Kaizen Sessions: Awareness for Improvement (Team Limitless)

*In this article, which is based on real data and experiences, the team name has been changed intentionally.

Team Limitless

We started to work with Murat from Team Limitless as of May 2017, after he received his first Kanban (Team Kanban Practitioner) training on April 27, 2017. Then he received the first education (KMP I) to get the Kanban Management Professional Certificate on October 12, 2017.

He is a person whom I enjoy working together with, who makes me happy with his efforts to better his system and with whom I learn. I’d like to thank him.

What I will write here is to provide an entirely outside perspective and get comments from other viewpoints. Although I try to comment, it will be incomplete. What is important is to help the team build an environment where they can discuss better; to create a feedback mechanism. The primary goal is to see the team support its own decision-making mechanism and make the right decisions with these graphs.

Summary

This report has been written to examine the following situations.

• How did the delivery rates increase by 7 times as of May 2017 and by 2.5 times as of June 2017?
• How were the works started to be completed in 5 weeks while they were completed within 10 weeks (80% probability)?
• Why no bugs have been reported/opened since September 2017?
• How has the team been able to increase their work capacity or balance the increased workload compared to their previous situation?
• Most importantly, how did they make leading decisions or manage other units to improve their systems?

When we compare the values obtained from the customer’s and the team’s perspective, we can foresee that the system will not improve only with improvements on the team. Improvement will be achieved with a more coordinated operation of teams or turnaround points within the system.

I believe there are two types of team models with my current experience;

1. To be a team to strictly and immediately fulfill the tasks and responsibilities assigned.
2. To be a team to fulfill the tasks and responsibilities by questioning what should be in the entire system, by measuring its effect, and proposing innovations.

Unfortunately, the most popular team model is the first one. If we want to move forward with an innovative perspective (Kaizen) I think we need to have a team model as stated in the second type.

I think that the Team Limitless achieved this by giving logical feedback from the project management to the departments they work with and by making themselves heard.

Metrics should be used as auxiliary tools that enrich the conversations without replacing the conversation. Metrics should be used as a common language between two foreigners.

You can send me anonymous comments from: sayat.me/intparse

If you share your e-mail address while commenting, I can reply on that page without seeing your e-mail address.

Details are below. Thank you in advance for reading.

—

Alper Tonga

Demand Competence Review (from the Customer’s Perspective)

The chart above shows us the values for the works started and delivered between January and November 2017. The reds represent works started and the greens represent the works delivered. The bar chart shows how many works have been started in that month (Ready 2 Prepare) and how many works have been delivered (Delivered). Example: 9 works started in May and 2 were delivered. The line graph shows the rates. Example: 0.42 works per day started in August (approx. 1 works in 3 days) and 0.39 works delivered per day (approx. 1 works in 3 days).

I marked some of the important events/days on the chart.

The question we want this graph to help us answer is; Can the system meet incoming requests at the desired level?

• If the red line (number of works started) is moving a little above (a little is a relative unit) the green line (number of works delivered), the system works without starvation. It means new works are pending.
• If the red line (number of works started) is well above the green line (number of works delivered), we can say that the system may be crushed under the incoming works demand.
• If the red line (number of works started) falls well below the green line (number of works delivered), we can say that the demand started to decrease, or the team can meet the demands that could not be met before by increasing the competence or the project nears the end.

Observation

Considering that the data were collected in May, it is better to make our comments as of June. Accordingly;

• Works delivered doubled between June 2017 and August 2017. It was 0.2 per day and increased to 0.4 works per day. (A 7-fold increase compared to May data.) This is a good increase if we consider that the first two weeks of August were holiday.
• It is observed that the speed of the team’s deliveries has decreased with the increase in works and abnormal work demand in September. (See Queueing Theorem: Little’s Law: as the number of works started increases, it takes more time to finish the work and work completion rates decline.)
• We can see that the team’s deliveries were affected by the Freeze in November.
• In November, we can see an increase in the start of works as in September.

It is more important to see how the team takes measures according to the circumstances.

These were the issues that concern the entire system. Let’s take a look at the factors in the team while these factors took place.

