Learn quality improvement by doing it
Build a paper robot to a customer’s order, against the clock. Record build times and defects, then work through the five steps of quality improvement to test changes and see what works. A practical activity for induction, teaching or an MDT team session.
- 30–45 minutes
- Teams of 4–6
- Facilitator-led
You’ll need one printed kit per team. Setup, videos, the timer and the run chart follow on this page.
Get ready to play
Prepare the printed materials, watch the short setup video and get your team ready. Once everyone understands the rules, you'll be ready for the timed robot builds.
Watch the setup video
Introduce the activity, organise teams of four to six and explain the rules before opening the order cards. The English subtitles are included in the video itself.
Read the setup instructions
Each team builds a paper robot to a customer's order, against the clock. Use the printed base and cut-out parts to assemble the robot. Every build is timed, checked by an inspector and recorded with its defect count.
Place one prepared kit, a score sheet and pen on each table. Keep the order cards face down until instructed. Assign a facilitator to coordinate the session and an inspector from another table to check the finished robot.
Start with baseline builds without making improvements. Later stages introduce a prediction, one change at a time, another timed test, and a comparison of results.
Prepare your robot kit
The original pack contains the robot base, 27 cut-out parts, eight customer order cards, score sheet and facilitator's run sheet.
Open printable kit (PDF) ↗Facilitator tip: Allow around 15 minutes of preparation per kit. Print and cut the parts before the session, ready for teams to use.
Before you begin
The four rules of the game
- Build the robot according to the customer order. Start the timer at “Go” and stop at “Done”.
- Place all nine parts in their correct spaces, right way up, with the requested pattern.
- The inspector checks for wrong patterns, wrong spaces, upside-down parts and missing parts. Record the time and defect count.
- Don't change the process during baseline builds. In later tests, predict the effect before changing one thing at a time.
The five-step challenge
Follow the stages in order, watch the short videos and build a robot against the clock. Record two baseline builds, test six changes one at a time, then see whether another team can repeat the result. Your figures and notes remain in this browser, not on SAM’s servers.
Stage 1 — Understand the problem
What is really happening?
Read video narration
Stage 1: understand the problem. Before we change anything, we need to see what is really happening. Start with two baseline builds. Build Order 1, then Order 2, with no planning and no changes. Write down the seconds and the defects for each one. On the second build, one person does nothing but watch. Where did hands wait? What did people have to search for? And who talked over whom? Now sketch the steps from Go to Done as you actually did them. Include the waits and the searching, because those are what you will improve. Take the slowest moment and ask why. Then ask why again, five times in all. The cause usually lies in how the work is set up rather than in the people. So you now have two baseline times, a map of the process and a cause. Step 1 gives a real project the same three things. Stage 2 is next: defining the problem and setting your aim.
What to do
Make two baseline builds: Order 1, then Order 2. Record seconds and defects without planning changes. On the second build, one person watches without touching the parts. Sketch the process from Go to Done, including waits and searches, and use five whys on the slowest point.
Give out Order 1, call Go and let inspectors count defects. Repeat with Order 2 and an observer. Allow time for the process sketch and each team to share one possible cause.
Watch for: No planning or sorting between baseline builds. Observers must not help. Treat defects as data, not criticism.
Done when: Two baseline builds, a process sketch and one likely cause.
Stage 2 — Define the problem and aim
What are we trying to improve?
Read video narration
Stage 2: define the problem and the aim. What exactly are we trying to improve, and how will we know when we have? Start by stating the problem with numbers. For example: “Our robot takes 74 seconds and has two defects, because the parts are scattered and nobody has a role.” Then write your aim. Make it specific and measurable, with a number and a deadline, such as: “Under 40 seconds, with zero defects, by build 8.” Now choose your measures. The outcome measure is seconds per build. The process measure is whether the parts are laid out before Go. The balancing measure is defects, so that speed does not cost quality. Plot builds 1 and 2 on your run chart. The line between them is your baseline median, and every change that follows is measured against it. Stage 3 is next: choosing change ideas.
What to do
Turn the two baseline builds into a problem statement with numbers. Write a specific aim with a time, a defect target and a deadline expressed as a build number. Keep track of outcome (seconds), process (for example, parts laid out before Go) and balancing (defects).
Ask each team to write a problem statement and an aim. Check that every aim includes a number and a deadline; ask one team to read theirs aloud.
Watch for: “Faster” is not a measurable aim. Do not omit defects as the balancing measure.
Done when: A quantified problem, a measurable aim and three named measures.
Stage 3 — Choose change ideas
Which small change will you try?
Read video narration
Stage 3: choose change ideas. Which changes are most likely to help? Start from the cause you found. Your five whys pointed somewhere, perhaps to the layout, the roles, the order of work or how the order card is read. Sketch a quick driver diagram. Your aim goes on the left, then the drivers: preparation, roles, sequence and communication. Put your ideas on the right, as many as you can think of in three minutes. Now pick one. Weigh impact against effort and choose the quick win. Test one change at a time, so that you know what made the difference. Before you test it, write down what you expect. For example: “Laying the parts out first will get us under 60 seconds.” Stage 4 is next: testing and measuring improvement.
