PDCA, A3 and Practical Problem Solving

Over the years, I have been party to at least three corporate-level efforts to bring “A3” or “Practical Problem Solving” into their toolbox. Sometimes it has other names, such as “Management by Fact” or such, but the approaches are all similar.

Typically these efforts, if they catch on at all, become exercises in filling out a form.

Actually, that shouldn’t be a surprise, because they are often taught that way – as process of filling in boxes in sequence, with a “module” for teaching each step.

Worse, it is often taught as an intellectual exercise, and once you are done with the three day class, you’ve been “taught.”

The various classes mention PDCA as being a crucial part of this process, but nobody really practices it.

People are sometimes taught that this process should be coached, but the “coaching” they get is typically organized as management reviews via PowerPoint.

The “problem solving team” shows their analysis and their “implementation plan” that is a list of tasks, and a timeline to get them done. The meetings become status reviews.

Sometimes the “coaches” offer suggestions and speculation about the problem, symptoms, or actions that might be taken. They rarely (if ever) get into the quality of the PDCA thinking.

This is one of the challenges we have in the west (and especially in the USA) where our culture is more one of “go it alone then get approval” rather than true teamwork with the boss. This often turns the “A3” into an exercise of getting approval for a proposal rather than a learning process.

Worse, it does nothing to teach the problem solvers to be better problem solvers.

Note that sometimes an A3 is used for a proposal, but the process of creating it is still coached, and part of the process is the consensus-building that happens before there is any meeting. But here in the west, we still seem to like to spring these things on a leadership team without a lot of that background work ahead of time.

Mike Rother and the “Toyota Kata” community have been discussing this gap lately, and working to close it.

The latest iteration is this SlideShare that Rother sent around today:

He clearly points out what people have been missing: The “A3” is really just another method to document the “improvement kata.”

The “Implementation” box, rather than representing an action item lists, is where the problem solver captures her PDCA cycles, what is being tried, what is being learned, as she drives toward the target condition.

The other boxes are capturing her understanding of the current condition, the target condition, and the impact of various problems and obstacles in the way of closing the gap.

One thing that makes this extraordinarily difficult: We are talking about more than the mechanics of problem solving here. We are talking about shifting the default, habitual structure of the interaction between people. That is culture, which is notoriously hard to change. Not impossible, but unless people are up front that they are actually trying to change at this level, there are a lot of obstacles in the way. This can’t be delegated.

Rapid PDCA

This sub-assembly line had a planned cycle time of 15 minutes. The most skilled and experienced assembler could almost get all of the work done in that time, but generally, two people were required to consistently deliver without stopping the main line.

Among other obstacles identified, the first assembly step was being done on a bench. This required the assembler to stabilize the main part with one hand, position the part being installed with another hand, and hold the tool with… you get the idea.

The idea was to design and build a jig that would hold the main part steady, in the right orientation, at a good working height.

Iteration 1: Mock up the concept.

Rodney (the lean manager) and Maegan (the line team leader) are showing their first iteration of concept.

But even before this, their real first experiment was to hold the part by hand and test where it needed to be stabilized. The cardboard mock-up was a confirmation.

01 Cardboard Concept

02 Cardboard Concept


Iteration 2:

They experimented with contact points and hold points and made a few adjustments…

03 Adjust Cardboard

Iteration 3: Have a more robust version made out of wood, try it with an actual part.

03 Wood #1

Learned: The part isn’t stable enough to work on.

Next experiment: Add a clamp, and mount it to a tool stand.

04 Wood on stand

Then Maegan tries it out.


What did we learn?

It is difficult to get the tools behind the clamp bar.


Iteration 4:

Add some wood shims to raise the part and see if that works better.


Expected result: Should be easier to get the tool in there. Now try it.


But we learned we need a better backstop for the clamp.

Iteration 5



Which seems to have done the trick.

This, plus a couple of other changes, got the cycle time under the 15 minutes, so one person could do this job.

All of this happened over less than a couple of hours. A far cry from the traditional tooling and jig design process.


This is the final version. As you can see, there is now a wedge under the clamp to change the angle + some additional clamps that allowed the tool to also be used for another subsequent operation.




Past Due Hours

This area was picked for the initial focus because they were way, way behind, and it was getting worse.

The initial work was done in mid-April. The target was consistent output at takt time.

As the team looked at the process, and identified the sources of disruption and variation, “changeovers” surfaced pretty quickly as a main issue.

“Which obstacle are you addressing now?” was the changeover times on the output process, so the target condition for the area’s team was to get the disruption to output for a change over down to a single takt time vs. the highly variable (up to 10 takt times or more) disruptions they were seeing.

One big mindset change was the concept of takt time. There was a lot of perceived variation in the run times of these parts. But upon study, the team realized the variation was a lot less than they thought. Yes, it is there, but over any given couple of hours, it all evens out most of the time.

