Sunday, August 09, 2020

Bringing Agile to non-IT work... those who don't remember history are doomed to repeat it

People tend forget that many of the agile approaches borrowed heavily from the Toyota Production System and its relatives, commonly known under the umbrella term "Lean."

These days we're experiencing an interesting development. People try to bring things they perceive as "typical agile practices" to non-IT work. For knowledge workers --a term coined by Peter Drucker in the 1960s-- this might perhaps be a valid approach. Even though I doubt it. For non-knowledge-worker work on the other hand, I would like to point out what happened here. Approaches taken from the Lean context of shop-floor management and vehicle design, related to continuous improvement and optimizing the flow of work, were translated into a very different environment -- that of software development. And even the considerable body of knowledge from other fields of expertise that is at the foundation of agile was put into a very specific context here. Actually, the so-called "agile manifesto" is called "Manifesto for Agile Software Development" and thus very specific with regards to the domain it targeted. So nowadays, when people try to "apply Agile to non-IT situations", they basically take the adaptations that have been made to non-IT approaches to make them helpful in Software Development and try to re-apply what's left of them back to non-IT work. Of course the original non-IT approaches have also evolved since the days that --just to pick an example-- Ken Schwaber and Jeff Sutherland read 1986 paper "The new new product development game" (sic!), and took parts of those ideas as a foundation of their own agile approach (Scrum). Hence it seems kind of silly to me to derive ideas for modern ways to organize non-IT work from the spin-offs from more than two decades ago instead of going directly to the source.

Of course sometimes re-applying the stuff we learned from agile software development actually works, but I think going directly to the source is a much better idea. Perhaps instead of trying to derive the helpful approaches non-knowledge-worker work from the shadows they cast unto the walls of the Agile world --to paraphrase Plato-- it might be a good idea to look at the origins and try to understand the original approaches to non-knowledge-worker work. Of course, oftentimes non-knowledge-worker work was simply called "work", back in the day. Directly adopting from approaches like Lean (from the 1950s) or New Work (which originated in the 1970s) might be an approach to improving work that avoids the 'Chinese-whispers' effect of the indirect approach via "Agile."

To end on a more positive note: the Kanban method is a great example of an approach that targets the challenges of the (non-it) knowledge worker and brings ideas from Lean and similar fields into a new context. And even though many people use the Kanban method in the realm of IT, it has many equally --if not more-- effective applications outside of IT. Maybe that's because the Kanban method avoids the triangulation via the older agile approaches and builds directly upon the common ancestors. I guess that is one of the reasons why David Anderson called the Kanban method "post-agile" even back in 2010.

till next time

  Michael Mahlberg

Sunday, May 24, 2020

Error: Customer unavailable! How do you manage flow now?

It’s Wednesday morning, you try to get an appointment with the stakeholder who is the most important customer for a work item that your team started to work on on Monday. And now you find out that they’re out of the office for the coming two weeks.

Now what?

You and your team invested 2 days already (for a 7 people team that would represent 14 person days of effort), and you all would really like to work on the topic.

After some inquiry about the requested feature it becomes clear that no one else knows what the requirements on this work item really are and now you’re faced with a tough decision: mark this card as blocked and become ‘idle’ as a team, or abandon the card (discard it) and let the customer bring up the topic later again?

There are drawbacks to both approaches –after all the situation is bad to begin with– but both follow the mantra to “accept reality” and are way better than following the wiscy approach where you start defining the result (e.g. by programming) without even knowing the requirements.

Let’s look at the implications of each approach.

For the following assume that we manage our work using ideas from the Kanban method and thus have the flow of our work visualised on a board including WIP-targets.

Option 1: put a blocker on the card if the customer is unavailable

Putting a blocker on a card is the normal way to make it visible to everyone that there seems to be no way to progress on the work represented by that card in a sensible way. At least if you apply the Kanban method. If your WIP-Target (mostly known as WIP-Limit) was already honored this creates a tremendous amount of slack for the team that worked on that card. In most cases more than you actually want and more than is wise from an economic point of view. But it reflects reality. If the relevant stakeholder is not available for an extended period of time this could be really expensive.

