Automatic time tracking accuracy: what it can and cannot know
· 7 min read
An automatic tracker knows exactly which app was in front and for how long; that part is as accurate as the Mac’s clock. It does not know whether you were working in that app, why you were there, or whether you had walked away. Read its output as “time the app was in front” and it is reliable. Read it as “time I worked” and you will be disappointed, and then you will start distrusting numbers that were never wrong.
What it measures precisely
At any moment, exactly one application on a Mac is active: the one whose menus are in the menu bar, the one that receives your typing. The system knows which one it is and tells any app that asks. An automatic tracker asks repeatedly, notes the answer and the time, and adds up the gaps.
That produces a record with no human error in it. There is no forgetting to start, no forgetting to stop, no rounding up the design block because it felt long, no leaving the timer running through lunch. If the editor was in front for four hours and twelve minutes, the record says four hours and twelve minutes.
The precision has a boundary, though, and the boundary is the word “front.” How the mechanism works is covered in automatic time tracking on a Mac: how it works and what it sees.
What it cannot know
Everything an automatic tracker does not know comes from the same gap: the front app is not the same as your attention.
Intent. The browser was in front for ninety minutes. Research, or drift? The record has no opinion. The same is true of a messaging app (work thread or friends), a terminal (a build running or a command being typed), and a video call (a meeting you ran or one you muted).
Presence. You left a document open and went to make coffee. If the display stayed awake, the document’s app was in front for the whole coffee, and the record says so.
Work happening off the Mac. A whiteboard session, a phone call, a conversation at the desk, reading a printed page. The front app during those is whatever you left there, and the record attributes your attention to it.
Second screens and second devices. The tracker knows the front app on this Mac. It does not know you were watching a reference on the iPad, or that the second display had a video playing while the editor was technically active.
The common accuracy traps
A few patterns trip people up repeatedly, and knowing them in advance keeps the numbers trustworthy.
The call that lives in another app. You take an hour-long call in a video app but keep your notes app in front the whole time. The record shows an hour of notes and no call. Nothing is wrong, but the “meetings” line will be low for that day.
The browser as catch-all. Many kinds of work happen in a browser tab: documentation, a web-based editor, email, a chat tool, and the video you watched at lunch. They all add up under one app name. If your work is mostly in a browser, the app-level record will be accurate and uninformative at the same time.
The walk-away. Leaving the Mac awake with an app in front while you step out is the single largest source of “that can’t be right” in an automatic log. The fix is at the operating-system level: let the display sleep after a few minutes, and lock it when you leave the desk.
The sleep gap. The opposite problem. The Mac sleeps during a long read, and the tracker records nothing while it is asleep, so a genuine hour of work that happened away from the keyboard is missing from the day.
The pattern across all four is the same: the record is right about the front app and silent about you. The difference between the two is the subject of the difference between time spent and time worked.
Idle detection: the fix that creates its own errors
Most automatic trackers address the walk-away trap with idle detection: if there has been no keyboard or mouse input for a set number of minutes, the tracker stops counting, and sometimes asks you afterward what to do with the gap.
This helps with the coffee break and hurts with everything else. Reading a long document produces no input. So does thinking, so does a call, so does watching a recorded talk for work. Set the threshold short and the tracker discards real work; set it long and the coffee break comes back. There is no value that is right for a whole day, and the “what do you want to do with these twelve minutes” prompt is exactly the mid-day question automatic tracking was supposed to remove.
It also needs to watch your input, which means a permission (Input Monitoring or Accessibility) that lets the tracker observe keystrokes system-wide. Some people accept that trade for better idle handling; others would rather let the display sleep and read the numbers with the limits above in mind.
How to read an automatic log honestly
- Read each line as “this app was in front for this long.”
- Find the walk-aways. A suspiciously round block in a single app across lunch is usually one. Subtract it mentally, or let it stand and remember it.
- Attribute the browser. Look at the browser line and decide, from memory, roughly what share was work. You were there; you will be close.
- Add back the off-Mac work. A whiteboard hour does not appear. Note it in a sentence if it matters.
- Look at the week, not the day. Day-level errors are large and random; week-level patterns are stable. The editor is in front for twenty hours a week, or it is not, and one coffee break does not change that.
Punchcard’s answer: app names, a receipt, no idle knobs
Punchcard is an automatic tracker that accepts these limits rather than papering over them. It notices which app is in front (app names only, never window titles, document names, URLs, keystrokes, or screen contents) and at the closing time you set prints a paper receipt of the day, itemized by app with a day total and a stamp. There are no idle-detection settings to configure, and it asks for zero macOS permissions, including none of the ones that would let it watch your keyboard. The honest reading of a Punchcard line is “this app was in front for this long,” and the receipt does not claim otherwise.
That means the walk-away trap applies, and the remedy is the Mac’s display sleep and screen lock, not a setting in the app. It also means Punchcard cannot split the browser by site, cannot tell a call from a note, cannot attribute time to projects or clients, and will not prompt you about a gap. If you need a tracker that guesses at your attention, this is not it. If you want an accurate record of the front app, read once at the end of the day, with a week roll on Sunday to steady the numbers, it is built for exactly that.
Questions
Why does my automatic tracker show more hours than I think I worked?
Almost always walk-aways: the display stayed awake with an app in front while you were elsewhere. Check the system display-sleep setting and the lock-on-leave habit. The second cause is the browser line including non-work time, which the tracker cannot separate.
Why does it show fewer hours than I worked?
Sleep gaps and off-Mac work. A long read while the display slept, a call on the phone, a meeting in a room. Those hours happened and the Mac did not see them. Note them in a sentence rather than trying to make the tracker record them.
Is a tracker with window titles more accurate?
It is more detailed, not more accurate. Titles let it split the browser line by page and the editor line by document, which answers “which” better. It still cannot tell whether you were present or paying attention, and it reads far more about you to get there.