As the team studied their changeovers, one of them had an insight that “We can do a lot of these things before we shut the machine down.” And in that moment, the team invented the SMED methodology of dividing “internal” and “external” changeover tasks.


Once that objective was grasped, they went to town, and saw lots of opportunities for getting most of the setup done while the last part was still running.

Their lot sizes were already quite small, the core issue here was the disruption of output caused by the increased tempo of changeovers. So that time was put back into capacity, resulting in the results you see above.

They continue to work on their changeover times, have steadily reduced the WIP between process stages, and (as you can see) keep outrunning their goals for “past due hours.”

But the really important bit here is that this was largely the team leaders, the supervisor, and the area manager. Yes, there was technical advice and some direction giving by the VP and the Continuous Improvement manager, but the heavy lifting was done by the people who do the work every day.

Consistent Output

This is from another company, in a completely different industry. Their issue, too, was that they were always behind. In this industry, the idea of a takt time is pretty alien. Even the idea of striving for a fixed level of output every day is pretty alien.

The team’s initial focus was maintenance time. They perceived that equipment reliability was causing them to fall behind in production, and so were shipping product to a sister plant every week to fill in the gaps.

The first question posed to them was “How much time is needed for production?”

In other words, they needed to figure out how much production time was “enough,” so they could then assess how much time maintenance could have. That would establish their target.

But in answering that question, they developed a takt time, then measured output cycles vs. that takt. What they saw was lots of inconsistency in the way the work was being done.

If they could hold to something close to their demonstrated lowest-repeatable cycle times, they could then know when they were “done” for a given shift, or day, and actually plan on maintenance time rather than seeing it as a disruption to production.

The focus shifted to the work cycles.

What is cool about this team is that the team leader / “learner” (in Toyota Kata terms) was the maintenance manager.

He gained a real shift in perspective. “Production” were no longer the people who wouldn’t let him maintain the machine, they were his customers, to whom maintenance needs to deliver a specified, targeted, level of availability as first priority.

The teamwork developed about the end of Day 2 of this intense learning week, and they have been going after “sources of variation” ever since.

Here is the result in terms of “daily output:”


As you can see, not only is the moving average increasing, but the range is tightening up as they continue to work on sources of variation.

The big downward spike on the right is two days of unplanned downtime. In retrospect, they learned two things from that.

  1. After foundering a bit, they applied the same PDCA discipline to their troubleshooting, and got to the issue pretty quickly. As a sub-bullet here, “What changed?” was a core question, and it turned out someone had known “what had changed” but hadn’t been consulted early on. Lesson learned – go to the actual place, talk to EVERYONE who is involved rather than relying on assumptions.
  2. Though they sent product out for processing, they realized they could have waited out the problem and caught up very quickly (with no customer impact) had they had more faith in their new process.

All pretty cool stuff.

These things are why this work is fun.

One-By-One; On Demand; As Requested

Once again we see another data point on the trend.

Why is “batching more efficient?”

Simple – you haven’t solved the problems yet.

Here is an ice cream shop that has no ice cream… until you order it.

Does this shop surface the next layer of issues to solve? Certainly. But by thinking it is possible and they trying it, she is learning fast.

As Requested; When Requested; Where Requested

Amazon.com’s competitive advantage over regular retail has typically been around good prices with the thing you are looking for being available. In essence, they are an online, extreme extension of the big box store.

The downside has been that if you want it now, you have to either pay extra for expedited shipping, or get it somewhere else.

Meanwhile, retailers have been attacking Amazon’s price advantage by working hard to put them in a position where they have to collect local sales taxes, leading to a number of court cases plus Amazon actually pulling their presence OUT of states to avoid having economic nexus.

Now Amazon has reversed this trend, and is starting to build very local distribution centers, promising, not next day delivery, but same day delivery – as in order it today, get it today.

The analysts out there are all saying this is the “death of retail” – assuming, I suppose, that although Amazon can change its strategy, brick-and-mortar retailers are, somehow, frozen in theirs. “Death of retail” really means “We analysts, while certainly not admitting we never saw this coming, can’t think of how we would respond.”

Ultimately someone will figure it out.

But I digress.

The point is this:

Short lead times AND high variety of offering AND good prices are what customers want, and that is what anyone in the business of fulfilling customer needs should be striving for.

The ultimate transaction is “I’d like one of these.” followed by “Here you go.”

Just because you can’t figure it out doesn’t mean others aren’t trying to.

Where does this ultimately lead?

This video speculates what would will happen when Amazon figures out how to deliver what you order the day before you order it. Enjoy. Smile


Simple Solution to Complex Problem

This video captures a crucial aspect of lean thinking – searching for the simple and elegant solution.

We can’t say this is an obvious idea, a lot of very smart people have been working on this problem for decades.

But it is a simple idea. Sometimes the search needs to be outside of the regular domain people are used to working in.

The web site is here:http://creativemachines.cornell.edu/jamming_gripper

The other aspect of lean thinking that is captured here is keep working to solve the problem. I run into so many technical people today who are so convinced that the problem is unsolvable with current technology that they either give up, or look to advance the state of the art in some (usually expensive) way.