Demand Competence Review (from the Team’s Perspective)

The chart above shows the values about the items that were started to be developed and the items that were ready for test or delivery between April and November 2017. The reds show items that were started to be developed and the greens show the items that were ready for test. The bar graph shows how many demands were started to be developed in that month (Ready 2 Start) and how many items were ready to be tested (Ready 2 Accept). Example: In August, 14 demands were started to be developed and 10 items were ready for the test or delivery. The line graph shows the rates. Example: In August, 0.45 items were started to be developed per day (approx. 1 item in 3 days) and 0.32 items were ready to be tested per day (approx. 1 item in 4 days).

Observation

• The team managed to keep the rate of items ready to be tested constant with the WIP limitation since June. It was 0.30 items per day in June and it rose to 0.37 items per day in November.
• Despite the development increase that started in August, they tried not to compromise the team output rates and to keep them stable. In this way, they may have preserved their predictable system structures.
• We can see from the WIP chart below that they had to increase the WIP values after September to meet the demand, but team output rates still remain constant. This means in-team competence increased.

So, what happened when they do that? Let’s look at the quality work demands, generated and solved defect rates.

Quality Demand Competence Review (from the Customer’s Perspective)

The chart below shows us how many percent of the works started and delivered between January and November has defect record (bug). The reds show that the percentage of defect records (bug) of the works started and the greens show the percentage of defect records (bug) of the works delivered.

IMPORTANT: Any increase or decrease in these percentages does not imply an increase in bug records. If 3 of the 6 works started in a month has bug records, this makes up 50% bug rate and 2 of the 3 works started in the following month has a bug record of 66%. In such a case, it would be wrong to conclude that bug records increased.

Observation

• While the percentage of bug records decreased during the months of April – June, the highest percentage of bugs were recorded in works started in August 2017.
• The percentage of defect records decreased in works delivered between May and July. We can say that effort on bug records declined and more emphasis was put on new functions.
• Bug records reported since September 2017 were reduced to zero. This may be due to the team’s ability to increase the quality of the works with higher competence, or we can also say that no bug reports came from business units.

We can improve our review by showing the daily rates of the works started and delivered in the same chart in order to understand whether the bug records increased or not. The following graph illustrates that. Look at the dashed lines.

Balanced Work Distribution Review

The chart above shows us the process and occupancy of the system on a monthly basis between April and November. We can see from this chart whether the demand flow is accelerated/slowed/stopped (bottleneck) or if there is a balanced distribution.

Observation

• It can be said that there has been a continuous increase in the number of works delivered between May and August (dark green zone). This increase slowed down in September, gained momentum in October, but stopped in November with freeze.
• We can make a better comment when we match the notes ​​in the chart with the values. The new project in September (gray zone) shows that priority might have been given to new jobs and caused a decline in deliveries between August and September.
• We see that works ready to be tested (blue zone) finished and works ready to start (orange zone) has begun to increase between July and August. This may be the reason for the decline in works delivered (dark green) between August and September. It is also interesting that the works ready to deliver (light green zone) increased without being affected. For a good comment, as always, we need to consult the team. Team interpretation: This result was obtained as the team observed a stoppage in works tested and decided to prioritize the works that did not need external testing, in order not to adversely affect the output speed.
• We observed that the percentage of the works ready to deliver (light green zone) is on the rise without being affected by the new project in September and the freeze in November.
• This is good; this means the team never stayed idle. In other words, team members did not wait for orders to work; they have a pro-active team structure rather than a reactive one.

Monthly Lead Times/Capacity/Output Review (from the Customer’s Perspective)

The chart above shows us the monthly lead times between April and November (average/75th percentile/85th percentile Lead Time) and the number of works that were started but not finished (in purple) as well as items delivered. A decline in lead time values shown in green, blue, and orange lines indicate faster output. The fact that lead time values are close to each other shows that stability and predictability are high. The dashed purple line indicates the number of works delivered in that month, and the purple line shows the number of works that were started but not finished (WIP: work-in-progress).

