Estimate project
← All posts

10 refreshing discoveries from going back to the Scrum and Kanban guides

10 things that struck me as refreshing or surprising while working through the Scrum learning materials.

13 min read
Project ManagementSoftware DevelopmentKanbanScrum

I learned Scrum 15 years ago, and ever since, I have regarded its set of rules and practices as more or less set in stone. Until recently, I had no reason to take another look at them.

Meanwhile, the projects I was involved in rarely followed proper Scrum. Instead, they tended to settle on a custom Agile process enriched with a few chosen Scrum practices.

The people I worked with, Developers and Project Managers alike, had all sorts of problems with Scrum. Some practices felt overly restrictive (Sprints lasting strictly a month, a fixed Sprint Scope), unclear (the distinction between high-level and low-level tasks), or of limited use (Story Points).

Not long ago, I was getting ready for the Professional Scrum™ with Kanban Certification. I treated it as a chance to brush up on my knowledge and set it against the current Scrum guides. To my delight, I found that Scrum had in fact evolved in response to these complaints in the meantime.

What follows is a subjective list of 10 things that struck me as refreshing or surprising while working through the learning materials:

  • When to ship
  • What the Sprint Review is for
  • Whether a live demo is required
  • Shorter Sprints are now official
  • The Sprint Scope can be negotiated
  • Quality cannot be negotiated
  • Story Points have become obsolete
  • Service Level Expectation
  • When tasks get too fine-grained
  • Value vs Work

Further on, I go through each of these points in detail. The takeaway is this – even a seasoned Agile practitioner benefits from going back to the basics every now and then.

Before that, though, let’s briefly go over a few fundamentals of Scrum and Kanban.

Scrum and Kanban

Scrum and Kanban are two widely used implementations of Agile practices. Although they can be seen as alternatives, they also combine well. Each tends to concentrate on a different side of the process. Both aim to make it more observable and adaptable, yet Scrum stresses the value of team collaboration, whereas Kanban concentrates on optimizing the flow. (A detailed comparison of Scrum vs Kanban can be found, for instance, here and here.)

In the past, I regarded Kanban as a lightweight and more flexible alternative to Scrum. In one project, we switched to Kanban because the client kept bringing in ever more urgent tasks in the middle of a Sprint. Keeping the Sprint Scope fixed was simply out of the question, and Kanban looked like the last option before sliding into complete chaos.

In 2020, the Scrum organization revised the official Scrum Guide. The revision brought some significant changes relative to what I had originally been taught. I was also pleased to learn that the Scrum organization had published official guidance on combining Scrum and Kanban into one coherent framework. This guidance is laid out in the Kanban Guide. It builds on Scrum in its entirety and adds Kanban on top to make the flow more efficient and predictable overall.

Note: There is another attempt at merging Scrum and Kanban, called Scrumban. Unlike the framework presented in the Kanban Guide, Scrumban is built around Kanban practices, supplemented with a handful of Scrum ceremonies. As such, it does not implement Scrum in full.

Scrum and Kanban complement each other nicely. I came to see them as a combination, and that is how I approach them in the sections below.

My reflections

When to ship

Scrum relies on the concept of an Increment. An Increment stands for a portion of work that can be built within one Sprint. Every Increment has to deliver value and be concrete, additive (with respect to all earlier Increments), and usable (see: Scrum Guide, “Increment”). Consequently, every Increment can, at least in principle, be released on its own.

So when, according to Scrum, is the right time to release an Increment? In particular, do you have to hold off on releasing until the Sprint Review is over?

No, you don’t. Releasing mid-Sprint is completely acceptable, even several times over. In the words of the Scrum Guide:

“A single Sprint can produce several Increments. All of them together are presented at the Sprint Review (…). An Increment can, however, be handed over to stakeholders before the Sprint ends.” (Scrum Guide, “Increment”)

The Kanban Guide says something similar:

“Many people wrongly believe that teams may deliver value only once in a Sprint. In reality, they are required to deliver value at least once in a Sprint.” (Kanban Guide, “The Sprint”)

This differs from earlier versions of the Scrum Guide, which suggested treating an entire Sprint as one Increment.

As a result, Scrum fits teams that release often or practice Continuous Delivery perfectly well. This ties in with the part the Sprint Review plays in the Scrum framework.

What the Sprint Review is for

The Scrum Guide describes the Sprint Review as a prescribed event at which the team shows key stakeholders what it has achieved.

But what if the stakeholders are unhappy with what they finally get to see at the Sprint Review? Wouldn’t it be better to hold the release until the stakeholders have had a chance to receive and sign off on the work?

