Version 0.2 · rebuilt from the evidence
The Unfrozen Loop
Most change models I studied start at the top and run once. The Loop starts in the middle and keeps running: a week-0 baseline, five parts on a rhythm, and a culture where it’s safe to speak.
Not yet proven as a whole. Each part shows the strongest evidence I’ve found and what no controlled study has tested yet. Grades are provisional.
Week 0, five parts, one culture
Week 0, before you start
Record how long bad news takes to reach you, who speaks up and who doesn’t, and what the team says the change will cost. Share it back with the team, then compare against it at every review.
- Molan et al. 2018, randomized trial across 49 small units of one disability-services nonprofit, rated High from the abstract, so the grade is provisional: staff who got their survey results back rated the relational side of their own service higher.
- Manthei & Sliwka 2019, field experiment in one bank, rated Medium (journal version read from its abstract): branches whose managers got objective performance data earned higher profits; about 2% more is the working-paper figure.
Still untested: whether a manager who records a baseline gets a better change outcome. No study I’ve graded so far (October 2026) tests it.
-
1Surface both ways
Bad news and ideas travel up as easily as plans travel down.
What the evidence says- Schaubroeck et al. 1994, field experiment in one plant, 173 people, rated Medium: explaining why a pay freeze was fair did not lift attitudes overall, but it took the edge off for the people the freeze hurt most, at least right after the session.
- Rigtering et al. 2019, field experiment, rated Medium: making an idea program opt-out raised participation without lowering idea quality.
Still untestedNo controlled test yet, in the studies I’ve graded so far (October 2026), of a manager-level routine that increases how much bad news flows upward.
A first moveThe research on this part, in the AtlasAdd one question to every team meeting for the next month: “What's getting in the way of this change?” Write the answers down and send them up unedited.
-
2Make sense together, top included
Managers and their teams work out what the change means for their own work. The people above say out loud what it costs.
What the evidence says- Williams et al. 2026, randomized trial of 21 clinics, rated High: training clinic leaders raised how well staff made sense of a new practice and built it into their routines.
- Blair et al. 2024, six pooled field experiments, rated Medium: strong backing from leaders didn’t stop officers carrying out a new policing program unevenly. Support from the top wasn’t enough on its own.
Still untestedNo trial yet, in the studies I’ve graded so far (October 2026), has senior leaders joining their teams to work out what a change means.
A first moveThe research on this part, in the AtlasAsk each manager to explain the change to their team in their own words, then ask the team what it means for their week.
-
3Start it close to the work
Running the change is the middle’s real job. The manager leads the rollout with the team: plans it, trials it small, checks how it’s going and removes obstacles.
What the evidence says- Williams et al. 2024, randomized trial of 21 clinics, rated High: in clinics whose leaders got an implementation-leadership program, clinicians viewed the measure’s feedback report in 23.1% of sessions on average, against 3.4% in clinics that had training and technical assistance only.
- Brookman-Frazee et al. 2025, two linked randomized trials of two different autism interventions, one in mental health programs and one in school classrooms, 65 programs and districts in all, rated High: a strategy aimed at program leaders raised how faithfully a practice was delivered. A separate strategy aimed at frontline providers had no significant effect on any outcome.
- Nearly all of these trials were run in health and human services.
Still untestedNo trial yet, in the studies I’ve graded so far (October 2026), compares a change designed by the front line with one designed by the manager.
A first moveThe research on this part, in the AtlasBefore the next rollout step, try it with one team for two weeks and change the plan based on what they find.
-
4Review on a rhythm, then hold the pace
Short, regular reviews that each end in one fix. Keep them running. Gains may fade when they stop. That is a hint, not a tested finding: in one clinic trial, read from its abstract, some early gains from a leadership program had faded 13 to 26 months after training. Then hold off adding new work while the change settles.
What the evidence says- Tannenbaum & Cerasoli 2013, meta-analysis, rated High: debriefs were linked to about 25% better performance on average than comparison groups; many of the pooled studies were quasi-experimental.
- Williams et al. 2025, randomized trial, rated High from the abstract: 13 to 26 months after training, the program’s edge in completion and symptom outcomes was not sustained. Among the 81 youths who received the care, fidelity stayed higher at program sites (24% against 1%, rated Medium).
Still untestedHolding off new work while a change settles has no trial yet in the studies I’ve graded so far (October 2026). It stays labeled as my judgment.
A first moveThe research on this part, in the AtlasStart a 30-minute review every two weeks, run by the managers, that ends with one named fix and one owner.
-
5The middle’s own weather
The manager’s own workload, strain and support shape whether the change holds. Peer support may help more when peers trade advice on the actual work; that is a working idea, not a tested finding. Support for staff may matter most where their own manager felt unsupported. That is only a hint, not a tested finding: it comes from a post hoc subgroup in one trial.
What the evidence says- Linos et al. 2022, field experiment, 536 dispatchers, rated Medium: 911 dispatchers prompted each week to swap advice with peers in other cities reported less burnout four months later; that result comes from the 152 of 536 who answered the survey. Counting everyone randomized, the drop in resignations was not significant; among those who stayed through the six weeks, resignations fell by more than half.
- Meredith et al. 2024, randomized trial, 2,077 health workers, rated High from the abstract: a broad peer-support program had no overall effect on distress.
- Jakobsen & Andersen 2013, field experiment in one city’s child-care centers, rated Medium: offering staff extra support had no average effect; a post hoc subgroup, rated Low, hinted it helped where their own manager had felt unsupported.
Still untestedOrganizational changes to a middle manager’s load: nearly all tested programs in the studies I’ve graded so far (October 2026) target the individual instead.
A first moveThe research on this part, in the AtlasTake one recurring report or meeting off each manager's plate until the change is in place.
Test it on your own change
Run it with your team
Send one link. Your team answers anonymously. You see the combined results.
Free lesson · about 50 minutes
Learn to read the evidence grades on this page
Module 0 of Leading Change with Evidence teaches you to grade a study High, Medium or Low with three questions, and to mark each step of a change plan as a tested finding or as practice. It’s free, and it opens right away when you sign up.
people have played Run the Middle · teams have run the Loop Check
Get notified when the book is out
Unfrozen, the book, is a week-by-week plan for new managers leading a change through its first 90 days. Its chapters point back to the Loop. Leave your email and I’ll send one note when it’s out. Nothing else.
Break it
Tell me about a change you've lived through where one of these parts was done well and the change still failed, or skipped and it still worked. If your example changes the model, I'll say what changed in the next version and credit you, with your permission.
Version history
- v0.2 · October 2026Rebuilt from the evidence. Added a week-0 baseline (measure it, share it back) and the culture ring. The middle’s own weather became a full fifth part. Part 3 now names the manager’s job: lead the rollout. Part 4 now says to keep the reviews running, and “hold still” became “hold the pace.” Added one rule across the Loop: training starts the change; follow-through finishes it. Each part now shows its strongest evidence and what is still untested. On October 26, after full-text reads, the Linos and Jakobsen & Andersen trials in Part 5 and the Althammer trial in the culture ring were marked down to Medium, the idea that support from above matters most where managers lack it became a labeled hint, and the Part 3 figure was reworded as a share of sessions.
- v0.1 · October 22, 2026First public version. Four parts, a sign for each, and one experimental part.
Every change to the model will be logged here.