Observation

• As of November, the lead time is about 25 days. The fact that lead time values ​​get closer to each other can be pointing towards a more stable and balanced system.
• When we look at the whole chart, we observe that lead times reach 25 days in June too, but there was a disintegration (in July) and this disintegration started to recover as of August. Note: Such a disintegration may be the result of the completion of old jobs. When the works that were not given enough priority/delayed due to the time distress caused by the workload on the team are finalized, the lead times are prolonged and cause such a disintegration/fluctuation in the graphics. If we have any backlog, these disintegrations will repeat at regular intervals. The team may have had time to focus on their accumulated works (cleaning) by creating slack time as a result of observed improvements, new policies or better understanding of the system as from April. And, as a result of this focus, the team turned a new page in August after the disintegration in July.
• As far as team capacity is concerned, the number of works in progress between April and May (WIP) ranged from approximately 8 to 9, and this figure went up to 12 from June to August. The fact that the number of works delivered in this period (dashed purple lines) is always increasing is an indication that the system (not only the team) has increased competence. I can say with confidence that Team Limitless leads the competence and capacity increase factors.
• We can observe how dramatically the number of WIP reached 25 with the new project in September and how this increase affected the number of outputs. This might be because the team panicked and immediately started analysis when the new project came. The more works are started, the longer the lead times will be. (See Queueing Theorem: Little’s Law)
• We can see that the system is trying to recover itself in October, but the negative impact of instability can still be observed from the customer’s perspective with the incoming freeze.

Let’s see what is going on inside the team in the meantime.

Monthly Lead Times/Capacity/Output Review (from the Team’s Perspective)

The chart above shows us the monthly lead times between April and November (average/75^th percentile/85^th percentile Lead Time) and the number of items that were started but not finished (in purple) as well as items delivered.

In the previous chart, we reviewed our observations from the customer’s perspective, and we will review them from the team’s perspective here.

Differences;

• The lead-time values show the time between the start of analysis and delivery from the customer’s perspective, while they show the time between the start of development and ready to deliver in this graph.
• The number of works in progress (WIP) is counted from the beginning of the analysis from the customer’s perspective, while the works that are started to develop and come to the test phase are counted as WIP in this graph.
• The number of outputs is the number of works delivered from the customer’s perspective, while the number of outputs (dashed purple line) shown in this graph is the number of works sent to test or ready to deliver.

Observation

• The period when the Team Limitless is the most stable and fast is August, works are ready to deliver in about 5 days.
• The disintegration we saw from the customer’s point of view in July was one month before, in June, within the team. The cleaning they made in June influenced the next delivery, shaking the lead time values. Of course, they triggered long-term recovery by doing this cleaning in advance and reducing the amount of work waiting for a long time. Personal Opinion: Sometimes we cannot see the long term to save the day and put ourselves in difficult situations. The fact that Team Limitless assumes its own responsibilities is something that needs to be appreciated. At this point, leadership comes to the fore; prove something and lead the path.
• The fixed number of outputs (dashed purple line; throughput) despite the increasing workload between August and September (purple line; WIP) indicates that the competence of the team has increased. However, we can see that lead time values increase and the gap between the 75^th and 85^th percentiles increased due to the increasing WIP caused by the pressure of the new project. Little’s Law shows its impact here: if you start new works before you finish the old ones, you will need more time to finish. Personal Opinion: I wonder the results and inferences of Team Limitless, which saw this result but had to progress any way. All in all, the team will have valuable experiences if it observes the situation and becomes the master of its own destiny instead of drifting around. I hope that others can learn a lesson from these inferences. This is why I’m very curious about the December data.

Last 9 Months Lead Time Histogram Review (from the Customer’s Perspective)

The chart above shows the lead time distribution of the works delivered within the last 9 months in weekly percentiles. Example 1: 76% of the works delivered between April and December were delivered within 6 weeks. The keyword here is “within”. We can say that 76% of the works delivered did not took 6 weeks but they took maximum of 6 weeks. Example 2: If we put all the items delivered in a bag and choose one randomly, we are 76% likely to choose an item that took maximum 6 weeks or less to be delivered.

