Switch course ▾
Session 5. Culture of Experimentation, Honest Results and Failing Safely
Objective. Build the conditions for a school culture of experimentation, run a blameless review of a pilot that did not work, write honest results with their limits, and present a five-minute briefing to a board or education department.
Whether a school keeps innovating depends less on its plans than on what happens when something does not work. If a teacher who reports a problem is thanked, problems come early and are cheap to fix. If the teacher is blamed, problems are hidden until they are big. This session helps you shape that response, review a pilot that missed its target without blaming anyone, report results so that a board can trust them, and share both successes and failures with other schools.
Key ideas
A school where it is safe to try
- Amy Edmondson defined team psychological safety as a shared belief that the team is safe for interpersonal risk-taking: people can ask a question, admit a mistake or report a problem without being embarrassed or punished. In a school, this decides whether a teacher tells you in week 2 that the pilot is not working or hides it until week 6.
- Safety does not mean low standards. Keep high standards for pupils' learning and honest data, and combine them with kindness to people who report problems early. The signals are concrete: problems arrive early, failed pilots are discussed in the staff room, and the same avoidable mistake is not repeated.
- Leaders create safety by what they do: ask questions before giving answers, say 'I do not know yet', tell a story of their own mistake, thank the first person who reports a problem, and separate the person from the result. A 10-minute 'what we learned' slot once a month costs almost nothing and makes the habit visible.
Failing safely: small, short, reversible, reviewed
- A safe failure is an idea that did not work as hoped, found out cheaply: small scale, a short period, a stop rule agreed in advance and pupils protected by the red lines from session 4. Harm to a child is a different matter: it is not 'learning', it is a red line crossed, and it is handled at once.
- Separate two kinds of failure. A well-designed small test of a new idea that did not work is useful. A failure that came from skipping a known step (no baseline, no gate, a warning ignored) is avoidable and should be discussed as a process problem, without blaming the person.
- Run a blameless review in four questions within two weeks: What did we expect? What actually happened (facts and numbers first)? Why did it happen (look for causes in the timetable, the tools, the instructions and the support, not in a person's character)? What will we keep, change or stop, who does it and by when? Write it on one page and share it.
Telling the results honestly
- One honest results page has: the problem in one line; what we did; the result with numbers (before, after and how many pupils); the limits; the cost; the decision; and the next step. Show counts such as '19 of 28' beside any percentage, and avoid percentages for very small groups.
- State the limits plainly: one class, five weeks, the same teacher teaching and measuring, no comparison group, pupils may have learned the format of the check. Use words that match the evidence: 'suggests' for a small pilot, not 'proves'. Nesta's levels of evidence help you say how strong your evidence is.
- Report the balancing measure and the disappointing numbers as well as the good ones, and credit the colleagues who did the work. Never include pupils' names, faces or marks. Honest reporting builds the trust that lets you ask for money and time next time.
Five minutes in front of a board or education department
- Start with the decision you ask for, in one sentence, then why now, the evidence with its limits, the cost, the risks and safeguards, and the next review date. Boards decide; they do not need the whole story, but they need to trust the numbers.
- Prepare for the three hardest questions in advance (cost in year 2, what if it fails, who is left out) and answer in two sentences. If you do not know, say so and promise a date for the answer. Send a one-page summary within 24 hours.
- Share failures outside your school too. A short note in the cluster meeting or a methodological association channel, with no pupil data, saves other schools weeks. Course 15.6 (Scaling Innovation Across a School Network) builds on this when ideas must travel beyond one school.
The learning cycle of a school that experiments
Small rituals that make learning normal
Culture changes through repeated small actions, not announcements. A monthly 10-minute 'what we learned' slot where one teacher describes a pilot, including one that did not work, makes failure ordinary. If the head goes first, others follow. A short 'before we begin' conversation, sometimes called a pre-mortem, helps too: imagine the pilot has failed in six weeks and write down three reasons; then fix the likeliest one now.
Reward the right behaviour. Thank the teacher who used the stop rule early. Count stopped pilots in appraisal discussions as evidence of good judgement. Do not ask 'whose pilot failed?' in a staff meeting; ask 'what did the pilot teach us?' Pupils' safety is not negotiable: a red line crossed is handled at once and is a different matter from a well-run small test that did not work.