Once more, the answer is no. If you aim for frequent releases, the work can be validated during the Sprint in other ways, such as code review and acceptance tests. The current Scrum Guide says so directly:

“The Sprint Review must never be treated as a barrier to releasing value.” (Scrum Guide, “Increment”)

The Sprint Review is all about communication. It lets the team and the stakeholders see where they stand relative to where they intended to be when the Sprint began:

“In the course of the event, the Scrum Team and the stakeholders look at what the Sprint achieved and at what has shifted in their environment. (…) The Sprint Review is a working session, so the Scrum Team should not reduce it to a mere presentation” (Scrum Guide, “Sprint Review”)

Whether a live demo is required

It is commonly assumed that a Sprint Review should include a live demo of the completed work. Even though this was never strictly mandated, it was frequently described as an integral element of Scrum practice (see: Agile Project Management with Scrum, chapter 5, “The Product Owner”).

As it turns out, the Scrum Guide says nothing explicit about a live demo being required. The team is therefore free to pick another way of presenting its results, one that works better for it. Here, too, the demo is not about getting a sign-off, but about enabling the discussion.

That said, a live demo brings significant value: it keeps the team focused on delivering a piece of work that actually works. In a demo, they cannot sidestep “the last mile” – all the annoying small details and dull side-tasks that have to be finished for something to really function. Preparing a demo scenario also prompts them to view their deliverables through the client’s eyes.

Shorter Sprints are now official

In the old days, Scrum mandated Sprints of exactly 30 days (see: Agile Project Management with Scrum, chapter 1, “Scrum Flow”). Combined with a frozen Sprint Scope, this made Scrum look rather inflexible, and some teams found it too inflexible for their circumstances.

Before long, some of them began working with shorter time spans. Common as it was, such an adjustment deviated from the official rules. Today, the Scrum Guide formally acknowledges shorter Sprints as a legitimate option. (In fact, they had already appeared in the very first version of the Scrum Guide, published in 2010.)

Monthly Sprints did have their advantages, though. Devoting two days to Sprint events at the start and end of a month-long Sprint can be reasonable. As Sprints get shorter, however, these events take up a proportionally larger share of the time. That is why the Scrum Guide now recommends scaling these events to the length of the planned Sprint.

The Sprint Scope can be negotiated

What surprised me even more is that the Sprint Scope can in fact be renegotiated partway through the Sprint. The Scrum Guide puts it this way:

“As the team learns more, the scope can be clarified and negotiated again with the Product Owner.” (Scrum Guide, “The Sprint”)

You can bring in a higher-priority task or drop part of the Scope, provided that the change does not compromise the overall Sprint Goal.

This illustrates well how Scrum has grown more flexible. All through the Scrum Guide, the authors rely on empiricism as the guiding principle. Empiricism rests on experimentation: some aspects of the work in front of you can only be discovered by experimenting with it. Once you accept this, you have to stay open to the idea that what you discover may change the plan.

With the Product Owner’s agreement, a Sprint may even be cancelled.

“Should the Sprint Goal become obsolete, the Sprint may be cancelled.” (Scrum Guide, “The Sprint”)

Quality cannot be negotiated

What remains non-negotiable is Quality – as set out in the existing Definition of Done. This has in fact never changed. Bear in mind that the Definition of Done itself is something the Scrum Team (the Developers, the Product Owner and the Scrum Master) agrees on, and it can be revised as requirements evolve.

Still, a task that does not satisfy the agreed Definition of Done can never be regarded as finished.

“A Product Backlog item that falls short of the Definition of Done may be neither released nor even shown at the Sprint Review. It goes back to the Product Backlog to be considered later.” (Scrum Guide, “Commitment: Definition of Done”)

Having unfinished items at the end of the Sprint does not have to mean throwing the work away blindly and starting over. Instead, before moving on, you should pause briefly, look around, and check that the priorities are still the same.

Story Points have become obsolete

When I was first taught Scrum, I learned to estimate with Story Points. Strictly speaking, Story Points have never been one of the official Scrum practices, but in those days learning materials were full of them.

