Working Framework · DOCX · Czech
PDSA: Change Validation — DOCX
Extended review template and working prompts for evaluating changes.
How to verify that a change helped
Before making a change, write down what you expect and what you will observe. Then test it on a small scale, compare findings with expectations, and decide on the next step. This is served by the PDSA cycle: Plan, Do, Study, Act.
Prepare a comparison
Select one change and one primary problem it aims to solve. Record the current state and how you assess it. Observe the trial run in the exact same manner. Beside expected benefits, pay attention to any new difficulties the change introduces.
You need a description of the change, expectations, and work logs. The output will be a brief decision with rationale and noted uncertainties. If the change already occurred and baseline records are missing, acknowledge this. You can describe current operations and prepare the next trial; do not retroactively fabricate exact before-and-after comparisons.
Follow the four steps of PDSA
- Plan. Record expected results, scope, and trial end. Agree on who executes the change, what will be logged, and when to pause the trial.
- Do. Test the agreed change. Record actual progress, deviations, and unexpected issues.
- Study. Compare logs with expectations. What improved, what worsened, and what remains unknown?
- Act. Based on findings, continue testing, adapt the change, or revert. Determine one next step.
Narrow scope helps during initial testing. Before expanding a change, verify it repeatedly under varying conditions. The cycle description builds on the Institute for Healthcare Improvement: Testing Changes methodology.
Example: is the project reference number easier to find?
Model numbers, not actual project results: during five handoffs of the same type, a colleague had to look up the project reference number three times. The team adds a requirement to the instructions to include the number in the folder name.
Prior to the trial, they state: "Over the next five handoffs, we expect at most one reference number lookup. We will also log any naming ambiguities." The designated colleague records the outcome of each handoff; if a project swap occurs, they pause the trial and resolve the issue.
In the model trial, lookup was needed once. However, two people asked how exactly to format the folder name. Expectations were met in this small sample, but a new ambiguity surfaced.
Decision: add one example of folder naming and run another small trial. It cannot yet be said that the adjustment will always work or that the difference was caused solely by it. Results may be influenced by project type or team experience. Five cases serve as initial learning, not a statistically conclusive proof.
Review logs with AI
Paste expectations and actual notes from the trial. Remove personal and confidential data before pasting.
Help me evaluate a single work change using the PDSA methodology.
Base your output solely on the provided records. Separate observations, estimates, and unknown data.
Do not invent measurements, baselines, or retroactively set expectations.
Briefly compare the plan, actual execution, and outcome.
Highlight new difficulties and differences in comparison conditions.
Propose whether to continue testing, adapt, or revert the change, and explain why.
If inputs are insufficient, state what cannot be decided and what needs to be investigated.
Conclude with one small next step. Do not execute anything or guarantee effectiveness.
CHANGE & PROBLEM IT AIMED TO SOLVE:
[Fill in]
PRE-RECORDED EXPECTATIONS & TRIAL PLAN:
[Fill in, or state missing]
BASELINE STATE & ACTUAL EXECUTION:
[Paste records, or mark missing data]
OUTCOMES, SIDE EFFECTS & CONDITION DIFFERENCES:
[Fill in]
Verify the conclusion
- Is it clear what was an expectation and what was an actual observation?
- Are you comparing similar situations in the same manner?
- Have you accounted for new difficulties and missing data?
- Does the next step align with what you actually uncovered?
AI can help structure records and highlight gaps. Verify record accuracy and conclusions with the people who performed the work. Even an unsuccessful trial can show what makes sense to adjust next time.
Next steps
Explanation & prompt
How to break a change into small steps
Define a single step that can be completed and verified.
Need to map the whole process first? Explore the guide Process first. Then AI.