Numbers that tell the truth
Show counts with the total, such as 19 of 30, because percentages of small groups look more impressive than they are. Put before and after side by side, use the same scale, and begin a chart at zero. State when and how the numbers were collected and by whom. If the teacher who taught the lesson also counted the results, say so.
Report what went wrong or was disappointing: the balancing measure, the pupils who did not benefit, the extra workload. A board that finds a hidden problem later will doubt every figure. Use cautious verbs such as 'suggests' and 'is consistent with' for small pilots, and state the next step that would make the evidence stronger, such as a trial in two more classes with a second person counting.
Examples from the classroom
A blameless review in Fergana
Scene. Aziz's peer-feedback pilot reached only 2 of 8 teachers. Instead of a long meeting he asks everyone for one fact on a slip. The slips show a 25-minute cycle, no timetable slot, Wednesdays used for substitutions and a reminder that five of eight never saw. The group agrees to a one-page template, a protected slot on Thursdays and reminders on paper. Nobody is named, and the second trial reaches 6 of 8 teachers.
Insight. Asking for facts first moves the discussion from character to conditions.
A five-minute briefing in Tashkent
Scene. Dilnoza has five minutes with the district education department. She opens: 'We ask permission for a one-term trial in two more classes.' Then: 'In Grade 7A, 19 of 30 pupils now name their own error, 10 before; one class, the same teacher counting, no comparison class.' She adds the cost of 8 teacher hours and one risk, that a pupil may feel exposed, with her safeguard. She answers two questions and sends a one-page summary the next morning.
Insight. A decision first, honest numbers with limits and a prepared safeguard earn trust.
Sharing a failure with the cluster in Samarkand
Scene. After a project week in which only 9 of 20 teachers took part, Malika posts a short note to the cluster group: the problem, what was tried, the result with its limits, the three causes found and what the school will change. She leaves out pupils' names and photos. Two other schools reply that they had the same problem with timetable slots, and one sends its revised schedule.
Insight. A plain account of a failure saves other schools weeks and builds the habit of honest sharing.
Common pitfalls
- Asking 'whose fault was it?' before asking 'what happened?'
- Hiding a pilot that did not work, so the school repeats the same mistake next year.
- Using percentages for tiny groups or leaving out the limits of the evidence.
- Presenting only successes to the board, so that every later figure is doubted.
Learn, Report and Present
- Pairs (15 min): in Table A, three sample sentences overclaim a result. Say what is wrong with each, then write one honest results statement with numbers and limits for the sample pilot data.
- Pairs (15 min): in Table B, read the sample story of a pilot that did not reach its target. Complete the four-question blameless review and rewrite three blaming sentences as system causes.
- Pairs (10 min): in Table C, mark eight sample leader behaviours as building or harming safety. Choose two you will start and one you will stop in your own school.
- Part B, alone (25 min): in Table D write a five-minute briefing for your school board or education department on one real pilot or portfolio decision: the ask, evidence with limits, cost, risks and next date, and your answers to the three hardest questions.
- Part B, trios (15 min): each presents the briefing in 4 minutes; the others play the board and ask the three hardest questions, then give one thing that was clear and one that was missing.
- Alone (5 min): write your leadership task and final portfolio list from the Kit.
Your facilitator will guide you through this activity in the session.
Questions to discuss
- Which sentence in your briefing would a sceptical board member challenge first, and how will you answer?
- What was the cause in the review that first looked like a person's fault?
- Which one behaviour will you start next Monday to make trying safer?
Reflect
- When did a colleague last tell you early that something was not working, and how did you react?
- Which result have you reported with more confidence than the evidence allowed?
- What would you say to a board if your best pilot had failed?
Try it tomorrow
- Start a 10-minute 'what we learned' slot next month and tell one story of your own that did not work.
- Before your next pilot, write three reasons it might fail and fix the likeliest one.
- Write a one-page results note for one pilot with counts, limits, cost, decision and next step, and show it to a colleague.
Further reading and links
Make it safe to try, state results with their limits, review failures without blame, and ask boards for a clear decision with honest numbers.