One of the really useful charts that I commonly use with Kanban implementations are lead time distribution charts. Given that lead time is one of the key metrics, at least from a customers perspective, understanding the nature of lead times for your Kanban system is a must for ensuring a more fit-for-purpose and customer focused approach. Understanding how to read these charts and take action to achieve that kind of outcome is an essential skill for any kanban practitioner.

One of the key points about lead time is hinted at in the above title – it’s a distribution, not a single number. If you take nothing else away from this article except this point, it will put you in good stead to start to make improvements. Too often, I hear about people talking about “when it will be done” and referring to a single number, like an average. This is inherently disadvantageous, because it denies the real nature of lead times, especially in knowledge work.

No doubt, you’ve seen before a lot of variation in the knowledge work you’re involved with. Items are not all the same size or complexity, plus you have to wait for others which causes delays which are often unpredictable. These kinds of things lead me to prefer to acknowledge the variation in lead times we see, rather than ignore them.

Here’s an example of a lead distribution chart (also known as a histogram):

Here is an explanation of what you’re seeing:

  • X-axis: Number of days from when something was committed to, to when it was completed (ie the Lead Time). Not that in the above they are recorded at 5 day increments. So, the value for 10, represents the number of items greater than the last measure (5) and 10.
  • Y-Axis: Count of the number of items that fall within that range. The number at the top of each column represents the total count for that item.

You can see from the above chart that a significant number of the overall items fall within the 5 & 10 day counts. 43 out of the 78 items (over half) fall within this range. Clearly the majority of work falls within this range, but would you feel confident giving a commitment based on just a 55% hit rate? Of course not, because as we can see, a large number of items are still counted up to 25 days. A little over 80% of items (for the purpose of this, we’ll round up to 85%) take up to 25 days. This is a little better, and you may feel comfortable in giving a commitment based on this, but keep in mind that its up to 2.5 times longer than our earlier sample. Had we given a commitment based on what we thought was more commonly occurring lead time (which is often where our biases may take us) we might be facing considerable customer disappointment and the consequences that go with that.

There’s also a section between 30-50 days, not a great deal of items, but you can see that there is a not insignificant chance that we might land an item this late (about 10% of items). Again, this is 5 times what our first sample told us. In other words, 1 in 10 customers would have to wait 50 days instead of the 10 days if we had promised that. How would you feel as that customer?

Lastly, there are our outliers from 70-85 days. Outliers are a normal part of your system – you shouldn’t dismiss them out of hand. In our case, they’re about 5% of the total, or 1 in 20 requests.

When I look at the actual data behind this chart, the average lead time is approximately 16 days. Perhaps instead of jumping straight to commitments of 10 days, or even 16 days, we might like to take the full distribution into account. Instead, how when discussing time commitments, perhaps acknowledge the distribution with something along the lines of:

“Over half our requests are finished in 10 days or less, the average being 16 days, with 95% of all requests being completed in 50 days.”

This is a better discussion and uncovers the real nature of the work and will lead you to start to make better commitments that are more reliable.

Future posts will cover how to use the Lead Time Distribution Chart for making improvements.

The lead time chart was produced by the freely available FocusedObjective spreadsheets downloadable from: http://bit.ly/SimResources

Cumulative Flow diagrams can contain quite a lot of information for observations. However, they can also be a little daunting for those who are not familiar with them. In this blog post I’ll share with you some tips and information for how to start to use cumulative flow diagrams.

Here’s an example of a CFD:

Each line is a count of the number of tickets moving through that part of the process – stacked on top of one another, hence the name “cumulative”.

  • Light Blue – Done items
  • Yellow – Deploying
  • Grey – Testing
  • Orange – Developing
  • Dark Blue – Analysing

Along the x-axis are the days – it’s quite common for CFDs to use days as the point of measurement.

Along the y-axis is the count of the tickets. For example, on day 0 there were no tickets in the Done area, whereas on day 1 there were now 3.

That’s actually one key part of being able to read these kinds of graphs. The distance in height between the two lines represents the number of tickets. Thus the larger the gap between the lines, the more tickets that are in that part of the process. Therefore, the further apart the lines the greater the Work in Process (WiP).

If you get lines that are continuing to grow apart, the you have a WiP growth problem and this is going to slow down your overall throughput. This is particularly useful if you use the top and bottom lines when viewing WiP growth because these represent the input and output of your overall system. If you’re not getting growth in you system, but lines within it a growing, it means that there’s likely a bottleneck inside the system.

Additionally, where these lines start to converge, it means you have decreasing WiP and may lead to starvation of a part / all of the system. This might not be a problem in the immediate short term, but left unchecked it may lead to an unnecessary spend / costs.

