What needs to be finished first
A concrete, measurable goal and a current baseline measurement exist, so improvement is observable.
What needs to be ready before start
Whiteboard or shared document with four quadrants (Plan, Do, Check, Act); data sources for measurement; template for hypothesis, experiment, evaluation, adjustment; timer for each cycle.
A PDCA owner as cycle owner; 2-5 people with execution and data access; a coach when introducing the method; a scribe for lessons learned.
Current baseline value of target metric; known hypotheses about levers; constraints (time, resources, risk); previous trials and their results; stakeholders who can confirm impact.
1-4 weeks per cycle, 60-90 min setup meeting, 30-45 min review meeting
Create quadrants for the current cycle. Hypothesis in Plan quadrant: "If we do X, then we expect Y, measured by Z by date." Define data source and success threshold. Fix review date directly during setup.
The one question this method answers
Which hypothesis do we test in the next cycle, which metrics determine success, and which adjustment follows from the evaluation?
Marker: Sektion
| Step | Duration | Action | Hint |
|---|---|---|---|
1Section 1: Plan | 60-90 min initial | Formulate hypothesis: "If we do X, then we expect Y." Name success criteria, define data source, set timeframe. Add actions, owners, schedule. Define risks and stop criteria. | If the hypothesis is "we want to improve" it lacks structure. Required form is If-then with measurable expectation. Without success criteria, check is impossible. |
2Section 2: Do | 1-4 weeks | Execute actions. Start small, observe, document, and record what actually happened (not only what was planned). Record deviations. | If execution is delayed, pause the cycle instead of falsifying data. A half-executed cycle is not learning. Better reduce scope than lose discipline. |
3Section 3: Check | 30-45 min after Do | Measure expected versus actual impact. Visualize data. Decide whether hypothesis is confirmed, falsified, or ambiguous. Identify side effects. Involve stakeholders if impact appears there. | A frequent mistake is treating Check as success reporting. If data disproves the hypothesis, this is valuable learning, not failure. Show data honestly, including unpleasant results. |
4Section 4: Act | 30 min, transition to next cycle | If hypothesis is confirmed: standardize and move into routine work. If falsified: derive new hypothesis. If ambiguous: sharpen methodology and repeat. | Act is not just 'continue doing it.' Standardization anchors learning. Without anchoring, PDCA has no effect in the next cycle. |
5Section 5: Next cycle | 60 min setup | Start next cycle with new hypothesis. Document linkage to previous cycle. Visualize trend across multiple cycles. | PDCA effect appears over multiple cycles. A single cycle is a test; multiple cycles are learning. Make trend visible. |
What comes out at the end
Cycle documentation with hypothesis, actions, data results, lessons learned, and standardization. One page per cycle with date, owner, participants, and status. Trend overview across multiple cycles.
One file per cycle, sequentially numbered. Standardization as separate entry with reference to the cycle. If a cycle is reopened with different outcome, create a new cycle number and keep predecessor unchanged.
Compact working template for PDCA Cycle with context, input, output artifacts, and next step.
Iterative improvement cycle made up of Plan, Do, Check, and Act.
When and for what do we use this method?
Which data, observations, decisions, or materials are available?
Short notes along the runsheet.
Owner, date, and success signal.
Concrete filled scenario, fictional example
Concrete filled scenario, fictional example
Owner: @lisa. Baseline: 32% drop-off between Day 0 and Day 1.
Hypothesis: If we show a personalized success message with a concrete next step at the end of Day 0, then Day 1 drop-off falls below 25% by 24.06.2026, measured via Amplitude cohort. Actions: Build component (engineering 5d), QA, A/B rollout 50/50. Stop criterion: if drop-off in variant rises above 35%.
10.-17.06.: component live in A/B test, 50/50, n=2,847.
Day 1 drop-off Variant A with component: 23.4%. Variant B (control): 31.8%, difference significant (p < 0.05). Side effect: positive NPS quotes about component, no negative impact on Day 7 activation.
Variant A is standardized, rollout to 100% on 25.06. Standard Work entry added to onboarding docs, documented by @lisa.
Cycle 5: Hypothesis that industry-specific personalization yields a further 5 percentage points. Setup meeting 26.06.
Recognize symptoms and steer against them
Plan contains actions but no measurable expectation with number and date.
Require if-then form with measurable expectation. If no metric is available, pause the cycle until measurement is set up. No check without measurement.
Data disproves hypothesis, team softens with alternative interpretations or shifts threshold.
Set threshold before Do and freeze it. Falsification is valuable learning, not failure. Document clearly and design a new hypothesis in next cycle.
Confirmed hypothesis remains individual knowledge, not embedded in standard work.
Act step must include integration into doc, process, or tool. Assign owner for standardization. Without this layer, PDCA impact is lost.
Plan phase takes 2 weeks, Do takes 6 weeks; method turns into mini-waterfall project planning.
Limit Do to 1-4 weeks per cycle. If longer is needed, split into sub-hypotheses. Smaller cycles learn faster.
Method runs once, no follow-up cycle, no trend, no learning.
Plan cycles as a series of at least 3-5 at setup. Learning comes from repetition, not one-off execution.
Done signals checkable in under a minute
Go to the profile for purpose, similar methods, and sources or continue to the next method in the catalog.