For about two years I logged everything. Start time, stop time, project code. At the end of a day the total would read something like 8h 10m, and I would feel like I had earned the evening.
Then I looked back at a month of it and tried to answer a simple question: what did I actually produce? The log could not tell me. It could tell me precisely how long I had been available to my own computer.
Hours are a measure of input, not output, and in knowledge work the relationship between the two is loose enough that optimizing for hours quietly makes the work worse. The exception matters and I will get to it: in billable and shift contexts, hours are the unit of the contract, and tracking them is not a mistake. For everyone else, the more useful question is how many real focus sessions you got, whether the important thing moved, and what conditions produced your good days.
An hour of logged time tells you one thing with certainty: a clock ran while you were nominally at work. It says nothing about whether you were in one continuous piece of thought or fourteen fragments separated by Slack.
That distinction is the whole game. Two people log four hours on the same problem. One spent it in two uninterrupted stretches, the other in twenty-minute slices between meetings. The timesheets are identical. The output is not close, because reassembling context after an interruption has a cost, and paying it twenty times consumes most of the day.
The log has no field for that. It is measuring the container rather than the contents.
There is a named idea here, and it explains most bad metrics.
Charles Goodhart, an economist at the Bank of England, made an observation about monetary policy that the anthropologist Marilyn Strathern later compressed into the version everyone quotes: when a measure becomes a target, it ceases to be a good measure.
The mechanism is simple. Before you target it, a number is a symptom of something real and correlates with what you care about. Once you target it, people optimize the number directly and the correlation breaks, because there is almost always a cheaper route to the number than to the underlying thing.
Hours are a textbook case. As a passive symptom, time at work correlates loosely with effort. The moment hours become the thing you are judged on, whether by a manager or by yourself at 6pm, the cheapest way to move the number is to sit there longer. You can do that while producing nothing, and the metric will keep reporting success.
This is not a moral failing. It is what happens to any proxy you push on hard enough.
Here is the case that made me stop.
You have a problem that has been sitting for a week. One morning you see it from a different angle and it takes twenty minutes. That twenty minutes contained everything: the week of background thinking, the experience that let you spot the angle, the judgment about what not to build.
On an hour-based system, this is the worst day of your month. You logged almost nothing. Meanwhile the colleague who ground through a brute-force version over three days logged twenty-four hours and looks tremendously productive.
Do that for a year and what you learn is not good. You learn not to mention that it took twenty minutes. You learn to pad. You learn that elegant solutions are professionally risky and the safe move is to look busy. Any system that rewards slowness will eventually get it.
Presenteeism, being present rather than productive, is a well established idea in occupational research, and the office version was easy to spot: the person who never left before the boss.
Remote work did not remove it. It moved it into a status dot and a response time. Green all day, replies within ninety seconds, a Slack message at 7:40am to establish that you were at it early. Same performance, cheaper props, and worse in one specific way: appearing available requires you to stay reachable, and staying reachable is incompatible with the concentrated work you are pretending to do. Office presenteeism wasted your evening. This version eats the middle of your day.
If you are tracking your own hours and nobody is asking you to, notice whether you are performing for a manager or for yourself. The internal audience is not obviously easier to please.
I do not want to be absolutist, because there are cases where hours are exactly right.
Billable work. If you are a lawyer, consultant, agency or freelancer selling time, hours are the unit of the contract and you have to track them accurately. Worth noticing that hourly billing has the same Goodhart problem built in, which is why many experienced freelancers move to project or value pricing: it removes the incentive to take longer.
Shift and coverage work. Support rotas, clinical shifts, anything where the job is being present and available. Presence really is the deliverable, and hours measure the deliverable directly.
Capacity planning. When you need to know whether a team has room for another project, hours are a reasonable rough instrument, as long as everyone understands they measure availability and not throughput.
Compliance. Some jurisdictions and contracts simply require records. No debate to have.
The pattern: hours work when time itself is the product, and stop working when time is merely the container the product arrived in.
Three things have been more informative for me than any hour total.
