The Interruption Log SOP
Purpose
To produce a defensible count of the unscheduled demands on one owner's attention over ten working days, coded so that the count points at a specific fix rather than at a general feeling of being busy.
The log exists because owners consistently misremember this in one direction. Long interruptions are remembered and short ones are not, so the recalled picture is a handful of big escalations, while the actual picture is usually dozens of short answers whose damage is the switching cost, not the minutes. Those two pictures lead to completely different fixes. One says hire; the other says write down three rules you have never written down.
Scope
Covers: every demand on the owner's time that was not on the owner's schedule when the day started, including calls, texts, walk-ups, radio or dispatch traffic, and messages the owner chose to open mid-task. It covers demands that arrive during field work as well as office time.
Does not cover: scheduled work that ran long, self-generated task switching with no external trigger, or time spent on the owner's own errands. Those are real problems and they are a calendar audit's job, not this instrument's. Mixing them in makes the log unusable, because the fixes do not overlap.
Duration: ten consecutive working days. Not five, because a single week is distorted by whatever went wrong in it. Not twenty, because the discipline decays and the last week's data is worse than no data.
Re-run: once a quarter, or four weeks after any rule written in response to a previous log, whichever comes first.
Roles and responsibilities
| Role | Who it usually is | Responsibility |
|---|---|---|
| Subject | The owner | Logs every entry within a few minutes of it happening. Does not clean up, defer, or batch entries to end of day. |
| Secondary logger | Office or dispatch | Logs interruptions they routed to the owner that the owner may not have recorded, especially calls the owner took while driving. |
| Reviewer | A partner, a lead, a spouse in the business, or an outside advisor | Reads the tally with the owner and asks the questions the owner will not ask themselves. Not optional. |
| Rule owner | Named per rule at review | Owns the written rule that comes out of the review, including telling the team it exists. |
The reviewer role is the one shops skip and the one that determines whether the log changes anything. An owner reading their own tally alone reliably concludes the problem is volume, because that conclusion requires nothing of them.
Procedure
1. Set up the capture before day one
One sheet, one field per column, on whatever the owner will actually carry. Paper on a clipboard in the truck beats a phone application the owner will not open with gloves on. Six fields, no more:
- Time in, to the nearest five minutes
- Channel: call, text, walk-up, radio, message opened mid-task
- From: the person's name, not their role
- Category, one letter from the code list in step 2
- Minutes, including the time to get back into the interrupted work
- Decidable elsewhere? Y / N / R, where R means "yes, if a written rule existed"
Do not add a "notes" field. It will be left blank by Wednesday and its absence costs nothing.
2. Use these six category codes
| Code | Meaning | Test |
|---|---|---|
| Q | Answer to a knowable fact | The answer is the same every time it is asked |
| U | Unblock or approval | Someone cannot proceed without your authority |
| D | Judgment call | Genuinely depends on the specifics of this case |
| E | Escalation | A customer is unhappy and it reached you |
| S | Supply or parts | Availability, substitution, ordering |
| M | Money | Billing, collections, payment, payroll |
Code once, immediately, and do not agonize. The distinction that matters most is Q against D: if you have given the same answer to this question before, it is Q, however much it felt like judgment at the time. Owners over-code to D, because D flatters the owner and Q says the answer should have been written down.
3. Record the minutes honestly, including reset
Log the clock time consumed plus the time it took to resume the interrupted work. If you have never measured your own reset, run three timed instances on day one and use that as your standard adder for the rest of the log. It is usually larger than expected on complex work like quoting and near zero on routine work like driving.
Under-recording reset is the single most common way a log understates the problem, and it understates exactly the category - short frequent Q hits - where the fix is cheapest.
4. Do not fix anything during the ten days
The instinct is to solve each pattern as you spot it. Resist it for the duration. Changing the system mid-log corrupts the second week's data, and you lose the ability to say what the baseline actually was, which is what you will need to know whether the fix worked.
Write observations in the margin. Act after the review.
5. Tally four ways at the end of day ten
Total events and total minutes, then break both down:
- By category, as a share of total events and, separately, as a share of total minutes. These two rankings are usually different and the difference is the finding.
- By originator, as a share of total events.
- By arrival hour, in two-hour bands.
- By the decidable-elsewhere field: the combined share of Y and R answers, out of total events.
