The lag
Counting from the right day
Lag and elapsed week are different numbers and almost every statement about a transition needs one or the other.
The standing slot for work with no ownerThe fastest of the five problems
week 5 · surfaced after 14 weeks
Reading supplier contracts before renewal
week 2 · surfaced after 22 weeks
The quarterly reconciliation of the project ledger
week 3 · surfaced after 31 weeks
Shadowing new starters through their first week
week 6 · surfaced after 47 weeks
Logging why a job was turned downThe slowest on the record
week 9 · surfaced after 64 weeks
Surfaced 5
Five removals were written down with the week each of them stopped. Every one of them produced an event inside the 78 weeks watched. Of these, five surfaced as a problem. All five removals whose absence surfaced as a problem, in the order they surfaced.
Two clocks run in any record of this kind, and they give different answers to almost every question. Keeping them apart is a small discipline that most accounts of shorter weeks do not observe, and the errors it produces are large.
The two numbers
Lag. Weeks between a thing stopping and anybody learning anything about it. A property of the removal.
Elapsed week. Weeks from the start of the transition to the same event. A property of the calendar.
They differ by the week the removal was made, which on this record ranged from zero to eleven.
| removal | week stopped | lag | elapsed week of the event |
|---|---|---|---|
| the standing slot | 5 | 14 | 19 |
| contracts | 2 | 22 | 24 |
| the reconciliation | 3 | 31 | 34 |
| shadowing | 6 | 47 | 53 |
| the refusal log | 9 | 64 | 73 |
Which one each question needs
Lag answers questions about removals. How long does it take to find out? Which kinds report fast? Does a contested removal report sooner than a quiet one? Every finding on this site about the behaviour of removals is in lag, because the number has to be comparable across entries that stopped in different weeks.
Elapsed week answers questions about the organisation. What had happened by the time we reported? Would a six-month pilot have seen this? When should the review be? Anything involving a date in a calendar needs this one.
The confusion arises because the two produce the same headline number for a removal made in week zero, and three of the thirty-eight were.
The error this prevents
The specific mistake, and it is easy to make: the five problems on this record have a median lag of thirty-one weeks. It does not follow that a thirty-one week review would have caught half of them.
Their elapsed weeks are 19, 24, 34, 53 and 73, with a median of thirty-four. A review at week thirty-one catches two of the five. A review at week thirty-four catches three, because the reconciliation lands exactly there.
The difference between "median lag thirty-one" and "median elapsed week thirty-four" looks like rounding and is three weeks of removals. On a larger transition, where the removals are spread over six months rather than eleven weeks, the gap between the two medians would be months.
A median lag describes the removals. A review window has to be set against elapsed weeks, because that is the clock the calendar runs on.
Why the distinction disappears from published accounts
Because most of them report neither.
The standard form is a duration and an outcome: "after six months, output was maintained". That sentence contains an elapsed period and no information about removals at all, so the question of which clock it is on does not arise — and neither does the question of what the six months covered.
A record that keeps both numbers can say something the standard form cannot: at week twenty-six, nineteen entries had produced no event, and the lags of the events that had occurred ran from one to twenty-four weeks. Both halves of that are needed. The first is in elapsed weeks, the second is in lags.
The practical rule
Record the elapsed week in the list, as a plain number, because it is observable and cannot be reconstructed later. Compute the lag when it is needed, which is any time entries are being compared with each other.
The reverse — recording lags and trying to recover elapsed weeks — fails as soon as anybody loses track of which week a removal was made in, and on this record four entries were logged retrospectively with approximate weeks for exactly that reason.
One more place it bites
When a removal is reversed and then reversed again.
Two entries on this record stopped, crept back, and were then deliberately reduced some months later. Each has three events and therefore three lags, all measured from the original stop, and an elapsed week for each.
Nobody has ever needed those numbers, and they are in the record because the alternative — overwriting the outcome column each time — would have destroyed the history that makes the entry interesting. An entry with one outcome is a row. An entry with three is a small story, and the only reason it can be told is that each event was written against the week it happened rather than against the state of the thing today.