Waste Walk (8 Wastes): How to Turn Eight Floor Observations Into a Ranked Improvement List

The Waste Walk tool does one job. It systematically identifies the eight types of lean waste observed in the process area. You go and look, you write down what you see, and you tag each observation against a waste category.

It belongs in the Measure phase of RDMAICS. By that point you have recognised a problem and defined its scope. You now need evidence about how the work actually runs, not how the procedure says it runs. A Waste Walk gives you that evidence in a form you can count.

Use it when you can physically or digitally observe the work. Use it when the team has opinions about where time goes but no record. Use it when you want a shared language for waste across a group that has never used one. Do not use it as a substitute for measurement of cycle time or defect rate. It sits alongside those measures and tells you where to point them.

Everything that follows is built on one fictional case. The data is sample data supplied with the tool, not a real organisation.

The area under observation is Accounts Payable, specifically invoice processing. The observer is James Patel, a Black Belt. The walk date is 2026-05-20. Eight observations were recorded, one against each waste category.

The case is useful because it is ordinary. Invoices arrive, people check them, approvers approve them, systems receive them. There is no dramatic failure. There is a slow accumulation of handling, waiting and re-keying that nobody owns. That is what a Waste Walk is designed to surface.

Before you record anything, fill the header. Area or process, observer, date. In the case that is Accounts Payable - Invoice Processing, James Patel (Black Belt), 2026-05-20.

These three fields do more work than they look like they do. The area field forces you to bound the walk. If you write a department name with no process attached, you will drift and your observations will not be comparable. The observer field matters because waste observation is subjective and the reader deserves to know whose eyes produced it. The date matters because a Waste Walk is a snapshot. A walk in a quiet week and a walk at month end will not look the same.

Walk the process in the direction the work flows. Start where the invoice arrives. Finish where it is paid. Resist the urge to fix anything while you are walking. Observation and intervention are separate activities and mixing them shortens the walk.

Each observation needs four things. What you saw, how often it happens, how hard it hits, and what you would do about it. The sample case records one observation per waste category.

Defects: invoice errors requiring manual rework, average 47 errors per week. Frequency frequent, impact high. Proposed action is a mandatory pre-submission validation checklist.

Overproduction: weekly reports generated for 12 stakeholders but only 3 read them. Frequency frequent, impact medium. Proposed action is a communications audit and a reduction to 3 targeted reports.

Waiting: invoices sit in the approval queue for an average of 3.2 days. Frequency constant, impact high. Proposed action is to delegate approvals under 500 dollars to the AP supervisor and escalate only above 5,000 dollars.

The tool produces a waste observation profile. It counts recorded observations by waste category and plots them on a scale from 0 to 4. In this case each of the eight categories carries a single observation, so the profile is flat.

That flat profile is itself a finding. It tells you the walk was balanced and that the observer did not anchor on one favourite waste. It also tells you that the count alone will not rank anything for you. The tool says this directly in its own note on the chart: use the frequency and impact fields to prioritise action.

So you read the profile for coverage and the fields for priority. Four observations carry high impact. Those are defects, waiting, inventory and motion. Three observations carry constant frequency. Those are waiting, motion and extra processing. The overlap is waiting and motion. Both are constant and both are high impact.

The tool also offers an AI review that explains what the profile shows. Treat that as a second reading, not as the decision.

A Waste Walk does not authorise spend. It narrows argument. With this case in front of you, a few decisions become easier to make.

First, sequencing. Waiting and motion are the two observations that are both constant and high impact. The approval delegation change and the single screen processing view are the obvious first candidates. One is a policy change. One is a system change. The policy change can usually start sooner.

Second, scope for the Analyse phase. The inventory observation gives you a number to chase, 460 unprocessed invoices and 12 days of work in progress at month end. That is a queue question and it points straight at a flow analysis. The defects observation gives you 47 errors per week, which points at a cause analysis of error types.

Third, stakeholder conversations. The overproduction observation, 12 stakeholders receiving reports and 3 reading them, is a cheap and visible win. The non-utilised talent observation gives you a reason to put frontline staff in the room for the rest of the project.

The most common mistake is writing conclusions instead of observations. Compare two versions of the same note. One says the approval process is inefficient. The other says invoices sit in the approval queue for an average of 3.2 days. Only the second can be argued with, tested or improved.

The second mistake is forcing one observation per category because the categories exist. The sample case has eight observations across eight categories, which is tidy, but a real walk may find three wastes in one category and none in another. Record what you see.

The third mistake is leaving frequency and impact blank. Without them you have a list, not a priority. The tool counts observations, and a count treats a constant high impact problem the same as a rare low impact one.

The fourth mistake is writing an action that is really a wish. Compare delegate approvals under 500 dollars to the AP supervisor against speed up approvals. The first names a threshold and an owner. The second names nothing.

The fifth mistake is walking alone and then presenting the findings as fact. Take someone who does the work. Their correction of your observation is often more useful than the observation.

It will not tell you the cause. The walk records that data is re-keyed 3 times and that operators move between 5 systems. It does not explain why those systems were never connected. Cause analysis is a separate piece of work in the Analyse phase.

It will not quantify benefit. There is no money in this output. The tool records 47 errors per week and 460 queued invoices, but it does not tell you what an error costs or what clearing the queue is worth. If a sponsor asks for a business case, you will have to build one elsewhere.

It will not validate the numbers you write down. The 3.2 day approval average and the 12 days of work in progress came from the observer. The tool accepts them as entered. If the observer estimated, the result inherits that estimate. Check anything you intend to quote in a decision.

It will not rank your actions. The profile counts observations and says so. Ranking comes from the frequency and impact fields and from your own judgement about effort and risk. API integration and a monthly improvement forum are not comparable pieces of work, and no chart will make them so.

It will not prove that the waste is still there next month. A walk is a snapshot taken on one date by one observer. Walk again. Compare profiles. That repetition, not the single result, is where the value sits.