The founder time audit: find your Tuesday number

The fiction of the remembered week

Every founder carries a mental pie chart of their week. Strategy here, sales there, a slice for the team, a sliver of admin. The chart is sincere. It is also, in my experience auditing real calendars, fiction — and the research on time-use agrees: when people's self-reports are checked against actual logs, the gap routinely runs to hours per day. We remember the work that felt like our job. We forget the connective tissue — the forwarding, the chasing, the re-explaining, the copy-pasting between systems that don't talk to each other.

The forgetting isn't random, and that's what makes it expensive. The tasks that vanish from memory are precisely the small, repetitive, low-judgment ones — individually too trivial to register, collectively a second job. A founder who'd never accept a $40,000 invoice without reading it will leak that amount annually in eleven-minute increments, and report honestly at dinner that the week was "mostly strategy."

You can't fix an allocation you can't see. So: the audit.

Why Tuesday

We use Tuesday for a reason. Monday is distorted by the weekend pile-up; Friday by the wind-down; days with launches or crises by the launch or the crisis. Tuesday is the most ordinary day of the week — no drama, no excuses, just the business operating at its actual default. If you want to know what a system really does, you don't measure it during the fire drill. You measure it on a Tuesday.

And ordinariness is the point, because your business is mostly made of ordinary days. The exceptional ones write the stories; the Tuesdays write the financials.

How much of your Tuesday could a machine do? Count it. That's your number.

The audit: one day, four labels

The protocol is deliberately primitive — no app required, and the friction of writing by hand is part of the instrument:

1. Log live, in 15-minute blocks. Keep a note open all day. Every switch of activity gets a line with a timestamp. Live logging is non-negotiable: reconstructing at 6pm reintroduces exactly the memory bias you're trying to remove. The 9:42 "quick question" that ate twenty minutes must be captured at 9:42, because by evening it will not exist.
2. Record interruptions as tasks. If a call breaks your focus block, that's two entries, not one. Context switches are real costs wearing camouflage.
3. At day's end, tag every block with one of four labels:
   • O — Only me. Work requiring your judgment, relationships, or authority. Closing the key client. Setting direction. Deciding what not to do.
   • M — Machine. Rule-based, repeating, structured: answering routine inquiries, scheduling, reminders, invoice chasing, moving data between tools, assembling the same report. The test: could you write instructions a careful stranger could follow? If yes, software can hold those instructions. (What that looks like in practice: AI agents, plain-English guide.)
   • H — Human, not you. Needs a person — judgment, warmth, presence — but not your judgment. The delegation column.
   • X — Shouldn't exist. Work created by a broken process upstream: the third reminder that exists because the first ask had no deadline; the meeting that exists because nobody wrote anything down. Don't automate these. Delete their cause.

One honest day. That's the whole data collection.

Scoring your Tuesday Number

Add up the M blocks. Divide by your total working blocks. That percentage is your Tuesday Number — the share of an ordinary day currently performed by the most expensive employee in the company doing work a machine holds without noticing.

Benchmarks from the audits we've run and the broader automation research (McKinsey's task-level analyses point the same direction): founders guess 15–20%. The logged reality, for most established service businesses, lands between 30% and 60%. The gap between the guess and the log is the whole reason this article exists — the leak survives precisely because it's distributed too thinly to be felt.

While you're there, glance at the other columns. If O is under a quarter of your day, the business has quietly converted its founder into its operations department. If X is fat, you don't have an automation problem yet — you have a process problem that automation would merely speed up. (That distinction is the difference between systemizing and paving the cow path.)

The math that makes it hurt

Percentages are easy to shrug at. Money is not. Three lines:

• Your real hourly value. Not your salary ÷ 2,000. The value of your decision-making hour — what an hour of your full attention on the right problem is worth to the business. For most founders of seven-figure companies this is conservatively $200–500.
• Machine-work hours per day. Your Tuesday Number × your working hours. A 40% number on a 10-hour day = 4 hours.
• The year. 4 hours × $250 × 250 working days = $250,000 a year — spent buying work that software performs for a few hundred dollars a month.

And the cash is the smaller loss. The larger one doesn't fit in the spreadsheet: those four hours were taken from the O column — the deals not pursued, the product not rethought, the rest not taken. Decision quality degrades with volume (decision fatigue is a measurable tax), so the machine-work doesn't just consume your hours. It consumes the quality of the hours it leaves behind.

What to do with the number

1. Sort the M column by frequency × pain. Daily beats weekly; customer-facing beats internal. The top item is your first automation — one workflow, end-to-end, done completely. (The selection framework in detail: what to automate first.)
2. Delete the X column's causes this week. Free, immediate, and it shrinks the automation project before it starts.
3. Hand the H column to humans with written instructions. The audit already wrote the job descriptions — each H block is a bullet point in one.
4. Re-run the audit quarterly. One Tuesday, four times a year. The number should fall each quarter; if it doesn't, the systems aren't holding, and that's worth knowing while it's cheap.
5. Protect what the machine gives back. The hours you recover will refill with new machine-work unless the O column claims them first — calendar them before the vacuum does.

This audit, in a more instrumented form, is the first thing we run inside every business engagement — before any technology is discussed, because the audit decides whether the technology is even worth building. If the numbers don't show a clear return, we don't build. The Tuesday log is where the truth lives.

Frequently asked questions

What is a time audit and how do I do one?

A written record of where hours actually go versus where you believe they go. Founder version: one ordinary Tuesday, logged live in 15-minute blocks, then every block tagged only-me / machine / delegatable-human / shouldn't-exist. One day to run, 30 minutes to score.

What is the Tuesday Number?

The share of an ordinary working day a machine could do — routine replies, scheduling, chasing, data transfer, reports. Founders guess 15–20%; logged reality usually lands 30–60%, which priced at a real decision-making rate is typically a six-figure annual leak.

How many hours a week should a founder spend on operations?

As few as the systems allow — the test is category, not hours. If most of the week is reactive (responding, routing, chasing), the business is spending its most expensive resource on its cheapest problems.

What should a founder automate first?

The highest scorer on frequency × structure × time cost. In practice: lead response, scheduling and reminders, routine customer questions, invoice chasing, data transfer. One workflow end-to-end before spreading across ten.