Waste Diagnostic · System 02 · the patterns worth knowing by name
The Wastes of Office Work
System 01 sorted every step into three piles. The third one, the work that exists to serve the process, arrives as a heap: a long list of things everybody agrees are annoying and nobody does anything about. Naming the patterns in it is what turns the heap into something you can act on.
The idea in one line: office waste comes in a small number of recurring shapes, and the shape is diagnostic. Once you can name what you are looking at, you know where the other instances of it are hiding, what evidence would size it, and whether it is worth a morning or a quarter.
Seen from outside, and from inside
Every piece of office waste can be described from outside, as something the customer experiences, or from inside, as something the work does to itself. Both descriptions are true and they lead different places.
A form that asks a customer for information the business already holds is, from outside, being asked the same question twice. From inside, it is one more field somebody types, checks, and later reconciles against the copy in the other system. The customer version of that finding gets fixed because someone senior hears the complaint. The internal version gets fixed because someone counted the hours. They are the same field.
Call them the outside view and the inside view, and run both over the same process. The outside view finds what is costing you customers, the inside view finds what is costing you capacity, and most businesses are noticeably better at one of them than the other. Whichever of the two lists below feels less familiar is usually the one with more in it.
What the customer experiences
All of these are visible to the person on the other side of the desk, and none of them visible on an internal dashboard.
Waiting on you
- Time the customer spends waiting for a reply, an approval, a document, or a delivery somebody promised. The clock runs on their side whether or not anybody on yours is working.
Being asked twice
- The same information supplied to a form, then to a person, then again at the next stage. Each repetition tells the customer the last one went nowhere.
Being sent around
- Four departments, four conversations, four separate explanations of the same request, because nowhere in the business is there one place to complete it.
Unclear communication
- Instructions, policies, or documents that different readers interpret differently. Every ambiguity you send out comes back as a question, a wrong submission, or a dispute.
The wrong thing available
- The slot, the capacity, the part, or the current version of the answer is missing when it is needed. Staff improvise from stale material and the customer receives yesterday’s position.
Opportunity lost
- The inquiry that went cold because the reply took three days and read like a template. Nothing broke, nothing was recorded, and the business never learns it happened.
Errors that reach them
- The wrong figure, the wrong entity, the wrong date. The correction costs more than the original work and it spends the customer’s confidence rather than yours.
The last one is worth separating from ordinary rework. Rework caught inside the business costs time. An error that reaches the customer costs time, plus the correction, plus a quantity of trust nobody writes down.
What happens inside the work
The Operations Diagnostic already looks for waiting, handoffs, rework, exceptions, bottlenecks, and single-person dependency. Those are the patterns that announce themselves, because something is visibly stuck. These are the ones that do not, because nothing is stuck and everybody is busy.
Producing more than anyone needs
- The weekly report circulated to forty people and read by three. Work made ahead of demand, in more detail than the decision requires, on a cadence nobody has revisited.
Doing more to it than required
- Three approvals where the risk justifies one. Formatting a document to a standard the recipient never looks at. The step is real; the amount of it is the waste.
Work piled up in queues
- Backlogs, unreviewed submissions, half-finished files, and drives full of near-identical versions. Every item in a queue is aging, and some of it will need redoing before it is touched.
The work touring the business
- One request traveling through email, a spreadsheet, a shared drive, and two departments before it reaches the person who decides. Each leg is an opportunity to lose or misread it.
Defects that survive the process
- Wrong data entered once and inherited everywhere. Duplicate records from two spellings of a name. The original mistake is cheap; the fifteen downstream corrections are not.
Expertise doing clerical work
- Your most experienced people assembling status updates, chasing signatures, and copying figures between systems. The work gets done well and by the wrong hands, and it is the most expensive waste on this page.
Waste specific to office work
These have no factory equivalent, which is why they rarely appear on anyone’s list. In knowledge work they are often the largest items on it.
Sorting and searching
- Hunting for the current version, the right file, the person who knows, or the decision that was made three months ago. Invisible because it is distributed across everybody in small pieces.
Measuring the wrong thing
- A number that is easy to count standing in for the outcome that matters, and behavior reshaping itself around the number. This one manufactures the other wastes on purpose.
Overload
- Demand, interruptions, and task-switching beyond what a person can absorb. Past that point, throughput falls and defects rise at the same time, which is why overload looks like laziness from a distance.
Interference
- Quick questions, drop-ins, escalations, and status pings that fragment the day. The interruption costs a minute and the recovery costs a great deal more.
The wrong frequency
- A report compiled daily for a decision made monthly. A review held quarterly for a risk that moves weekly. Too often and too seldom are both waste, and the first is more common.
The wrong order
- Work sequenced by who asked loudest rather than by value, risk, or what is blocking somebody else. The internal presentation gets polished while the customer issue ages.
Idle capability
- A specialist waiting on intermittent requests because intake is scattered and the work is never leveled. Capacity you are paying for and cannot use.
Ambiguity, the cause behind most of them
Uncertainty about the request, the owner, the standard, the decision right, or what finished means.
A request that says “please review” with no acceptance criteria, no owner, and no date generates a clarification loop, a wait while the clarification is answered, a meeting to settle it, a deliverable built against the wrong assumption, and the rework that follows. Five entries on your list. One origin.
That is why ambiguity gets counted separately. When several patterns trace back to the same unclear request or the same unowned decision, record the ambiguity as the finding and the rest as its consequences. Do it the other way and you will size the same problem five times, fix the five symptoms individually, and watch them regenerate.
The test: for each item on your list, ask what somebody would have had to know, decide, or own for this never to have happened. If the same answer keeps coming back, that answer is the finding.
Which ones are worth acting on
Running these lists over a real process produces more items than any business can act on, and all of them genuine. Sorting them matters more than lengthening the list.
Then hold the result to the family’s standard: anything that cannot be sized stays a question. It goes on a separate list, with the evidence you would need to settle it written beside it.
Get those numbers from your own work. Count a real two weeks: log what happened in the process, code each event into one of the patterns above, and add it up. It is unglamorous and it is the only version of the number that survives contact with a skeptical colleague. Tell the people doing the work what the log is for before it starts, in plain words: it measures how the work is arranged rather than how fast anyone works, most of what it will find is time nobody is spending, and no names go in the write-up. The complete system has the exact wording.
Published figures for how much office time goes on non-productive work are not a substitute for that. They measure different things, from narrow ones like time spent searching for information to very broad ones like all time spent coordinating, which includes plenty of coordination that needed doing. They overlap, so they cannot be added together, and several of the most-quoted come from companies selling the remedy. A business case resting on somebody else’s percentage is the easiest thing in the room to argue with. Your own two weeks are not.
Open ChatGPT, Claude, or whatever you use, and upload your two-week log. Then send these four messages one after another in the same chat. Paste in the pattern names from this page so it sorts with yours rather than inventing its own.
- “Here is a two-week log of what happened in this process. Sort each event into one of these patterns, and where an event fits two, name both and say whether each is an outside-view or inside-view reading.”
- “Group the sorted events by pattern and give me count, total time, and whose time. Use only what is in the log. Where the log doesn’t support a number, say so instead of estimating one.”
- “Which of these patterns look like consequences of a single unclear request or unowned decision? Say what I’d have to check to confirm that.”
- “Don’t recommend any fixes yet. I want the list sized before I decide what to do about it.”
You now have named patterns with numbers against them, which is enough to argue about properly and still short of a plan. Some of this work should stop entirely, some of it has to keep happening and can be made much cheaper, and telling those two apart is a different judgment from the one that found them. That is System 03, and it starts by refusing the question everybody asks first, which is what to automate.