Skip to main content

4 Key Flow Metrics and How to Use Them in Scrum's Events

May 10, 2018

Metrics in Scrum Events

In the Kanban Guide for Scrum Teams and the Professional Scrum with Kanban workshop, we introduce 4 key flow metrics that we believe Scrum teams can use to improve their flow:

Work in Progress (WIP)

The number of work items started but not finished (according to the Scrum Team's definition of "Workflow").

Note the difference between WIP and the WIP Limit. The WIP Limit is a policy which the Scrum Team uses as a "constraint" to help them shape the flow of work. The goal of the WIP Limit is to reduce the amount of actual work in process (WIP). The team can use the WIP metric to provide transparency into their progress towards reducing their WIP and improving their flow.

While teams can directly visualize the WIP levels over time (which I recommend), most people use the Cumulative Flow Diagram to visualize the WIP.

cumulative flow diagram

Cycle Time

The amount of elapsed time between when a work item "starts" and when a work item "finishes."

This metric is a lagging indicator of flow. It is available only after an item is actually finished from the workflow perspective (e.g. reached a Done lane on the Kanban board). It is typically used to drive improvement work as well as to be able to establish internal/external expectations as to the team's turnaround time on specific items. The main chart/report used to visualize and analyze Cycle Times is the Cycle Time Scatterplot where teams can understand their Cycle Time trends, distributions, look at anomalies.

cycle time scatter plot

Throughput

The number of work items "finished" per unit of time.

Note the measurement of throughput is the exact count of work items, without any compensation for item size - which is a major difference between throughput and story-points based velocity. Throughput is measured at a certain step in the workflow, typically at the finish line of the workflow. Throughput can be visualized via a separate run chart or by looking at the angle of curves on a Cumulative Flow Diagram.

throughput run chart

Work Item Age

For currently active items - The amount of elapsed time between when a work item "started" and the current time.

WIP and Cycle Time are classic metrics every Kanban practitioner is probably familiar with and throughput is somewhat similar to Velocity.

Work Item Age is the new guy on the block. Work Item Age complements Cycle Time. If Cycle Time is a lagging indicator only relevant for finished items, Work Item Age is a leading indicator only relevant for non-finished items. The basic idea is to provide transparency to which items are flowing well and which are sort of "stuck" even if not formally blocked.

I've been using some variant of this metric with most Kanban teams I've worked with. I also worked with several Kanban tool vendors to introduce some way to visualize card/item age.

Age on its own is interesting but not enough. We also want some indication of flow health. One common thing to visualize is the age in the current step in the workflow also known as "cards that didn't move recently".

Another way to look at it would be to look at the overall age but combine it with where the work currently is in the workflow as well as what the team expects their cycle time to be (We call that expectation Service Level Expectation (SLE) in the Kanban guide for Scrum teams and the PSK class). Combining all this information can help the team focus on the items that are at the most risk of missing the team's expectations/SLE. For example, let's say a team has an SLE of 16 days with 85% confidence. If one of the cards on their board has an age of 10 days, is that ok? is it a problem? The answer is that it depends. If that card is very close to the end of the workflow it is probably not a problem. If it is very close to the start of the workflow it is probably an indication of a problem that requires attention. The "Aging Work in Progress" chart below provides this perspective of both where active items are in the workflow, what the typical cycle times for this team are, and based on that which items are indications of flow risks (obviously orange-red means very low probability of finishing within the team's flow expectations).

To sum up - Work Item Age is the best metric to look at if you want to determine when an item that has already started is going to finish. This is in contrast to an item that hasn't started - where your best bet is your historical Cycle Times. The Service Level Expectation is just an expectation set by the team to themselves answering the question "What Cycle Time do we expect to see for an item of this type, and what is our confidence level for this?".

aging work in progress

Note: The charts above were created using the demo version of ActionableAgile Analytics - a tool created by my co-steward of the Professional Scrum with Kanban class - Daniel Vacanti. You can access the demo yourself and play with these metrics and think about how they would help your Scrum team.

Using the Flow metrics in the Scrum events

 

So how can these flow metrics be used to improve the Scrum events? This is one of the key learning objectives in the Professional Scrum with Kanban class. In a follow up discussion with some of the Professional Scrum Trainers who attended last week's class, we came up with a matrix mapping the metrics to the events. (credit Maarten Kossen)

Matrix

I'll explain -

Sprint Planning mainly leverages Throughput in order to create a realistic forecast for the Sprint Backlog. Work Item Age might be relevant when you have some items left over from the previous Sprint and you want to decide what to do about them.