What to do
Generate several ideas without judging them first. Consider impact and effort, then choose ONE change for build 3. Write down what you will change and predict the build time before you test it.
Give teams three minutes to generate ideas and one minute to choose. Confirm they have selected just one change and predicted a time in seconds.
Watch for: No bundles of changes in a single build. A prediction should be a number, not “as fast as possible”.
Done when: One selected change and a numerical prediction for build 3.
Stage 4 — Test and measure improvement
Are your changes working?
Read video narration
Stage 4: test and measure improvement. The question now is whether the changes are working. Each test is one cycle. Plan the change. Do the build. Study the time and the defects against your prediction. Then act: keep it, tweak it or drop it. That is a Plan, Do, Study, Act cycle. Plot every build as a point on your run chart, and look at the whole line rather than the last number. When points fall below the baseline median, one after another, the change is working. A single fast build on its own tells you nothing yet. Keep an eye on your balancing measure. If the defects rise as the seconds fall, the change has only moved the problem. Builds 3 to 8 give you six cycles, one change each, at about two minutes a cycle. Keep what works, tweak what nearly works and drop the rest. Stage 5 is the last one: sustain, spread and share.
What to do
Complete builds 3–8: six small Plan–Do–Study–Act cycles, using a new customer order for each build. For each cycle, write ONE change and predict the seconds before Go. Afterwards record seconds, defects and the decision to keep, tweak or drop the change. Look at the full run chart, not only the most recent point.
Run each cycle in this rhythm: change and prediction → Go → Done → inspect → record → review → decide. Allow about two minutes per build, and keep checking defects.
Watch for: Avoid multiple simultaneous changes. Do not stop recording defects when the build gets faster; one fast point is not proof of improvement.
Done when: Builds 3–8 recorded with a change, prediction, defects and decision for each.
Stage 5 — Sustain, spread and share
Can another team repeat your result?
Read video narration
Stage 5: sustain, spread and share. Will your improvement last, and can it help others? First, make it standard work. Write the new way on a single page, with the layout, the roles and the sequence, so that a new team can follow it. Now swap tables. Give your page to another team. Can they match your time with a new order at the first attempt? If they can, the change has spread. Then share your run chart. Each team has one minute to show the baseline, the changes, the result and what did not work. Then back to the ward. You have just run all five steps. A real project uses the same steps on a real problem, with a run chart over weeks rather than minutes. Start your first real project with the Beginner kit. Good luck.
What to do
Write your new standard way of working on one page: layout, roles and sequence. Swap the instructions with another team. Ask them to use a new order and time one build without help. Then give a one-minute report showing the baseline, tests, outcome and what did not work.
Give two minutes for the standard-work notes. Swap between tables and time one independent build. Invite each team to share its full chart, including the changes it abandoned.
Watch for: Instructions that only their authors can follow are not yet standard work. Show the whole chart, not only the best build.
Done when: A standard-work summary, one swap build and a one-minute team report.
Record your baseline builds
0 of 9 builds recordedOutcome: seconds per build · Process: for example, parts laid out before Go · Balancing: defects per build.
Write instructions another team could follow without asking for help.
Builds 1 and 2 are baseline measurements: no changes between builds.
Run chart — seconds per build
Add two baseline builds to establish a baseline median. Later points will be plotted against it.
| Build | Order | Seconds | Defects | Change | Prediction | Decision |
|---|---|---|---|---|---|---|
| No builds recorded yet. | ||||||
Your notes stay on this device and browser. Use the printable score sheet for a shared or facilitated session; do not enter patient information.
From robots to better patient care
The robot is the exercise. Improving care is the goal. Use your results to reflect on what helped, what didn't, and what you would do differently in your clinical team.
Did speed and quality improve together?
Look at the run chart and defect counts. Was a faster build always a better build? What variation did you notice?
What did small tests teach you?
Which changes helped, which didn't, and what happened when you tested one change at a time?
Could another team reproduce your result?
Reflect on the swap build. Were your instructions clear enough for someone who hadn't developed the process?
What would make improvement last?
Think about standard work, shared ownership and continued measurement — not just one successful test.
Now apply the same thinking to acute medicine
Real improvement also starts with understanding a problem, agreeing an aim, involving the people who do the work, testing manageable changes and checking whether gains are sustained.
For example: If your team wants to improve the reliability of handover on the acute medical unit, start by understanding omissions and variation. Agree what good handover looks like, test one small change, and measure reliability alongside staff and patient experience.
The same five-stage learning journey, applied to a real care process.
Where will you take your next improvement idea?
Apply the Five Steps
Turn your idea into a small, measurable QI project using SAM's Quick Guide, Workbook and practical templates.
Explore SAM's QI guidesImprove together
Good QI needs the people who understand the process — different professions, patients and carers — to help design and sustain better care.
Review your team's results