Using this distribution, we can make a quick estimation about how long the new works will take. Sharing the distribution with the stakeholders within the system shows possible project deadlines by giving information about how long the works will take (in the first stage before the detailed analysis is made) and about the risks. I can say that this supportive method is better than promising a deadline without knowing anything. (This approach was used in history. See German Tank Problem, Troy Magennis; Percent Likelihood)

I can tell my customer who is asking when a new request will be delivered by using this distribution: Note: This distribution chart shows the distribution from the customer’s perspective, which means that it does not include only the values of Team Limitless; the whole system (analysis, testing, delivery) is here.

• There is a 76% chance that it will be delivered within 6 weeks (~1.5 months).
• There is an 85% chance that it will be delivered within 13 weeks (~3 months).
• There is a 94% chance that it will be delivered within 17 weeks (~4 months).

This distribution is a histogram, i.e. a history. Looking at a long range of 9 months may not produce meaningful results due to the rapid change of work environment, teams or projects. Therefore, it is best to look at the last 3 months or a specific period based on the structure of the system. If the teams have changed or started a new project and the platform of this new project is completely different from the previous project (such as the transition from a Java project to a .NET project), it is useful to make an estimation in the light of new data without using the old data.

The classification of works based on the stages they go through or the way they are handled (classes of services) may further strengthen this estimation. Classifying by size, as seen in the literature, will not provide an advantage. Example: You have a small project and regular tasks in your work list. When the small project continues, a big project is sent to you, and you are instructed that it should be finished as soon as possible. Although the small project is already underway, it stops, and you focus on the big project. The big project is finalized in a month with extraordinary effort. Small project is completed with a delay of 1 month. Do you think that the size of the work plays a big role in the completion of the works faster?

Another point is that the longer the line extending towards right, the more unstable and unbalanced our system will be. If we want to shorten that line, we need to limit our works. (Limit WIP)

Let’s try to interpret what the team Limitless did by looking at the 3-month (quarter) distributions. Thus, we can see if there is improvement by using distributions.

Q2 Lead Time Histogram Review (from the Customer’s Perspective)

The chart above shows the distribution of the lead time of the works delivered within the second quarter (April – June) in weekly percentiles.

Observation

• Customer or stakeholder did not have any works delivered in 1 week. The works could be delivered after waiting for at least 2 weeks. And the works delivered in these 2 weeks make up 10% of the works finished in the second quarter.
• 70% of the works completed in the 2^nd quarter were delivered within 5 weeks; 80% were delivered within 8 weeks, and 90% were delivered within 14 weeks.
• Since deliveries are monthly, it can be a good method (fitness criteria) to expect more works delivered in 4 weeks to produce more value and to work effectively. We can see that there is a 50% chance that works can be delivered within 4 weeks. Personal Opinion: 50% chance like flipping a coin. Instead of spending time planning, we can flip a coin as well. Starting the works that I cannot finish will slow me down and decrease the value I produce. Thinking about how to improve quality and how to prevent waste rather than aiming 100% system occupancy rate (trying to keep everybody busy) will increase system efficiency. Instead of focusing on 100% project work, 70% project and 30% improvement/personal training/experiment can be a more effective capacity distribution.

Q3 Lead Time Histogram Review (from the Customer’s Perspective)

The chart above shows the distribution of the lead time of the works delivered within the third quarter (July – September) in weekly percentiles.

Observation

• Our customer started to see works delivered within 1 weeks. And these works account for 11% of works finished in the third quarter. The chance of the works delivered in 2 weeks increased to 33%.
• In the 3rd quarter, 70% of the works were delivered within 6 weeks (It was 70% in 5 weeks); 81% of the works were delivered within 10 weeks (It was 80% in 8 weeks); 89% of the works were delivered within 17 weeks. (It was 90% in 14 weeks)

Lead times started to increase compared to the previous quarter. This may be due to the fluctuation in July caused by the completion of previous works, which we mentioned in the previous reviews.

