What happens when a device goes quiet

Your watch dies on the second morning of a week away. You did not bring the charger, because you were not planning to train, and the whole point of the week was to stop thinking about any of it.

You come home. You charge it. You open the app, and there is a flat line through the middle of your month. Load down. Fitness down. A cheerful note about how well rested you are.

You are not rested. You walked eleven kilometres a day up and down a coastal path with a bag on, drank more than usual, slept badly on an unfamiliar mattress, and did a two hour hill walk on the Thursday that had your calves talking for days.

The gap in the chart says none of that. It says nothing at all, which is the problem: a system that reads nothing as zero has just made a confident claim about a week it knows nothing about.

Silence has at least six explanations

You went on holiday and left the charger behind. You got ill and did not train for nine days. You are injured and have stopped running entirely, which is the case where getting this wrong matters most. Your phone updated overnight and the sync integration silently lost its authorisation, which happens more often than any of us would like. You trained perfectly normally and stopped wearing the thing, because the strap was annoying you. Or you swapped devices and the new one has been recording happily into an account nobody has connected yet.

Six stories, one shape in the data. From inside a system, a rest week and a broken OAuth token look identical.

That kills the tempting answer straight away. You cannot infer the cause from the gap. Anything that claims to is guessing, and dressing the guess up as a number does not make it less of one.

The two ways to get this wrong

Most tools pick one of them.

Absence read as zero. No sessions logged, so training load is zero. Zero load for nine days, so your fatigue drains to nothing and your fitness declines on the usual curve. You come back, the system thinks you are fresh and slightly detrained, and it writes you a week accordingly.

If you actually rested, that is roughly right. If you were on your feet all week, or ill, or carrying an injury you did not log, it is wrong in the most expensive direction: the plan hands a solid week to a body that needed an easy one. This version hurts people in the week after the gap rather than during it, which is why nobody ever connects the two.

The plausible fill is worse, and worse for a subtle reason. The system notices the gap and quietly patches it. Carries your last known resting heart rate forward. Interpolates your HRV baseline across the hole. Estimates the load from your average week and drops it in.

The invented numbers are now indistinguishable from the measured ones. You look at your chart, the line is continuous, and nothing on the screen says the middle section was imagined. Every downstream calculation treats fiction and measurement identically, your baseline is partly built out of numbers nobody ever recorded, and the pollution outlives the gap by weeks, because baselines are computed from windows and that window now has a lie in it.

At least a hole in a chart is honest about being a hole.

What silence should do instead

The principle is easier to state than to build: absence of evidence is not evidence, and the system should say so out loud. In practice that comes out as four behaviours.

Confidence drops, and the drop is visible. Every signal we compute carries its own confidence, and coverage is one of the inputs to it. Four days without data does not produce a slightly stale readiness number. It produces a readiness number that says it is running on four day old inputs. The number and its trustworthiness travel together, so you can tell the difference between a good answer and an answer that had nothing to work with.

The baseline does not silently reset. Your normal is built from your own history, and a gap is not a reason to throw that away and start again. When the data comes back we are still comparing you against the same you, with the gap treated as a gap. A quietly reset baseline is how a runner comes back from two weeks off and gets told everything is fine, because fine is now being measured against the two weeks of nothing.

The gap is a question, not a conclusion. Trailwise will ask what happened: away, ill, injured, or just not wearing it. One sentence from you resolves in three seconds what no amount of inference would have resolved at all. You are the missing sensor.

The plan gets more careful rather than more confident. Coming back from an unexplained gap, the sensible default is a conservative return that reassesses quickly rather than a plan built on the assumption you are freshly rested. Tell us it was a genuine holiday and that assumption becomes reasonable and the plan can act on it. Until then it stays an assumption and gets treated like one.

Why you see more silence than you expect

Most people assume that connecting a platform means the data arrives. It very often does not.

Different sources give you very different amounts. Some paths carry detailed intraday series, the beat by beat material everything else is computed from. Others carry a daily summary and nothing underneath it. Some carry sessions but no overnight data at all. And several impose hard limits on how far back you can ask: one popular source will only ever hand over the last seven days of detail, another caps at twenty eight, and once that window has passed the data is not late. It is gone.

Which produces the failure mode that catches everyone eventually. Your integration breaks on the first of the month. You notice on the twentieth, because nothing tells you and the app looks the same either way. You reconnect, and you get the last week back, because that is all the window allows. The other two weeks never existed as far as your history is concerned.

A broken connection deserves to be loud. A silent integration failure looks like a small technical issue that resolves itself the moment you reconnect. What it actually is, is permanent data loss on a delay, and you find out long after the point where you could have done anything.

What to do about it, and where we are still imperfect

Log the reason in one line when you know you are going dark. “Holiday, walking daily, no watch” is enough. Written at the time it takes ten seconds. Reconstructed in November it is a shrug.

Check that your connections are actually connected, once a month. Open the thing, look at the last sync date, close the thing. That is the whole habit, and it is the difference between losing a day and losing a fortnight.

And do not read a flat chart as a rested body. A week with no data is a week you have no information about, which is not the same as a week that was easy. Your own memory of that week is better evidence than the chart, and the chart is the thing that will quietly convince you otherwise.

Three things I would rather set out than let you discover. If you never tell us what a gap was, we will never know; the system stays appropriately unsure and keeps the plan conservative, and there is a real cost to a cautious week you did not need. We take that cost on purpose because the other error is worse, but it is a cost. Data lost past a source’s retention window is genuinely lost, and we will not pretend to reconstruct what was never handed over. And partial silence is harder than total silence: a device that reports sessions but stopped reporting overnight data does not look broken, it looks like a normal week with slightly thinner recovery inputs, and it takes longer to notice than a clean stop does.

Trailwise is not a medical device. Neither is it a replacement for a doctor or a physiotherapist. Trailwise flags what is worth a conversation. It does not diagnose, and it will tell you when it does not know.