How to Measure Productivity Without Counting Hours

The best productivity measure is one you can read off your own system in under a minute, without a dashboard, a manager or anyone’s permission. That rules out almost everything the search results will offer you. Hours logged, activity rates, keystroke monitoring: all of it measures employees, and you are not trying to audit an employee. You are trying to find out whether this week actually produced anything.
The honest conclusion, up front: if you already run a GTD system, you do not need a new tool to work out how to measure productivity. Your project list and your waiting-for list already hold the numbers. Five of them, read in about fifteen minutes a week, inside the review you are supposed to be doing anyway.
What you are actually trying to measure
Formally, productivity is a ratio of output volume to input volume. That is the definition used in the OECD’s productivity measurement manual, and it works beautifully for factories, where output is units and input is hours or capital. It falls apart the moment the output is a decision, a negotiated contract or 1,200 words that finally say the right thing.
So the ratio is not available to you. What is available is the distinction underneath it: activity versus output. Hours at the desk, messages sent, tasks ticked, inbox cleared, all activity. Projects moved to done, commitments kept or honestly renegotiated, decisions made that were stuck last week, that is output.
Before you adopt any measure, apply one test. If this number went up next week, would anything of value have changed? Messages sent fails it. Hours logged fails it badly, because the number rises on your worst, most distracted days. Projects closed passes. Oldest waiting-for item in days passes, because a fall means something genuinely unstuck.
One more constraint, and it matters more than it sounds. The measure has to be visible to you alone, from data you already generate. The second measurement needs a plugin, a trial licence or a conversation with somebody, you will stop doing it. Probably inside a fortnight.
The data your task system already holds
Every serious task manager timestamps completion and does not make a fuss about it. OmniFocus keeps a Completed perspective. Things has the Logbook. Todoist has completed tasks and an activity log. None of that is marketed as analytics, which is why most people never look at it.
Four seams are worth knowing about:
- Completions, at both project and action level, with dates attached.
- Stalled projects: anything on your project list with no available next action. Not “no next action ever”, but nothing you could do today if you wanted to.
- Waiting-for ageing: delegated items, sorted by the date you added them, with the oldest at the top and a name attached to each.
- Inbox behaviour: how much arrived, how much you clarified, how much is still sitting there unprocessed when you open the review.
Export the completion data rather than relying on memory, and rather than screenshotting a list. Memory flatters good weeks and catastrophises bad ones. A screenshot loses the dates, which are the only part you actually need, and it cannot be pasted into a spreadsheet or handed to a summariser later.
Six measures worth tracking week to week
Five of these go in a weekly line. The sixth is sampled occasionally, not logged.
1. Projects closed
Counted at the project level. This is the whole point. A project is a commitment with an outcome, and you cannot manufacture more of them by reorganising your list, which is exactly what makes it honest.
2. Stalled project count
The number of projects with no available next action. You want this falling. If a third of your project list is stalled, your weekly review has lapsed, and no amount of extra effort during the week will touch it. This is the single most diagnostic number in a GTD system, and almost nobody tracks it.
3. Oldest waiting-for item, in days
Not the count, the age of the oldest one. Nineteen days means somebody has been sitting on your work for nearly three weeks and you have been quietly absorbing the delay. Hours logged will never show you that. Working harder will not fix it either. A chase email might.
4. Commitments renegotiated versus commitments quietly dropped
The integrity measure. GTD treats a commitment as live until you complete it or renegotiate it with whoever you made it to, including yourself. Dropping it silently is the thing that erodes trust in your own system. One renegotiated, zero dropped, is a good week even if nothing closed.
5. Weekly review completed: yes or no
Binary. No partial credit. Track it for a quarter and you will find it predicts the other four better than anything else you could measure, because a missed review is how stalled projects accumulate unseen.
6. Time-to-captured (sampled, not logged)
Seconds from having a thought to that thought sitting in one trusted inbox. The only input measure worth bothering with, because it is the single gap between an idea and the system that acts on it. Do not log it continuously. Test it now and again, and compare routes rather than averaging them: voice assistant, phone quick capture, index card in a pocket. Time each from thought to item landed. The slowest route is the one you have already stopped using without noticing, which is a capture problem worth fixing properly before you worry about anything downstream.
Measures that look useful and are not
Hours logged and activity rates. Trivially gameable, and they reward presence over finished work. A nine-hour day spent re-reading the same brief scores higher than a two-hour day where you finally made the decision.
Raw task completion counts. This is the most common self-measurement mistake, and it is sneaky, because the data is right there and the number feels objective. It is not. You control how finely an action is sliced. “Draft proposal”becomes six subtasks and your weekly score triples while the same amount of work gets done. Keep action counts as context if you like. Never as the score.