• Our hypothesis is supported by the increase of our line to 26 weeks and its increase compared to the second quarter. 26 weeks means about 6 and half months. Considering that we are looking at a 3-month period, it is reasonable to say that previous works were cleaned. We can see the benefits of this cleaning in the long run. We will see the benefits of this cleaning in the long term (in the next quarter).
• There are still monthly deliveries and if we look at the possibility of getting works delivered in 4 weeks, there isn’t much change with a value of 52%. In other words, we can say that the effects of changes made in the system on monthly deliveries remain the same. The experience was not noticed in the system, but it had a valuable impact on the Team Limitless. In addition, these experiments help determine the improvements that will have long-term returns in the system. Personal Opinion: These changes made by Team Limitless within the system create a good experience and improvement environment with Kanban as they are evolutionary and minor changes. Performing a controlled experiment within the system rather than attempting experiments that will have a big effect in the organization creates win-win situation for both parties.
• We can also see that the number of works in the 3rd quarter is higher than the 2nd quarter.
• Number of works delivered in Q2: 10
  Number of works delivered in Q3: 27
• These data can be found in the Demand Competence Review in the first part of the report.

Q4 Lead Time Histogram Review (from the Customer’s Perspective)

The chart above shows the distribution of the lead time of the works delivered within Q4 (October – December) in weekly percentiles. I guess it’s not a coincidence to see this improvement in the system.

Observation

• It is important to remember that the system was in Freeze since November. This Q4 chart actually only contains October. It is important for the system stakeholders to know how the work to be delivered at the end of December will affect this chart. The effect of this change on a system that has begun to gain stability and balance will be negative on the distribution and thus on the foresight.
• In the 4th quarter:

71% of the works were delivered within 4 weeks (70% was delivered in 6 weeks)

82% of the works were delivered within 5 weeks (81% was delivered in 10 weeks)

88% of the works were delivered within 6 weeks. (89% was in 17 weeks)

• The highest lead time the system is 13 weeks. The cleaning of previous works, WIP limitation, and other decisions taken by the team may suggest the beginning of improvement. We can say this is because Team Limitless took responsibility of the system together and to carried out the necessary recovery (Kaizen) actions to maintain that. Of course, Team Limitless can make better comments and show data.
• We can see that the number of works delivered in October is 62% of Q3. What would happen if there was no freeze in November?
  Number of works delivered in Q2: 10
  Number of works delivered in Q3: 27
  Number of works delivered in October only: 17

These data can be found in the Demand Competence reviews in the first part of the report.

What was the change in the team when all these developments were in the system? Let’s look at these distributions from the team’s perspective.

Last 9 Months Lead Time Histogram Review (from the Team’s Perspective)

The chart above shows the lead time distribution of the works sent to test or made ready to deliver within the last 9 months in weekly percentiles. Example 1: 80% of the items sent to test or made ready to deliver between April and December were sent to test or made ready to deliver within 2 weeks. The keyword here is “within”. We can say that 80% of the items sent to test or made ready to deliver did not last for 2 weeks but they lasted maximum of 2 weeks. Example 2: If we put all the items sent to test or made ready to deliver in a bag and choose one randomly, we are 80% likely to choose an item that took maximum 2 weeks or less.

Using this distribution, we can make a quick estimation about how long the new works will take. Sharing the distribution with the team members within the system shows possible project deadlines by giving information about how long the works will take and about the risks. With this supportive method, the team can compare how big or small the responsibility is.

I can tell my demandant who is asking when a new request will be ready to test or deliver by using this distribution: (Note: This distribution chart shows the distribution from the team’s perspective, which means that it does not include the times of analysis team or test team.)

• There is an 80% chance that it will be ready to test or deliver within 2 weeks (In 6 weeks from customer’s perspective)
• There is an 87% chance that it will be ready to test or deliver within 4 weeks (In 13 weeks from customer’s perspective)
• There is a 94% chance that it will be ready to test or deliver within 6 weeks (In 17 weeks from customer’s perspective)

When we compare the values obtained from the perspective of the customer and the team, we can foresee that the system will not improve only with improvements on the team. Improvement will be achieved with a more coordinated operation of teams or turnaround points within the system.

