Denise continues to think about the three types of experiment and wonders what would the experimenting record look like for each of the three? She draws a small spreadsheet with the two columns obstacle and next step.
Since Denise and her team leaders, Joe and Mark, started distinguishing between an implementation obstacle and a process obstacle, they had decided to only write down the currently addressed process obstacle on the experimenting record. They had also developed the habit of drawing a dash in the obstacle column as long as the next process obstacle that required addressing was unknown. In such a case, the next logical step would be a type one experiment — find process obstacles to address, Denise thought. She filled out the first row in her spreadsheet.
In most cases, the next step for finding obstacles was a more detailed process observation. Or a trial run simulating the desired process pattern described in the target condition such as working with one-piece flow, zero buffers, or one operator less.
Running the process like that, of course, did not work perfectly but made obstacles clearly visible. This had become a trick they used quite frequently to understand better what they had to work on next to get closer to the next target condition. It was a lot better than assuming and discussing what the obstacles might theoretically be.
That was also why Denise had rephrased question three on her card with the Coaching Kata questions. She had the “do you think” part. If the answer was, “I don’t know for sure” or Joe and Mark came up with hypothetical obstacles, she often insisted on running the process with the desired process pattern to reveal real and relevant obstacles.
The process obstacles they identified that way were put down in the obstacle parking lot. Then they selected one of them to address and wrote it into the obstacle column in the next row of the experimenting record.
The next step then logically had to do with understanding the cause of the chosen obstacle. Typically, that required another observation, only more accurate. Regularly, several steps with test runs and measurements were needed until the actual cause became apparent. Denise filled out the next row in her little 2×3 spreadsheets.
Once the cause was clear, a type three experiment was needed. The obstacle remained the same, but the step now was something like 'test solution ABC to remove the obstacle.
After Denise had filled out the third row accordingly, she thought, what about the expectation in these three cases? She added a third column to her little sketch.
For a type 1 experiment, finding obstacles, the logical expectation was something like obstacles are known. Ideally including the unwanted effect of each obstacle, Denise thought as she noted the expectation on the table.
So, for type 2 experiments, an observation, or a test to investigate the cause of an obstacle, the expectation was, of course, we know the root cause. Just like this morning, Denise thought. This is precisely how Joe had formulated the expectation for his next step. And that step was a type 2 experiment, as Denise was pleased to recognize. Her little table seemed to work well.
A type 3 experiment, testing a specific solution to remove an obstacle, therefore had the expectation, process metric changes by value x and not just, solution works. Only when there was a measurable effect on the process metric had the obstacle indeed been removed.
Looking at her sketch, Denise suddenly has an idea. If I know what the understanding of the obstacle is, I can predict what my improvers answer to question 4, the next step and expectation, should be. Again, it’s about finding the threshold of knowledge. What is it we don’t know about the obstacle, she thought.
If no obstacles have been identified yet, the next step should be an observation or test-run with the expectation to find obstacles and understand their magnitude.
If an obstacle has been identified, the next step should be an analysis with the expectation to understanding the cause.
And if the cause is clear, the next step should be to test a change to remove the obstacle, expecting a positive change in the process metric if the obstacle is removed.
This is cool, Denise thought. Now I know in advance what the answer from my improvers regarding next step and expectation should be. If their answer is different, I can ask deepening questions to help them to clarify for themselves. Then they can come up with a more precise step and logic and measurable expectation.
It’s a little like being able to foretell the future, Denise thought and started smiling. She called her new trick: The fortune-tellers crystal ball.