Inbox zero as a trophy. Zero is a by-product of clarifying, not an achievement in itself. You can hit it by filing everything into a vague list called “Someday”, which is how a clean inbox and eleven stalled projects end up coexisting.
Streaks and gamified scores. They keep you busy on small, safe, tickable items. Meanwhile the project you are avoiding sits untouched, because it is hard and there is no points bonus for being frightened.
Tool-hopping. A new app feels like progress for about ten days. It also resets your history and hides your trend, which is the real cost nobody mentions when comparing apps. If you have been measuring inside the tool, migrating wipes it.
How to measure productivity inside your weekly review
Give it fifteen minutes at the end of the review you already do. Pull the five numbers, write them as one line, compare with last week. A real line looks like this:
2 closed, 11 stalled, oldest waiting-for 19 days, review yes, 1 renegotiated / 0 dropped.
On its own that line tells you almost nothing. Four of them stacked up tell you everything. Single weeks are noise: holidays, illness, a client who went quiet and a week when three things happened to land at once all occupy the same column.
Make the numbers readable in seconds rather than calculated. In OmniFocus Pro, the perspective rule builder will give you a Stalled Projects view filtered on projects with no available action and grouped by folder, plus a Waiting For view sorted by date added so the oldest item sits at the top. Both read off in about five seconds each. If you have not built custom views before, the perspectives guide on this site walks through the rule logic, and the OmniFocus interface does shift between versions, so check the current builder before copying anything exactly.
Keep the record itself somewhere plain and exportable. A single note, or one spreadsheet with six columns. Not inside the task manager. People who track inside the tool lose the trend every time they migrate, and you will migrate eventually, because everybody does. The measurement habit should outlive the app, the same way the rest of a system worth keeping does.
And when a week reads badly, resist the obvious response. Adding more next actions to a stalled project is how it got stalled. Take the project back through the natural planning model instead: purpose, outcome, brainstorm, organise, then next action. Nine times in ten, the project is not stalled because you ran out of actions. It is stalled because you never settled what “done”looks like.
Where an AI assistant helps and where it misleads
Useful jobs, genuinely: summarising a long export of completed items into something you can read; spotting which contexts or people keep recurring across a month; drafting the clarifying questions for an open loop you have written down vaguely four weeks running. That last one is the best use of the lot, because a vague project is usually a thinking problem, and a decent set of questions breaks it open.
Jobs it ruins: deciding what mattered. Judging your week. Producing counts from an incomplete export, which it will do confidently and wrongly. Ask for a total and you may get a plausible number that nobody derived from anything. Always check any count it reports against the source list.
Practically, strip client names and commercial detail before you paste any task export. Keep the summary with the work rather than lost in a chat log: a project page in ChatGPT works well as project support material for review notes, which is one of the patterns covered in the piece on which GTD jobs ChatGPT does well and which it wrecks.
None of this requires new software, and nothing here asks you to move apps. Knowing how to measure productivity honestly comes down to counting things you cannot inflate, writing them somewhere boring, and reading them a month at a time. If you want the method underneath these numbers rather than the metrics on top, start with the GTD fundamentals covered across this site and come back to measurement once the weekly review is holding.
Frequently Asked Questions
How do you measure productivity when your work has no obvious output?
Measure commitments rather than units. Projects closed, commitments kept or renegotiated, and decisions that were stuck last week and are not stuck now. The OECD’s output-over-input ratio was built for measurable units of production, and knowledge work rarely produces any, so borrow the activity-versus-output distinction and leave the formula alone.
Is time tracking a good way to measure productivity?
It is a good way to measure time, which is useful if you bill by the hour or want to know where a week actually went. As a productivity measure it is poor, because it rewards presence and rises on your most scattered days. Track time for billing or for a one-off audit, then stop.
How often should you review your productivity metrics?
Record them weekly, inside the weekly review, and read them monthly. Weekly recording keeps the data honest while the week is fresh. Reading a single week will mislead you, because illness, holidays and other people’s delays all land in the same column, so judge on three or four weeks together.
What is the difference between measuring activity and measuring output?
Activity is motion: hours at the desk, emails sent, tasks ticked, inbox cleared. Output is a change in the world: a project finished, a decision made, a commitment honoured. Activity measures can go up while nothing of value moves, which is exactly why they stay popular.
Can an AI assistant measure productivity for you?
It can summarise and pattern-spot across an export you give it, and it is good at drafting clarifying questions for a project you have described badly. It should not judge your week or generate counts, because it will produce a confident number from partial data. Check every figure against the source list before you write it down.