Since then, I have encountered a good deal of criticism of Story Points. Andy Hiles gives a thorough list of arguments in his book (see: Applying Scrum with Kanban, chapter 1, “Story-Points”):

  • “The first issue is how much time teams spend estimating the work (…) That time would have been much better used actually doing the work and discovering its complexity.”
  • “The second issue is that teams stop concentrating on producing outcomes; as time goes on, Scrum Teams discuss the Points instead of (…) the outcomes they are creating.”
  • Moreover, by aiming for precise estimates, Story Points create a false sense of certainty. Every estimate carries some uncertainty, yet Story Points offer no means of conveying that risk to the client.
  • Story Points can even harm the work culture: when managers expect every task to be estimated in advance, the team may grow too afraid to admit that for a particular problem they just don’t know yet and have to start working on it before they can come back with an estimate.
  • And despite all that effort, estimates expressed in Story Points are not especially reliable. In Andy Hiles’s words, they amount to guessing at guessing.

Service Level Expectation

This time, I came across Service Level Expectation as a different way of approaching estimation. Unlike Story Points, SLE relies on historical data (facts), namely historical Cycle Time:

“Cycle Time: the time that passes from the moment a work item is started to the moment it is finished.” (Kanban Guide, “The Basic Metrics of Flow”)

For instance, if you know that 80% of all the work items you have finished so far took 5 days or less, you can turn this into a forecast: any future task should likewise take 5 days or less, with 80% probability.

As you can see, SLE inherently carries information about probability, which gives you a handy way to communicate risk to the client. If you deal with tasks of different categories, the method generalizes easily: simply calculate SLE for each category on its own.

Naturally, the method is sensitive to variance in Cycle Times. It works best when the variance is low, i.e. when most items take about the same amount of time to finish. Interestingly, Kanban does not attempt to reduce the variance of Cycle Times. Rather, it concentrates on optimizing the Cycle Times themselves, i.e. on getting each item done as fast as possible.

The importance of optimizing Cycle Times is backed up independently by the “State of DevOps” report (see: Modern Software Engineering, chapter 3, “The Importance of Measurement”). Cycle Time is very similar to Lead Time – one of the four measures that, according to the report, correlate statistically with team performance.

When tasks get too fine-grained

In the past, I was in fact fairly satisfied with Story Points, particularly compared with estimating in man-days. I noticed that estimates improve once you break an item down into smaller tasks (the overall estimate typically rises, at times considerably). With Story Points, you may be inclined to split items as finely as you can during Sprint Planning.

This can work well for repetitive work in familiar areas, but it is exposed to unforeseen problems and surprises (the “unknown unknowns” vs “known unknowns”). One could argue that trying to settle every question during the Planning session produces tasks that are too fine-grained and leaves no room for experimentation during development.

Curiously, Kanban metrics and Service Level Expectation can also result in work items that are too fine-grained. This would happen if optimizing the Cycle Time metric were not balanced by another rule – namely, that every work item has to represent value rather than work.

Value vs Work

Both the Scrum Guide and the Kanban Guide draw a line between value and work, and both put the emphasis on value:

“Within a Sprint, the Scrum Team transforms a chosen portion of the work into an Increment of value.” (Scrum Guide, “Scrum Definition”)
“Kanban: an approach to optimizing how value flows through a process” (Kanban Guide, “Definition of Kanban”)

Value is what gets discussed with the client and during Sprint Planning, and it is what the Product Backlog and the Sprint Backlog consist of. Consequently, every Backlog item has to deliver concrete value to the client on its own. It will, for instance, stand for an entire User Story rather than a technical task that forms part of that Story. A Backlog item also does not need to be described in great detail; in particular, you do not need to settle every implementation detail in advance. This leaves space for experimentation and decision-making during the Sprint.

What does this look like in practice? You might keep two levels of items on your Board:

  • high-level items that represent value,
  • and the associated low-level items that represent pieces of work.

From a Scrum point of view:

  • the Sprint Scope would consist of the high-level items,
  • adding or dropping a high-level item in the middle of a Sprint would amount to changing the Scope, which calls for negotiation within the Team,
  • whereas adding low-level items costs nothing, as it does not affect the Scope.

From a Kanban point of view:

  • Kanban requires limits on work items in progress; you may well choose to apply those limits to the number of open high-level items,
  • and tracking the Cycle Time of high-level items matters most, because it reflects the time needed to deliver real value.

(A more in-depth discussion of Value vs Work can be found here.)

Summary

These, then, are the reflections I jotted down while refreshing my understanding of Scrum and Kanban in preparation for the exam. They certainly do not cover the subject exhaustively. Overall, I am happy I took the exam, as it gave me an excellent chance to refresh what I knew and pick up something new.

Below are the sources I relied on in this article:

And if you are thinking about taking a Scrum exam, I can recommend the Professional Scrum with Kanban Certification as an alternative to the standard Professional Scrum certificates. The Scrum site offers a thorough list of learning materials for the exam.