The classification of works based on the stages they go through or the way they are handled (classes of services) may further strengthen this estimation. Classifying by size, as seen in the literature, will not provide an advantage. Example: You have a small project and regular tasks in your work list. When the small project continues, a big project is sent to you, and you are instructed that it should be finished as soon as possible. Although the small project is already underway, it stops, and you focus on the big project. The big project is finalized in a month with extraordinary effort. Small project is completed with a delay of 1 month. Do you think that the size of the work plays a big role in the completion of the works faster?

Another point is that the longer the line extending towards right, the more unstable and unbalanced our system will be. If we want to shorten that line, we need to limit our works. (Limit WIP)

Let’s try to interpret what the Team Limitless did by looking at the 3-month (quarter) distributions. Thus, we can see if there is improvement by using distributions.

Q2 Lead Time Histogram Review (from the Team’s Perspective)

The chart above shows the distribution of the time that an item takes until it is sent to test or made ready to deliver within Q2 (April – June) in weekly percentiles.

Observation

• In Q2,
  60% of the works were made ready to test or deliver within 1 week,
  80% of the works were made ready to test or deliver within 2 weeks.
• 100% of the works were made ready to test or deliver within max. 5 weeks

Q3 Lead Time Histogram Review (from the Team’s Perspective)

The chart above shows the distribution of the time that an item takes until it is sent to test or made ready to deliver within Q3 (July – September) in weekly percentiles.

Observation

• In Q3:
  59% of the works were made ready to test or deliver within 1 week (It was 60% in 1 week)
  78% of the works were made ready to test or deliver within 2 weeks (It was 80% in 2 weeks)
  85% of the works were made ready to test or deliver within 4 weeks
  96% of the works were made ready to test or deliver within 8 weeks (It was max. 5 weeks)

Lead times started to increase compared to the previous quarter. This may be due to the fluctuation in July caused by the completion of previous works, which we mentioned in the previous reviews.

• Our hypothesis is supported by the increase of our line to 12 weeks and its increase compared to the second quarter. 12 weeks means about 3 months. Considering that we are looking at a 3-month period, it is reasonable to say that previous works were cleaned. We will see the benefits of this cleaning in the long run (in the next quarter).
• We can also see that the number of works ready to test or deliver is more in Q3 compared to Q2.
  Number of works delivered in Q2: 15
  Number of works delivered in Q3: 28

This data can be found in the Demand Competence reviews in the first part of the report.

Q4 Lead Time Histogram Review (from the Team’s Perspective)

The chart above shows the distribution of the time that an item takes until it is sent to test or made ready to deliver within Q4 (October – December) in weekly percentiles.

Observation

• The fact that the system has been in Freeze since November does not prevent the Team Limitless from making works ready to deliver. Therefore, this distribution includes the values ​​obtained in 3 months. It does not only include “October data” as it did from the customer’s point of view.
• In Q4,
• 76% of the works were delivered in 1 week (It was 59% in 1 week)82% of the works were delivered in 2 weeks (It was 78% in 2 weeks)94% of the works were delivered in 4 weeks. (It was 96% in 8 weeks)
• The maximum time to make an item ready to test or deliver is 5 weeks. The cleaning of previous, WIP limitation, and other decisions taken by the team caused this improvement. We can say this is because Team Limitless took responsibility of the system together and to carried out the necessary recovery (Kaizen) actions to maintain that. Of course, Team Limitless can make better comments and show data.
• We can also see that the number of works ready to test or deliver in the 4th quarter is less than the 3rd quarter (although the 4th quarter consists of October and November).

Number of works ready to deliver in Q2: 15

Number of works ready to deliver in Q3: 28

Number of works ready to deliver in Q4 (October and November): 22

This data can be found in the Demand Competence reviews in the first part of the report.

Thank you for reading and interpreting all the comments through 15 pages. Let’s never forget that we can still get the best and most realistic comments from the team itself.

Metrics should be used as auxiliary tools that will enrich the conversations without replacing the conversation. Metrics should be used as a common language between two foreigners.

For deeper comments, it is worth talking to the Team Limitless. Team Limitless has more detailed metrics, and they can also tell others about their capacity and what competence they need.

You can send your anonymous comments at this address sayat.me/intparse

— Alper Tonga