The focus of Daily Scrum is the ongoing flow within the Sprint so naturally what we care about is what's currently going on. Therefore, Current WIP and Work Item Age are the most important metrics in the Daily Scrum.

Sprint Review includes a review with stakeholders of both the Increment as well as overall flow behavior of the team - trends in Cycle Times and Throughput are interesting. Throughput can also be used as part of release planning/road-mapping discussions, especially when combined with Monte-Carlo simulations provide some better visibility/confidence into "What can be done by when". NOTE: It is always important to emphasize that these are projections/forecasts, not commitments.

Sprint Retrospective is all about inspecting and adapting the process and the workflow. Therefore it is the place to look at WIP, Cycle Times, Throughput from the perspective of looking for improvement areas.

To learn more about flow metrics and how they can be used in a Scrum context, check out my Professional Scrum with Kanban workshops or contact me directly

 


What did you think about this post?

Comments (18)


Erez Morabia
06:11 am May 13, 2018

Thanks Yuval for the great post. The images you have attached are a great help for understanding. When it comes to cycle-time, it seems you emphasis that this is an after-fact metric and not a prediction. The prediction metric in that aspect is the SLE (where cycle-time-like prediction is inside it). Is that correct?
Interesting point it the relation between cycle-time, SLE and sprint length. Sprint length should define the desired upper limit for a cycle-time. Therefore, we would like our average cycle time to be less then the sprint length. SLE defines probability and period (16 days with 85% confidence) - the period should be shorter then the sprint length. Any guidelines on how shorter it should be? for example, should we look on the 95% confidence period and make it doesn't exceed the sprint length?


yuvalyeret
09:50 pm May 13, 2018

Erez - thanks for the feedback. Yes cycle times are an after-the-fact metric that can be used to figure out a service level expectation (SLE) that is predictive ( I wouldn't necessarily call it metric btw ...)

I typically say that the average cycle time should be quite a bit shorter than the sprint length. We don't provide specific guidelines there but I would like to see cycle times and SLEs that have a length of 60-70% of the sprint length with as high a confidence as possible. e.g. for a 14 day sprint cycle time/SLE at the 9-10 day range with a confidence level that matches what the team can currently accomplish.


gazza8
10:33 am May 18, 2018

Thanks for this article, very useful to know other 'metrics' than the Burndown Chart.
However, regarding the 'Throughput', its description could be more expilcit: I do not understand how to measure it.


Curtis Slough
08:00 pm June 5, 2018

"Throughput: The number of work items "finished" per unit of time."
Say your team selects 13 items for a sprint and completes 10 of these selected items; the throughput is 10. Think about football and the different number of points for each way to score. Touchdown is 6 points, extra point kick is 1 point, safety is 2 points, 2-point conversion is 2 points, field goal is 3 points. The final score of your team for the game is 25. It may have taken your team 6 or 8 or 11 successful scoring attempts to makeup that 25 points within the game. Therefore the throughput for the game is the number of times your team scored in the game overall. Hope that helps.


Steven Spruce
02:03 pm August 10, 2018

Confusing analogy if you don't live in the USA :-)


Curtis Slough
04:40 pm August 10, 2018

Steven, you don't have to know American football to understand but I'll further simplify it and hope it helps. The basic point I was making is that you are awarded different points depending on the type of goal you score. Let's take and compare with Scrum. Within Scrum, you assign story points to each product backlog item. So for a sprint, the team plans for 15 stories worth of 45 points. At the end of the sprint, the team completes 12 stories with a total of 32 points. The Throughput for this sprint would be 12 because it's the number of stories/items that they completed. Since Kanban does not estimate items via story points, the Throughput is used to gauge how much work was completed in a given time.


Stefan
09:01 am August 31, 2018

Thanks for this article which complements well the reading of the book "Actionable Agile Metrics" from Daniel Vacanti.
One thing i'm wondering after having read most of the documentation around the Professional Scrum With Kanban is :
- What would be the most relevant (and compliant to Scum principles) kanban 1st stages in order to act the "starting" of the cycle time. Is it when Product Backlog Items are commited by the team for the sprint during the planning ? Is it when Product Backlog Items are qualified as "ready to plan" after some grroming sessions between the PO and the dev-team ? This part is not clear for me.
If you have some best practices or interesting references to share ....

