Automatic tracking
An automatic time tracker records the day you actually had
A manual timer records the day you remembered to record. Those are different days, and the gap between them is the part you never invoice.
In short
An automatic time tracker runs in the background and records which applications, windows and projects you spent time on, without anyone starting or stopping a timer. It measures input activity and window focus, separates idle stretches from working ones, and produces a timeline you correct afterwards rather than a form you fill in beforehand.
- Timers to start
- 0
- Before idle is split out
- 30 s
- Focus sampling
- 3 s
- Input sampling
- 250 ms
The problem is not laziness, it is recall
Everyone who bills by the hour has had the Friday afternoon where they open a spreadsheet and try to reconstruct Tuesday. You know roughly what you did. You do not know whether the call ran forty minutes or seventy, whether the thing you fixed took an hour or most of the morning, or how many times you dipped back into a project you had already written off as finished.
So you round. And people round down, because a number you cannot defend is a number you would rather not have to defend. Every rounded-down hour is revenue that existed and was performed and was then given away at the moment of writing the invoice.
A manual timer does not solve this, it just moves the moment of failure earlier. Now the question is whether you remembered to press start when the client called, and whether you remembered to press stop when you got up.
What an automatic tracker measures
Four signals, sampled continuously. None of them require you to describe what you are doing while you are doing it.
- Which window has focus
- Sampled every few seconds. This is what turns a nine-hour day into "four hours in the design file, two in the repo, ninety minutes in a browser, forty in mail". It is also what lets time be assigned to a project after the fact, because the evidence of which project it was is in the record.
- How much input there was
- Counts of keystrokes and clicks, not their content. Two hours at the keyboard and two hours with a document open while you were in a meeting look identical to a timer and completely different here.
- Idle time, separately
- Thirty seconds without input and the stretch is marked idle rather than active. Idle time is kept and shown, not silently deleted and not silently billed. You decide which side of the invoice it belongs on, with the information in front of you.
- Real elapsed time, including sleep
- A laptop that was closed at six and opened at nine slept for three hours. Getting this wrong produces either a three-minute session or a three-hour one, and both are wrong. Stunda reads two system clocks and ends the session at the sleep boundary.
Manual timers against automatic tracking
Neither approach wins every row, and the second column is not a strawman. Manual timers are genuinely better at some things.
| Automatic | Manual timer | |
|---|---|---|
| Time you forgot to record | Already recorded, waiting to be assigned | Gone |
| Effort during the work | None | Start, stop, switch, remember |
| Effort after the work | Review and assign the timeline | None, the entries are already labelled |
| Intent | Inferred from evidence, corrected by you | You said so, so it is definitive |
| Offline work, meetings, whiteboards | Needs a manual entry, same as anywhere | Needs a manual entry, same as anywhere |
| Evidence if a client queries the invoice | A timeline with apps, windows and activity levels | A number somebody typed |
Stunda has manual entry too, because the honest answer to the fifth row is that some work happens away from the machine. Automatic tracking is the floor, not the ceiling.
Automatic does not mean unsupervised
The worst version of this category is a tool that decides what you were doing and files it without asking. Automatic capture is a strong signal and a weak judgment: it knows a browser was focused on a client domain for forty minutes, it does not know whether that was billable research or reading the news on a subdomain that happens to match.
So the timeline is a draft. You confirm, split, merge and reassign, and the tool learns the assignment for next time. The difference from a manual timer is not that the machine decides, it is that the raw material exists at all when you sit down to decide.
Questions
What is an automatic time tracker?
Software that records which applications and windows you use, and how much input activity there was, continuously in the background. It builds a timeline of your day without you starting or stopping anything, which you then review and assign to projects.
How does automatic time tracking know what project I am on?
It infers it from evidence: which application had focus, which window or document, and for browsers which site. You confirm or correct the assignment, and the rule is remembered so the same evidence maps to the same project next time.
What happens to time when I step away from the computer?
After thirty seconds without keyboard or mouse input the stretch is marked idle and shown separately. It is neither deleted nor billed automatically. You decide, with the timeline in front of you, which of it was a meeting and which was lunch.
Is automatic time tracking accurate enough to bill from?
It is more defensible than a remembered number, because there is a record behind it: which app, which window, how much activity, and when. The judgment about what is billable stays with you, which is where it belongs.
Do I still need manual time entries?
For work away from the machine, yes. Client calls on your phone, meetings, and thinking time on a walk are not visible to any tracker. Stunda supports manual entries alongside the automatic timeline for exactly this.