Using these observations, you can start to adjust WiP limits and control the point of input to either prevent extra work coming into the system / parts of the system, or alternatively allowing more work to flow through. Effectively, this is what managing flow is about – trying to match variations in demand and capability to achieve a balance.

Another key observation about CFDs is that the average lead time is denoted by the width of the graph. For example, if you look at the above graph at the dark blue mark on day 1 – this is when item 11 entered the system. If you track horizontally across the chart, you’ll see that this item left the system (the light blue line) on day 7. To determine the system lead time, you can deduct the entry time from the exit time giving you a lead time of 6 days.

Of course, this time will vary as things change in the system. If you look further up the chart to day 7 and the analysis count there and then track across to done, you’ll notice the lead time is down to approximately 4 days. That’s about a 30% reduction in lead time and can often be attributed to actions such as limiting WiP in the system (as Little’s law tells us that decreasing WiP can also lead to decreased lead time) or other improvements to the system.

The third key attribute of this is the angle of the bottom line of the graph. This is known as the delivery rate or throughput of the Kanban system (as an average). The flatter the angle, the less the throughput. The greater the angle (moving towards vertical), the greater the delivery rate / throughput.

You can see these three key attributes in this chart:

You don’t need to always have electronic tools – if you have physical boards you can maintain a physical chart with this data. Alternatively, you can capture this in a spreadsheet. There are also other tools out there such as SwiftKanban, Kanbanize and LeanKit that do these things for you as well.

You can now start to use cumulative flow diagrams to help make more informed decisions about your system of work. You can see bottlenecks that you can action, you can start to limit or increase input to balance flow and you can limit work through the process to help smooth the flow. You can also start to make plans around how to reduce lead time or increase throughput which will have positive impact to your customers and your financial bottom line.

The cumulative flow diagram is an essential tool for managing flow and improving the outcomes of your service delivery. If you aren’t using one, you now have the basic knowledge to get started.

RETURN TO BLOG

Lead time scatterplots are an important tool for a team working with Kanban. Sometimes these are referred to as a “Control Chart” which is often the case in manufacturing. However, in knowledge work I think it’s less about the “control” mechanism and more about using it to improve the capability of the term providing the services. Thus I usually just refer to these as “scatterplots” and if you look at the example below you can see why.

There’s quite a bit going on here, so lets look at each part in turn.

X-Axis – This contains a series dates, you can see they progress over about an 8-9 month period moving from left to right chronologically.

Y-Axis – Time in calendar days. This is the amount of time a work item took when it was completed. This is calendar days, so it includes weekends and public holidays.

Orange dots – These represent Standard class of service items. Each time an item is completed, an orange dot is added at the date of completion and the total lead time for the item.

Grey triangles – These represent Fixed date class of service items. Similarly to the orange dots, they represent when an item is completed and the lead time to completion.

Orange dotted horizontal line – This represents the 95th percentile mark for Standard class of service items. That is, 95% of the work items fall under the line and 5% above. This line is optional, plus you can add lines at different percentile’s as per your needs (eg 85%).

Grey dotted horizontal line – This represents the 95th percentile mark for Fixed date class of service items.

You may notice a few things:

  • Lead times are not necessarily uniform. This is often the case in knowledge work – thinking everything is exactly the same is a common misconception
  • Sometimes there are gaps – there may be holiday periods where there is less work going on. You should understand this as being part of the normal cycle / capability of your team.
  • Some items take a large amount of time – Often items that are blocked or have external dependencies cause work items to have their lead time blow out.
  • There are much fewer Fixed date items than standard.
  • Fixed date items have a much lower lead time

Things to look out for

  • Growing lead times – this would tend to indicate an underlying problem in the system that you need to attend to
  • Where batches are completed together – there may be some bottleneck in the process preventing things from getting done, or the items might not be independent – look at how you slice for independence.
  • In progress items – this only shows you the “Done” items, it doesn’t show you the ageing of currently open items. You can use a separate chart for this – it’s important to action problems early before it’s too late!

You don’t need a whole lot of data for this to get started. All you need is the start date and end date of each of your work items. A few years ago I did a lightnight talk at Agile Australia 2017 that talks about this. If you’re looking for a tool to help you do this, you should check out Troy Magennis’ “Throughput and Cycle Time Calculator” excel sheet in his Focused Objectives repository.

It’s good to understand the basics about how these scatterplots work. I’ve touched on some of the ways they can be used, but there’s more detail to it than that and it requires a few other charts, so I’ll save this for a future blog post.

RETURN TO BLOG