Many Thanks


Abhilash Pandey
09:13 am September 25, 2018

So here are my 4 sprints:
Sprint Story Points Throughput
S1 50 10
S2 45 15
S3 40 10
S4 35 5

What does the team learn from these numbers?


Curtis Slough
03:11 pm September 25, 2018

Simple, they learned the number of stories that made up the story point totals. Throughput is not a planning metric so unless every story you do has the exact same story point estimation, it cannot be used for planning. It just answers the question of “how many stories made up the story point total”.


Yuval Yeret
07:58 pm November 7, 2018

I recently wrote about this as part of another follow up blog post https://www.agilesparks.com....

What I say there is that I would typically consider the item started once the team starts to refine the story in prep for the Sprint.


Steve Loren
04:02 pm August 7, 2019

Hi Yuval, This has been a great article and I found it at a time when we are just about to transition to Kanban with a large Scrum team.
In the article you share an image twice of the flow on the sample team with overlapping banners highlighting the Scrum Events and a few Flow Metrics. Is there any chance you could share that same image yet without the Event banners or Metrics? It would prove quite helpful for our purposes.


TheGadgetGuy
08:14 pm June 17, 2020

Does Kanban with Scrum have a time boxed Sprints?


Patrick
10:59 pm July 21, 2020

Yuval,

Thank you very much for the post and for all your fine work. It’s very helpful as I prepare for the PSK-I.

My question relates to the Aging Work in Progress Chart, and what determines the height of the red/orange/yellow/green color bars.

Please correct me if I’m mistaken: I surmise the rationale for the height of the color bars is that the earlier in the Workflow state, and the longer a Work Item takes to move to the next state, the greater the chances of it missing the SLE.

Therefore, I get that the “Analysis Active” column has the tallest red bar (and any Work Item in that bar would be in “danger” of missing the SLE). Plus, each succeeding red bar is shorter because it means Work Items are closer to meeting the SLE.

Would you please tell me, though:

1. What is the *calculation* that helps determine the height of the red/orange/yellow/green color bars?

2. Can this calculation be performed easily, or does it require some kind of statistical analysis or tool other than a calculator? LOL.

Thanks in advance!

Patrick


Kremena
05:14 pm November 20, 2021

Dear Yuval, thanks a lot for sharing the matrix mapping the metrics to the events, I find this very comfortable as this will help me to better organize what to show and focus in each event :) Kudos!


Kremena
05:20 pm November 20, 2021

May be in S4 somebody was on vacation?


Katrina Latyshava
09:44 am June 16, 2022

So far the most helpful article amongst those recommended for PSK I . It was a quick read, although learnt something new, something that wasn't mentioned in D.Vacanti's book "Actionable Agile Metrics for Predictability". THANK YOU!


Jimmy Mathew
06:27 pm August 7, 2023

When we discuss an item during sprint planning, won't historical cycle time helps directly or in-directly(influencing other factors that help)?
In my humble opinion, When we consider a work item for planning ( in any context, Kanban or real life) historical cycle times of similar items will help / influence our go/no-go decisions. So my opinion is - in an exam, if cycle time is among the options, and a slot ( select best n) is left after selecting the direct ones, cycle time is one choice. Also, there is an interesting correlation among average values of cycle time, throughput and WIP.


Chand
12:39 am August 23, 2023

Sometimes, stories are split to a finer detail level into multiple stories, e.g. one team would have a story of 5 story point, whereas some other team would have split it is 2 stories of 3 story points and 2 story points. if this is extrapolated to many stories, then the number of stories would, and therefore, Tthroughput be different for the same work done.

e.g. Team 1 had story pointed stories as: 5, 5, 5, 5 which is 4 stories with sum of 20 story points, whereas, Team 2 had stories as: 2,3, 2,3, 1,2,2, 2,3, which is 9 stories with sum of 20 story points.
Even though the work done is same, the through put would be 4 for 1st team and 9 for 2nd team.
And that is why Throughput in terms of number of stories does not give correct picture. Instead sum of story points, i.e. velocity makes more sense.