random “what are you doing?” pings show where your time really goes
No timers to start and forget. Leave this tab open and it pops a quick question at random moments — answer with a tag, get back to work. Each answer stands for one slice of your day, and the pie and timeline add them up into an honest picture. Saved in your browser, exportable to CSV or JSON.
Sampling is on
No pings in this range yet. Leave sampling on and answer a few pop-ups — the picture builds itself.
Nothing logged yet.
Ask anyone how they spent yesterday and you will get a story; ask their calendar and you will get a plan. Neither is what actually happened. Conventional time trackers don't help much, because they depend on the one thing nobody reliably does — remembering to start and stop a timer. The result is a log of the days you tracked carefully, which are precisely the days that were not normal. If you have been hunting for a simple, lightweight time logger — an unobtrusive popup that just asks what am I working on? rather than making you run a stopwatch — that is exactly what this is.
Sampling flips the problem around. Instead of you reporting to the tracker, the tracker interrupts you: at random moments it pops a one-line question — what are you doing? — and you answer with a tag or two. Each answer is one honest data point about one real moment, taken without warning. Stack up a few days of them and the proportions of your answers converge on the proportions of your life. It is the method behind the cult-classic TagTime tool and the experience-sampling studies psychologists run; this is that idea as a plain web page — no install, no account, and every answer stays in your browser.
The statistical trick is the same one that lets a poll of a thousand people describe a country: a random sample is an unbiased estimator of the whole. In random mode the gaps between pings are drawn from an exponential distribution, which has a property worth pausing on — it is memoryless. At any instant, the chance of a ping in the next minute is exactly the same whether the last ping was thirty seconds or two hours ago. You cannot learn the rhythm, so you cannot — even subconsciously — save your focused work for the moments the tracker is watching. Whatever fills your hours gets sampled in proportion to how much of your hours it fills.
The arithmetic that turns answers into hours is honest, too. If pings average one per 15 minutes, each answered ping stands for about 15 minutes of estimated time. Eight pings tagged writing across a day means roughly two hours of writing — not because any ping measured a duration, but because that is what eight random hits imply. The pie chart you see is exactly this calculation, and when a single ping carries two tags, its minutes are split between them so the total still adds up to the time you sampled.
The estimate carries statistical noise, and the app does not pretend otherwise: error shrinks with the square root of the ping count, so a day's pie is a sketch while a few weeks' is a measurement. What it never carries is the bias of memory — the flattering rounding-up of deep work and rounding-down of scrolling that makes hand-written time diaries fiction. Missed pings even get their own gray slice, because an unanswered question at 3 PM says something about 3 PM.
Answer for the instant the ping fired, not for the block of time around it. If the pop-up appears while you are fetching coffee in the middle of a work session, the honest answer is coffee — the work sessions will collect their own pings in proportion. This feels pedantic and is exactly what keeps the estimate unbiased. Keep your vocabulary small while you are at it: a dozen reusable tags beat a hundred specific ones, because a tag that only ever receives two pings tells you almost nothing. The autocomplete chips exist to pull you back toward words you have already used.
Pick the gap to match the question you are asking. Auditing a workday in fine grain, 10–15 minutes is comfortable for a day or two; running ambient tracking for weeks, 45 minutes (TagTime's own default) fades into the background while the data quietly accumulates. And resist the urge to switch to fixed mode because predictability feels nicer — anything you always do on the hour will be systematically miscounted by a tracker that always looks on the hour. Random is the whole trick.
It is the same statistics opinion pollsters rely on: a random sample is an unbiased estimate of the whole. If pings arrive at random moments averaging one per 15 minutes, then each answer stands for roughly 15 minutes of your day, and the share of answers that say “email” converges on the share of your time that email really takes. No single ping proves anything; fifty of them paint a picture that is hard to argue with.
No — this is a plain web page, so it can only ping while the page is open somewhere. A backgrounded or minimized tab is fine: the ping is detected, a desktop notification can surface it if you have allowed notifications, and the tab title flashes. If the page was closed or your machine slept through ping times, those pings are reconstructed when you return and logged as missed rather than silently dropped. The practical setup is to pin the tab and let it sit.
No. Notifications are optional and off by default. Turning them on asks your browser for permission once, and after that they are used for exactly one thing: surfacing a “What are you doing?” prompt when a ping fires while the tab is not in front. Nothing else is ever sent, and you can revoke the permission in your browser at any time. Note that some mobile browsers do not support page notifications; the in-page pop-up always works.
They are kept, not hidden. A recent missed ping waits in a small queue so you can still answer it — the prompt shows the time it fired, and you answer for that moment. Anything older is logged with the tag missed and appears as its own gray slice in the chart. That is deliberate: knowing that a third of your afternoon pings went unanswered is itself information about your afternoon.
Yes. Separate tags with spaces or commas — “meeting project-x”, “cooking podcast”. In the pie chart the ping's estimated time is split evenly across its tags, so totals still add up to the time you actually sampled. The autocomplete suggests every tag you have used before, which nudges you toward a small reusable vocabulary instead of a hundred one-off labels.
Random, unless you have a reason not to. Random gaps are drawn so that a ping is equally likely at any moment, which means you cannot subconsciously schedule your slacking between pings — the estimate stays honest. Fixed intervals are predictable, which feels calmer, but anything you always do at the half-hour mark gets systematically over- or under-counted. Fixed mode is there for people who use the tracker as a recurring check-in rather than a measurement.
It improves with the square root of the number of pings. As a rule of thumb, an activity that received 10 pings has about a 30% margin of error, at 50 pings roughly 14%, and at 200 pings about 7%. So treat a single day's pie as a sketch and a few weeks' as a measurement. The big shares stabilize quickly; tiny slivers stay noisy the longest.
Yes — two buttons, no strings. CSV gives you one row per ping with an ISO timestamp, local date and time, tags, and the estimated minutes that ping represents; it opens straight into any spreadsheet. JSON gives you the same data plus your settings, suitable for scripts or backup. Both are generated locally in your browser and downloaded as plain files; nothing is uploaded anywhere.
No account and no cost. There is no sign-in and no email field. The tool is free and ad-supported, and everything you track is saved only in this browser.
Once the page has loaded, everything — pings, answers, charts, and export — runs entirely in your browser with no connection. Your answers are written to local storage on this device and never uploaded, which also means clearing browser data or using private browsing will reset them. Export to CSV or JSON if you want a permanent copy.
Apps like Toggl, Clockify, and TimeCamp are stopwatch-and-billing tools: you start a timer for a project and stop it later, which only captures the days you remember to. This is a lighter idea — an unobtrusive, lightweight time logger with nothing to start. A small popup asks what am I working on? at random moments and adds the answers up, so you get an honest breakdown of where the day went without running a single timer. If you have looked for a simple time tracking app, or a popup tracker that samples instead of nagging, that is the gap this fills.