Research

The Time Audit

You cannot state where your hours go. Your logs can.

By Merryl DMello, DMello Advisory · 4 min read

The claim: owners and operators of small companies systematically cannot say where their working time actually goes, and the number they would guess first, total hours, is the least useful one. The number that changes decisions is active minutes per unit of output: how many minutes a human is actually at the keyboard per explainer video, per bid pack, per outreach round. That number is measurable from work logs most companies already have and never read, and in the one audit run so far it reclassified a one-person operation as covering seven conventional roles in 25.1 active hours a month.

This paper describes the audit method and its first results, run on dmello-advisory's own practice under the same rules a client would get.

The measurement, not the guess

The audit lists every recurring task, then measures two things per task: units per month, and active minutes per unit. Active time is time a human is actually at the keyboard deciding, checking, or approving. Elapsed time is deliberately excluded: a render that takes an hour and 15 minutes of review counts as 15 minutes, because the hour needs no human.

Three results from the first audit, which covered 22 recurring tasks:

The monthly totals: 15 explainer videos, 7 bid packs, 3 outreach rounds, 10 qualified leads, 17 LinkedIn posts, and 25.1 active hours of human attention across all 22 tasks. The same output conventionally sits across seven named roles. An explainer video costs 15 minutes of review; a studio budgets two days for it.

Every "usual estimate" in the table is marked as an estimate. That is a rule of the method, not a disclaimer: a line only counts if it has a frequency, every estimate is labeled, and every number in the register carries a source.

Why the believed number is wrong

Before the audit, the honest answer to "how much time does the pipeline take" was a shrug with a number attached. The audit found two structural errors in how operators account for time.

First, machine time is invisible in memory but dominant in logs. Nobody remembers the two-hour render fondly, but everybody reports the pipeline as heavy because of it. Active time strips it out and the work shrinks to what actually needs a person.

Second, the emotional maximum stands in for the average. The outreach round that took three hours once becomes "outreach takes three hours." The audit measured the heaviest genuine manual block at 4.5 active hours a month, concentrated exactly where the method predicts: steps that are still done by hand (sending, tracking, logging replies) while the drafting is already automated. The next automation target fell out of the ledger without a strategy meeting.

The receipts discipline

An audit that produces flattering numbers is worse than no audit, so the method enforces redaction and sourcing before analysis: names of tools, models, and people replaced with placeholders, monetary figures recorded as present but redacted, verbatim quotes capped at 40 words and always with file name and date. The audit runs on the operator's own session transcripts, 10,436 prompts, which record what was actually done rather than what is remembered as done.

There is a precedent for this standard of evidence. The author's reporting-dashboard work at a vaccine manufacturer served 35+ stakeholders and consumed 10 hours per person per month of manual effort before automation; that figure survives scrutiny because it was measured in the same way, per unit and per person, not remembered. The Time Audit applies that discipline to a whole operation in an afternoon of scripting.

What an operator should do with this

Run the register on one month of your own logs: every recurring task, its frequency, and active minutes per unit from your own timestamps. The first finding will be that your heaviest manual block is not your most annoying task, it is your least automated one. The second will be that you are already doing the work of several roles and cannot prove it, which matters exactly when you negotiate budget, price, or headcount.

What would change my mind

• An operator who can state their active minutes per unit for their top ten recurring tasks within 20% of their logs, without measuring. Nobody has passed this yet.
• Evidence that self-reported time diaries agree with log-measured active time. The literature on self-report says otherwise, and the first audit's before/after gap agreed with the literature.
• A case where the audit found the believed allocation correct. One audit is one audit; the method invites replication on any company's logs.

The author runs dmello-advisory and performs the Time Audit (German: Zeitprüfung) as a fixed-price engagement for small and mid-size companies: your own register, measured from your own logs, every number with a source. Eighteen years building and leading products at Intel, HP, Rackspace, BioNTech, and Phoenix Contact. dmello-advisory.com.

All research papers

Last updated: 16 October 2026