Guide
Where the hours actually go, and how to find out in a week
Every business we talk to knows admin is eating the week and almost none can say where. That matters, because the tasks people nominate as the worst are reliably not the ones costing the most time. The annoying job that happens twice a month gets named. The three minute job that happens forty times a week does not, because three minutes never feels like anything. This is a one week method for finding out, and what the answers usually look like.
Why the guess is always wrong
Memory records irritation, not duration. A task that is frustrating, interrupts something else, or ends in an argument gets remembered as expensive. A task that is dull and quick gets remembered as nothing at all.
The arithmetic runs the other way. Frequency multiplies everything, and it is the variable people underestimate most.
Source: Arithmetic. Forty repetitions per week against one per month, at 46 working weeks.
The method, in one week
No software, no time tracking tool, no consultant. The objective is a rough count of frequency and a rough estimate of duration, and rough is genuinely sufficient: you are looking for the difference between 90 hours and 24, not for precision.
- Print a tally sheet for each person whose week you are measuring. One row per recurring task, a column for marks, and two blank rows for whatever gets discovered.
- Pre-fill the obvious rows. Answering the phone, replying to enquiries, entering an order, chasing a document, sending a quote, following up a quote, booking something in, invoicing, answering a status question.
- Mark a tally each time, not a time. Writing a duration takes thought and people stop by Wednesday. A pen stroke survives a busy day.
- Estimate durations once, on Friday, together, as a group. Group estimates of familiar tasks are close enough and take ten minutes.
- Multiply and sort. Frequency by duration, biggest first. This is the entire analysis.
- Ask one question of the top five: does this need a person's judgement, or does it need a person's hands. That question decides what is automatable.
What the answers usually look like
Across the businesses we have done this with, the same four things arrive at the top with remarkable consistency regardless of sector.
- Answering the same questions. Hours, availability, status, price ranges, where are you. Highest frequency of anything, and almost no judgement required.
- Typing information that already exists somewhere else. From an email into a system, from one system into another, from a form into a spreadsheet. Universally underestimated.
- Chasing. Documents, approvals, replies, payments. It is invisible because each instance is thirty seconds and it happens all day.
- Reconstructing. Working out what was agreed, what happened, what stage something is at. This is a symptom of nothing being recorded at the moment it occurred.
The judgement test
Once you have the top five, the useful split is not by task type. It is by what the person is actually contributing.
| Kind of time | What it looks like | What to do |
|---|---|---|
| Hands | Copying, retyping, looking up something that exists, sending a message you have sent a thousand times, chasing | Automate. This is the whole target and it is usually most of the hours. |
| Judgement | Deciding whether to take a job, pricing something unusual, handling an upset customer, choosing between two approaches | Protect. Give it back the hours the first row was consuming. |
| Hands, wrapped around judgement | Assembling a quote: gathering everything is hands, the number is judgement | Split it. Automate the assembly, keep the decision. This is where most real builds land. |
The counting mistakes that matter
Four things that will distort the result enough to send you at the wrong problem.
- Not counting interruptions as their own cost. A three minute task that breaks concentration costs more than three minutes. You do not need to quantify it, just note which tasks interrupt.
- Measuring a quiet week. Pick a normal week. If you have a season, measure inside it, because that is the week the system has to survive.
- Only measuring the owner. The owner has the least representative week in the business. Measure whoever answers the phone.
- Letting people self-report at the end. Friday recall produces the guess this whole exercise exists to replace.
What to do with the answer
You now have a ranked list with hours attached. Two things to do with it, and one not to.
- Take the top item that is hands rather than judgement. That is the first build, almost regardless of what it is.
- Write down the number before you start. Hours per month, dated. Without a before, there is no after, and every project afterwards becomes an argument about whether it worked.
- Do not start with the most annoying task unless it also happens to be near the top. It will feel better and return less, and it is how automation projects end up with nothing to show.
Where this connects
The automation scanner is the fast version of this. It asks nine questions about how your week runs and ranks the three jobs worth automating first, with hours attached, in about two minutes. The tally sheet is the accurate version, and doing the scanner first tells you whether the accurate version is worth the week.
Once you have the ranking, the ninety day roadmap is the order to build in, and the back office blueprint describes the layers the hours usually sit in.