The important point here is that blocking the card fosters learning inside the system. People might come up with policies to make sure that the customer is actually available for before starting to work on the work item. Or they might come up with completely different ideas on how to mitigate this situation. Remember: Documentation isn’t bad per se.

Option 2: discard –or abandon– the work

On the other hand there is always the option of starting something new. In a system with a known capacity that usually means abandoning the old work – after all there are good reasons to limit the work in progress in a system. Amongst other reasons, allowing more work in has a very clear and negative impact on the lead time for the system as a whole. Discarding the work doesn’t necessarily mean that all the work already put into this item will be lost. It just means that the work item has to start its journey again from the far left side of the board and it might even take a couple of replenishment meetings (where stakeholders decide wich item to commit into the system) before work is started again. Some of the work will probably have to be redone because the world will have changed by the time work on that item again.

Here the systemic aspect is that discarding or abandoning a card fosters learning outside the system. Especially upstream, at the customer’s end. Customers might come up with the idea that it is a good idea to have more than one person who knows what the reasoning behind a requirement is. Or they might develop other smart ideas on how to handle this. Remember: there once was a whole body of knowledge called requirements engineering.

Either way: it is a chance for learning

Both approaches offer a chance to improve the system. While the former is targeted more at the system itself, the latter is target more towards the customer interface. Both are legitimate ways to “accept reality” and thus are way better than the alternative: wishful thinking and hoping you will guess the customer’s needs correctly.

till next time
  Michael Mahlberg

Sunday, April 26, 2020

Handling recurring work - some options from Lean and TPS

As someone who draws from the Kanban Method and its Lean and TPS ancestors when supporting change in organizations I regularly come across the question of “how to handle recurring work?”

And while neither the Kanban Method nor I have definite answers there are some options from Lean and TPS that can be integrated nicely into the workflow of 21st century knowledge workers.

I originally got the basic idea to adapt Kamishibai boards (see below) from @LukasDonSchmidt and I would have preferred to just link to a blogpost of his, but unfortunately I can’t find that out in the wild. Yet. (Didn’t you intend to write something, Lukas?)

Capacity allocation and recurring work

Assuming that you visualized your work using some kind of board and you use WIP-targets to manage flow, there is one challenge (amongst many others) that often arises. Small, recurring units of work that “need to” interrupt the flow of the big main work items. In an ideal world we might just strive for the knowledge worker’s equivalent to single-piece flow, but in the real world that is not always attainable or even feasible. To take a small business, non-IT example: the big Work Item might be an essay that needs a week to finish, and the small items could be fetching the mail and paying invoices.
One approach to handle this is to combine the concepts of different swim lanes for different work item types and capacity allocation. Let’s look at a small scale example for that.

Starting with this fictional, initial board for a very small –let’s say three person– writing agency, we see that the company has committed themselves to limit their sold “Options” ((Options for others, but things the people from our little agency committed to) to six “announced” articles and agreed to have a target number of articles they’re “writing” concurrently to two (thus allowing some work to be already in the “ready for B” column). Furthermore they agreed to have only one article at a time in the “copy edit” stage and one in the “delivery” stage. The “done” column in the example is unlimited (∞) and the invoicing process seems not to be part of this board.

But is this kind of work really the only thing that has to be done in this small company? Of course not.

Let’s take invoicing for example. Even if we assume that writing the invoice and sending it out is part of “delivery” then there still is the little thing of checking if all open invoice get paid some time. And most probably there are also invoices coming in that need to be paid. So there a many small things that need to be done and that we might want to visualize if they are important enough to justify the effort.

Taking the approach of capacity allocation a first solution might look like this:

The idea here is to simply start by allocating part of the day to the big chunks and reserving a smaller part for all the small chunks. Those small chunks might be things like fetching the mail, checking the payment of invoices or paying their own bills.

Following the original ideas behind the Kanban Method principle of “visualize work”, this whole setup is rather buggy. The description of the columns doesn't match with the actual work that's performed in that column, and probably the whole workflow is quite different for those smaller pieces of work.