6. Apply the triggers, in this order
Originator trigger, first. If one person accounts for 30% or more of total logged events over the ten-day window, the finding is person-specific. Handle it with that person directly - scope, training, or a clarity problem about what they are allowed to decide - before you conclude anything about the shop.
Order matters here. A category that turns out to be one person is not a system gap, and writing a shop-wide rule for it teaches everyone else that you now need to be involved in something they were already handling.
Category trigger, second. Any single category at 25% or more of total logged events over the same ten-day window earns one written decision rule. Step size: one rule per log cycle, for the largest qualifying category only. Several rules issued the same week get tested at their edges simultaneously, every edge case comes back to you anyway, and you conclude that written rules do not work.
Routing trigger. If Y and R answers together exceed 50% of total logged events, your load is a routing problem, not a volume problem, and hiring will not fix it. New capacity routes to you the same way the old capacity did.
Arrival trigger. If any single two-hour band carries 40% or more of total logged events, that band is where coverage goes. It is also, by definition, the worst place in the day for uninterrupted work, whatever the calendar currently says.
7. Run the review with the reviewer
Thirty minutes, four questions:
- Which trigger fired, and in what order?
- For the qualifying category, what is the rule, in one paragraph, that would have handled most of those events without you?
- Who owns that rule and who tells the team?
- What is the re-log date?
8. Re-log after four weeks and compare like with like
Compare event counts by category against the baseline, not minutes. Minutes move for reasons unrelated to your fix - one long escalation swings a ten-day total. Counts are the cleaner signal that a rule is holding.
A completed log
Ten working days. 58 logged events, 12.75 owner hours.
| Code | Events | Share of 58 events | Hours | Share of 12.75 hours | Average minutes |
|---|---|---|---|---|---|
| Q Answer | 21 | 36% | 2.40 | 19% | 7 |
| U Unblock | 13 | 22% | 3.20 | 25% | 15 |
| D Judgment | 8 | 14% | 2.00 | 16% | 15 |
| S Supply | 7 | 12% | 1.00 | 8% | 9 |
| E Escalation | 6 | 10% | 3.50 | 27% | 35 |
| M Money | 3 | 5% | 0.65 | 5% | 13 |
Read the two rankings against each other before anything else. Q is the largest by events at 36% of the 58, and nearly the smallest by time at 19% of the 12.75 hours. E is nearly the smallest by events at 10% of the 58, and the largest by time at 27% of the hours. Those are two unrelated problems. Q is a switching problem: 21 hits scattered through ten days, each cheap, and the damage is that no hour is whole. E is a time problem: 6 events averaging 35 minutes each, and the damage is that six times in ten days a chunk of the day disappeared.
An owner reading only the hours column would fix escalations and still never finish a quote.
By originator: techs 34 of 58 events (59%), customers direct 11 (19%), office 8 (14%), suppliers 5 (9%). Within the tech total, one apprentice accounts for 19 events, which is 33% of the 58 total.
Apply the triggers in order. The originator trigger fires first: 33% is at or above 30%, so this is person-specific before it is anything else. And the check that follows is the one that changes the whole conclusion: of that apprentice's 19 events, 14 were coded Q. Strip him out and Q across everyone else is 7 events. Take care with the denominator here, since that is this whole SOP's subject: removing his 19 events from the numerator means removing them from the base too, so the honest figure is 7 of the remaining 39 events, about 18%, not 7 of 58. Either way it lands under the 25% category trigger, but quoting 12% would be comparing a stripped numerator against a whole-log base.
So the category trigger, evaluated on the shop rather than on the log as a whole, does not actually fire. The correct action was not a shop-wide rule about answering questions. It was one conversation with one apprentice about what he is expected to decide himself, plus the two-page reference he should have been given in his first week.
By arrival hour: 25 of 58 events, 43% of the total, landed between 7:00 and 9:00. Above the 40% band trigger, so that is where coverage belongs, and it also settled a running argument with himself about whether the early morning was his best thinking time. It was not. It was his least defensible two hours.
Decidable elsewhere: Y or R on 33 of 58 events, 57% of the total, above the 50% routing trigger. Which, together with the apprentice finding, is a shop whose owner does not need another person yet.
References
- See related: The Owner's Calendar Audit
- See related: What an Owner Should Stop Doing First
- See related: How to Protect Planning Time When the Phone Keeps Ringing
- See related: Batching vs Switching: The Cost of Interruptions
- Trade-standard practice for work sampling and time-study measurement