traded away. People accept a decision they lost far better when they can see the rule — and the rule also stops the next twelve one-off exception requests.
Skipping the frontline translation. A perfectly accurate SOP written by someone who has never worked nights will be technically followed and practically ignored. Someone from the line must re-explain it, including the edge cases the author did not know existed. If your trainer cannot name three things that go wrong on that shift, they are not the right trainer.
No before-measure, so no argument in month six. If you did not take the baseline before week one, you will be unable to distinguish reversion from noise, and your audit degrades into an attendance check. One measure, taken twice, beats a dashboard nobody reads.
Treating go-live as the finish line. The rollout budget usually runs out exactly when the honeymoon does. Book the audit cadence and the ownership handover into calendars before launch, with names on them; a rollout with no named owner in month seven is a rollout with an expiry date.

Worked example

The scenario: you run operations across four shifts. Handover errors are causing rework, and a new structured handover checklist would fix it. The night shift has resisted the last two changes and, as it happens, this decision goes against their preferred routine. You have six months and an interviewer asking how you would roll it out.

Weeks 1–2: consult all four groups plus the team leads and the quality lead, before the checklist is drafted, asking what broke in the last two rollouts and what the checklist must not do. You change two fields and drop a required second signature because of what nights told you — and you say so by name. Week 5: publish the decision and the criteria — under ten minutes, no second signature, auditable — and name what was traded away. Week 7: pilot on nights, run by their own shift lead, with a baseline error rate you took in week 0. Week 9: training delivered by that same shift lead, in their language, with the three edge cases the draft missed. Week 12: weekly audit for four weeks then monthly, owned by the shift leads. Week 19: ownership formally handed to the shift lead group, with the measure in their standing agenda.

Adoption is slower for the first two months than a directive would have been — that is the honest cost. But the band across the four groups stays narrow, month four passes without a slide, and in month six you can say: handover errors down from the week-0 baseline, audit still running, owner named, and the sceptical shift is the one enforcing it. If the same change had been imposed from above and you thought it wrong, the shape is identical — you argue upward with the pilot data, and consult downward on everything still genuinely open.

Check yourself

Your director says the new escalation rule is final and asks you to "socialise it" with the shifts next week. What is the most useful next move?

Six weeks after go-live, compliance is 82% and everyone is pleased. Which fact best predicts where it will be in month six?