So let's take another page from the Kanban method and let's have a look at what we actually do now (start with what you do now ) for those small items that we work on we find that we don't have a general workflow but instead the workflow for each card follows the steps of “pick a card”, “understand card”, “perform task”, and “reschedule card as necessary.” And even though this might be the actual workflow, for such small tasks it would not make much sense to separate “understand card” and “perform task” – especially since they happen repeatedly.

So, a more realistic model might only include columns like “options”, “doing”, and “ready to reschedule.”

Picture of a two swimlane board with different workflows for Standard and repeating work

In this picture we now have a Kanban board with two swim lanes. Each lane with its own work item type and both with separate work in progress targets.

Even though this board would actually work (and in many cases I've seen boards like this being really helpful) the actual workflow here is so generic that it doesn't really pull its weight to have a separate board for this.

Depending on details of the cadences (i.e. the rhythms in which work happens) and how much time each item takes to complete, the aforementioned challenge might also be addressed via the lean concept of Heijunka (workload levelling) in the way I mentioned in my earlier post.

Enter kamishibai…

Luckily, the Kanban board is not the only thing that made its way over from TPS into other applications of lean thinking. One of the important parts of the Kanban Method –and probably the most visible one– is the first principle: “Visualize (the work and the workflow).” Using visual representations for things important for the job to be done is a key tenet of visual management. Which in itself is a key component of lean production. Like other things from lean (e.g. 5S), visual management has to be adapted to knowledge work to reap its benefits in this realm as well. Kamishibai boards, which are sometimes used for recurring audits, can be used quite nicely to manage recurring tasks with low complexity like the ones described in the example above.

A simple kamishibai-board just consists of a place to put cards with different colored sides (e.g. red/green). Having a card with the green side up means it’s done, having it with the red side up means it is not yet done. Also, you have an indication of the cadences (the Rhythms in which you want to repeat the tasks) and the very moments those cadences start. Each time a cadence start you simply turn all the cards in that area to show the red side.

Now each time you intend to work on the recurring tasks, you just go to the board, pick a red one, do it and place it back again with the green side up.

Picture of a very simple Kamishibai-Board

In this example the mail has been fetched (green card), the receivables have been checked (green card) and the invoices have yet to be paid. All daily tasks will be turned to red at 9 o’clock and all weekly cards will be turned to red on Sundays at 9pm.

Do you have to do actual board design for this?

Certainly, even board with almost no explicit design like the one pictured above will already be sufficient for many situations. That doesn't mean that there are no approaches to designing such a board that can be helpful. First off let's start with the areas. In the example above we concentrated on weekly and daily recurring tasks – that might be completely different in your situation. Whatever the cadence of your recurring tasks is: make an area for that. For example, daily weekly and monthly or biweekly or any other cadence you actually need. A helpful practice here is to note down the point in time when cards should be turned back to “undone” next to the cadence name, so that everyone knows when the cards need to be turned back and the organization doesn’t become dependent on the knowledge of one single person who acts as a “keeper of the board.”

If you have more than just a few recurring tasks, it becomes a good idea to hang the board on the wall instead of wasting precious horizontal flat surface otherwise known as desk space. something metallic with magnets to hold the cards is a good idea here. Or you might to want to invest in professional T card holders as they are used for example in manufacturing (or just build them yourselves…).

Picture of a DIY T-Card-Holder

What about the cards?

Even with a simple board there are a lot of things you can do with the cards. By the way: a simple way to create cards with two sides of different color is to just stick two post-its together, back to back… Or to take a white card and mark the top off in the colors of your choosing. Lots of room for DIY ideas here.


Is it really worthwhile to talk about the colors? Can’t we choose whatever looks fine? E.g. blue for undone and yellow for done? Or whatever our mood suggests? Yes that would be an option and it would even work. But with visual workplace management we want to make things easy to recognize with as little mental load as possible. So I suggest to stick with color codes that fit with everyday expectation, and Green for “done” seems to be pretty universal while “to be done” works with any color that’s far enough away from green like Orange, Red, Purple and the like.

