Plan-Do-Check-Adjust(Act) Cycle. Make a plan, execute as planned, check the result and decide for the next step. Darn simple – in principle. Question is why so few people actually use the cycle. don not mistake the P-D-C-A Cycle as „trial and error“. Trial and error works without the input of intelligence. When you do not plan, you will move into the train of trail and error, no matter how intelligent you are. Experiment without plan (target result) is trail and error.
Let us have a look at learning by experimenting:
- Plan: Making a plan is always the pre-requisite for anything – unless you are planning to just fool around and play. “Fail but fail fast” is a well-known mantra. There is no fail if there is no plan. Fail is deviation from expected result. No expected result, no learning. Very simple. This is the first possible pitfall. This also separates us from just playing around and making accidental discoveries.
- Do: We have a plan and execute. Stick to the plan, otherwise you will see a deviation from the expected result, but have no clue where the deviation originates in. There is one thing however: if you made a plan and you notice things really go wrong while you go. You expect to experience severe pain if you continue and you just cannot stand to continue. Then you observe a symptom of a problem. The problem is that the cycle you planed is too long. You go too far without possibility to correct. Now, just don’t be stupid. Stop the cycle, go to step 3.
- Check: After the planned time interval (a.k.a. timebox) compare actual result to expectation. If it matches, congratulation, your model of your system is good. Unfortunately, you only learn that. All under control, you can continue. If the result deviates from expectation, you now have the chance to generate new knowledge. You made a prediction, but it did not manifest. Something else happened. This has to potential to generate new knowledge. Use it.
- Adjust/Act: Some time has passed since you planned the last step. Check the environment if the situation has changed. It may have. If necessary, adjust the target condition (refer to Mike Rother’s Toyota Kata). When the target condition is either still good, or adjusted, go back to step 1.
There is a lot of information on PDCA-cycles on the internet. The originators are Walter A. Shewhard and/or W. Edwards Deming. Some people trace the idea back to Francis Bacon.
Just as in extreme programming, planning is a continual process.
Classical pitfalls thus are:
- Not making a plan.
- Not sticking to the plan.
- Not checking the results.
- Not comparing the results to the expectation.
- Not analyzing the observed deviation from the expectation.
- Not having an expectation.
- Not reviewing the target.
- Planning too big steps.
- Instead of working iteratively, having a list of x steps in an “if … then …” or “if … then … else …” structure.
- …
“Fail but fail fast” simply means: the smaller the steps you make are, the faster the learning can be.