One thing I am glad to see is a trend in engineering schools toward making things rather than just designing them.

Organize, Standardize, Stabilize, Optimize

As I work in the field, I continue to encounter a desire to get right to optimal results – “Why aren’t we working to eliminate all of the waste?”

Once people start seeing opportunities, there is a great temptation to try to simply engineer a new process and implement it. This is especially true in value stream mapping processes. The future state map seems like an engineering specification – just make it look like this, and all of these issues will go away.

The “kaizen bursts” are supposed to represent the obstacles in the way, but they often show up as “action items” instead as people are tempted to make this into a project.

But the messy reality in the actual workplace (gemba) requires a far more methodical approach.

The title of this post is from four steps that a consultant friend shared with me years ago. As I converge my own practice toward guiding people through implementing the daily improvement culture vs. implementing the tools, I am finding these words a handy mantra. This is how I have found myself explaining it lately:


Organize includes all steps required to make the current condition apparent. This can include 5S activity, but is by no means limited to that.

You want to know enough about what is going on that you can tell the normal, intended, pattern from one which has been disrupted by an obstacle or issue.

In other words, you know how the process should operate if people don’t encounter any serious issues.

This allows you to standardize.

This is not a standard in terms of something you are trying to enforce.

Rather, it is a standard that you are striving to achieve each and every time the process is run.

“Did we make takt time with this work cycle?”

“Did this changeover go the way we intended?”

“Did we make a good part this time?”

(for Brian): “Did we get the quote out in two days?”

Once you know how the process is supposed to work, and what results you are supposed to get, you can clearly identify those times when it doesn’t work that way, or the results are not what you wanted.

imageStabilize: Now you can begin marching up the PDCA ramp toward the standard as you work to stabilize the process by systematically eliminating the issues.

Of course, for this to actually work, you must have a process to flag problems right away, pounce on them, clear them, and then feed them into a follow-on process that works to solve them.

As you apply countermeasures to sources of instability, you are adjusting the standard itself to include those countermeasures, making it more robust.

So Stabilize and Standardize play off one another in successive PDCA cycles.

imageNow… here is the trick part. Optimize can be indistinguishable from “Stabilize.” It is more of a semantics issue than anything else, and honestly, it isn’t worth arguing about.


Because as a process becomes stable, you are going to keep driving toward perfection, True North. You are going to have intermediate-term goals you are trying to reach that keep you moving forward. You are optimizing as long as you are moving in the right direction. So there isn’t really a distinctive point of shifting gears, there is only the most important issue to address right now to hit the next target condition.

Scrap Bin Kaizen

If you are in a production operation, be it a factory or even administration, take a look in your scrap bins (or trash cans). What you find there can be very informative.

Is all scrap treated the same? In a lot of metal working operations, I see routine drop scrap mixed in with scrapped parts and products – things that were not intended to end up there.

If that is the case, what does it tell you about the attitude around scrapping a part? Is it routine, just part of getting the job done, like drop scrap?

If you find a scrapped part, can you find the defect or flaw that caused it to be unusable? How much work (and cost) was added after that defect was created? All of that is additional cost that added no value.

Was it painted? Partly assembled? Did it go through a bottleneck operation and consume capacity of the entire factory?

What about your trash cans or recycle bins in the office?

Do you see signs of frequent paper jams in printers and copiers? These events suck time away, and more than just the time to clear them. They cause a mental shifting of thought pattern away from the productive work that takes additional time to recover.

Here are a couple of thoughts.

Separate the routine drop scrap from bad parts that were supposed to be good parts. If you sacrifice parts to adjust in changeovers, separate those out into their own category.

Each of these is a different line of inquiry starting with the question “Why?”

Consider taking your bad parts from the last 24 hours and bringing them to a central location (or a location in the department). Conduct a “morning market” and get to the bottom of at least a couple of the root causes. You don’t have to solve all of them at once, just pick one and drive it to ground. Then another.

Scrap is not simply a material and time waste. It is a reflection of how well you really understand your processes.

Learning vs. Knowing (or not)

PC once again left a provocative post in the Lean Thinker’s Community, and gave us a link to this Tim Harford TED talk that drives home the point that learning and improvement is more about rapidly discovering things that don’t work than about designing things that do.

Trial and Error

Tom Wujec makes the same point in the Marshmallow Challenge. In that video, Tom talks about how 5 year old kids out perform most adult groups in a problem solving / learning game. While the adults engage in a single cycle of “know-build-fail” the kids engage in multiple cycles of “try-fail-learn-try again.” In the improvement world, we call this process PDCA.

Harford’s key point is that learning only happens through a process of trial of large numbers of ideas, followed by the selection and further trials on the best ones.

Hmmm… that sounds a lot like the 3P process of “Seven Ideas” as well as “Set Based Design.”