What to put on the card

Apart from a short description of the task, it turned out to be a good idea to also put some additional information on the card. Two of the most important improvements or adaptations for knowledge work are sizing and tracking.

Assumption is the mother of all screw-ups

One helpful piece of information to add to the card is the time it should take to complete the task. Also, to make it more plausible that the work gets done reliably, it is a good idea to note down how long it actually took. A good practice here is to write the expected time on the “to be done side” and the time it actually took on the “done” side. And make sure to adjust the time it should take in accordance with those numbers. At least each time you have to replace the card because there is no more space for more records. Then we’re not talking about estimations anymore but about actual data from measurements.

Card sizing (kind of a hack actually)

If you have a dedicated amount of time to take care of your recurring tasks, it is a good idea to have a rough idea how many of those you're going to handle in any one session. One way to visualize this is to just have different sizes for the cards reflecting their average effort (doesn’t work with the T-Card approach, btw). This way, you can easily see how much of your time slot for recurring work will be filled up by the cards just by having a space at your own work area reflecting the time you allocate for these recurring tasks by it’s size.
You still should try to have a healthy mix of cards from different cadences e.g. daily and weekly, so that you don't have to do all the weekly stuff at the end of the week (or whatever the largest cadence is – could well be quarter in many enterprises).

Kamishibai – a little more complex

Even though simple recurring tasks do not merit the creation of a board that reflects their workflow, some recurring tasks can become quite complex. For those we can combine the whole idea of kamishibai with some insights from the checklist manifesto. The aforementioned task of checking invoices or receivables may actually turn out to be a bit more complicated than it might seem at first glance. Especially when the business is prospering and the client base grows. It might look like fetch the list of open receivables, add invoices we wrote since the last time we checked our receivables, fetch bank statements, cross out invoices where we got the money from our list of receivables, update last-check-date on the list of open receivables. As we can see, this workflow is in no way a process of knowledge gathering but actually just a collection of steps to be performed. Therefore, it wouldn't be natural inside the Kanban method to have a column for each of the steps. Furthermore, this process is very specific to this one recurring task. So, what's more natural than to put it directly on the kamishibai card? This way you might end up with kamishibai cards with things like cadences written on them, with an “Effort”, with little checklists and with any additional information helpful in the specific context of that task. Quite enough to handle most simple recurring tasks. Should the tasks become so complex that they warrant their own workflow the quest for an appropriate solution starts again of course.

The right approach?

Is it more appropriate to takle recurring tasks based on the kamishibai ideas? Well - it depends. It depends on the maturity of the people using the boards, it depends on the type of the recurring tasks –the approach lined out here assumes that the recurring tasks are small–, it depends on the complexity of the recurring tasks –having checklists on the task-cards makes it easier for people to swarm on recurring tasks- etc.
In my personal opinion one of the biggest advantages is that the approach helps in separating “value items” or “products” from items that are mere “tasks“ and are neither visible to the customer nor appreciated in themselves.

This illustration shows the approaches “use Kanban boards for everything” (one way) and “Kanban boards for value stream work and kamishibai boards for recurring tasks” (another way) can be used to represent the exact same situation.

A multi-process Kanban board broken down into value-stream Kanban board and kamishibai board

Both the light blue and the yellow scenarios represent the same situation. With regard to the recurring tasks: Two cards to do, one card in progress and one card “done” (at least done until the next cycle starts).

The second scenario uses a kamishibai representation for the recurring tasks. The recurring cards that were in the “committed” state on the kanban board of scenario “one way” became “undone” (red) cards on the kamishibai board in the scenario “another way.” An additional advantage is that the different cadence for both cards –one occurs weekly, the other one each day– is immediately visible. The “in progress“ card from the Kanban representation is not visible in this instance of a kamishibai board because the person working on it has taken it along e.g. to use the checklist on it. The card from the “done” column became a green (“done“) card in the kamishibai version. Again with the advantage that the cadence of that card is obvious.

Try kamishibai

Before you try to handle all work alike, maybe give the approach of kamishibai for recurring work a chance – it might help more than you think.

till next time
   Michael Mahlberg