Number of real focus sessions. A session means one task, distractions actually out of reach, and a defined start and end. Three of those in a day is a strong day. It is a far better predictor of what I produce than whether the day contained six hours or nine.
Completed units of work. Not tasks ticked, which is easy to game by ticking small things. Units that mean something: a feature shipped, a chapter drafted, a client problem closed, a decision made and communicated.
Whether the important thing moved. The question with the best ratio of usefulness to effort. Most weeks contain one thing that matters and a lot of maintenance. At the end of the week: did the important thing move, yes or no. You can answer that honestly in four seconds.
| Measure | What it captures | Where it fails | |---|---|---| | Hours logged | Availability, presence, billable time | Blind to fragmentation, rewards slow work, easy to inflate | | Tasks completed | Momentum, throughput on small items | Gameable by shrinking tasks, ignores difficulty | | Focus sessions | Protected, uninterrupted attention | Says nothing about whether the work was worth doing | | Did the important thing move | Actual progress on what matters | Coarse, subjective, useless for billing |
None of them is complete. That is the honest position.
Output measurement is harder and fuzzier than hour measurement, and it can turn into its own anxiety.
Hours have one enormous advantage: they are bounded and objective. You put in eight, you are done, and you can stop with a clean conscience. Output has no natural ceiling. If the standard is "did I produce something meaningful today," there will be days where the honest answer is no, because you spent the day reading background material, or untangling someone else's code, or thinking. Those days are real work and they look like failure under an output metric.
Creative and research work is especially badly served here. Some of the most valuable weeks produce nothing visible at all. If you switch from hours to output without accounting for this, you swap the guilt of not sitting long enough for the guilt of not producing enough, and the second one is more corrosive because there is no number that discharges it.
So here is the middle path I actually run:
Most people reading this are measuring themselves, and self-measurement has a different purpose: you are not evaluating, you are learning the shape of your own work.
Worth knowing:
Vanity data is everything else: streak counts you stop caring about, total lifetime hours, daily bar charts you glance at and never act on. The test is whether a number has ever changed a decision. If it has not, it is decoration.
I should be straight about the tension, since Deep Focus is my own app and it does record data about your work.
It records session count, total focus time, today's focus, average session length, a focus heatmap across days and times, and which apps tried to interrupt you. That is tracking, and I am not going to pretend otherwise.
The claim is narrower: session-shaped data answers better questions than hour-shaped data. Total focus time only counts time inside a session with blocking actually on, so it cannot be inflated by leaving a tab open all afternoon. Average session length tells you your real sustained capacity rather than your intended one. The heatmap shows where your good hours are, which is the single most actionable thing on the page, because you can move your hardest work into them. Distraction tracking names the specific applications that pulled at you, which is more useful than knowing that you were, in some general sense, distracted.
It is still a proxy. A completed session says nothing about whether the work deserved the attention, nothing in an app can tell you that, and any dashboard implying otherwise is selling you a feeling.
Probably not as a performance measure. Tracking your own hours mostly produces a number you use to judge yourself with, and it rewards sitting still. Tracking sessions, or simply noting what you produced, gives you information you can act on.
Yes, when you bill hourly, because then hours are the product. It is worth separating the two uses: track hours for invoicing, and judge your weeks by output. Many experienced freelancers eventually move to fixed or value-based pricing precisely because hourly billing penalizes getting faster.
Number of protected focus sessions, completed units of meaningful work, and whether the one important thing this week moved. The first is countable, the second is concrete, and the third is the one that catches a week where you were busy the whole time and nothing advanced.
It can. Hours are bounded and objective, so they can be finished. Output has no ceiling, and days of reading, thinking or debugging look like failure under a pure output metric. A session target avoids most of this: bounded like hours, but impossible to satisfy by sitting still.
Most people find two to four genuinely protected sessions is a full day, and the research-flavored consensus around deep work tends to land in the same territory. More than that usually means the sessions are not as protected as they look.
The month of logs I went back through was not a lie exactly. Every entry was accurate. It recorded, with real precision, how long I had been sitting near my work.
I had never once asked the log a question it could answer.
Get the latest productivity tips and Deep Focus updates delivered to your inbox
