Learn

Almost every argument about sleep is an argument about hours. Seven or eight. Whether you can bank them on the weekend. Whether five is survivable during a hard stretch. The largest study to test hours against timing put both in the same model, and timing won. Once researchers accounted for how consistently people slept and woke, adding total hours slept explained nothing further about who died and who didn't. The Evidence Behind The Consistency Claim The study ran on roughly 61,000 adults wearing wrist sensors, close to ten million hours of recorded sleep. People with the most regular sleep and wake timing had 20 to 48 percent lower all-cause mortality than the least regular group. Run head to head, timing consistency came out the stronger predictor of the two, and layering hours on top of it added no explanatory power. The finding is roughly two years old now and still sits almost entirely outside official sleep guidance. A perspective paper published this month makes that point directly: regularity keeps outperforming duration in the outcome data, and duration is still the number every guideline is built around. Which means the single highest-evidence move available here costs nothing, needs no hardware, and appears on no dashboard anyone is trying to sell you. Why The Morning Score Deserves A Demotion Pool the research putting consumer trackers next to what people actually report about their own sleep and the devices account for somewhere between 3 and 16 percent of the variation. Most of the picture is missing. For people who sleep well, device and person agree about 82 percent of the time. For people with insomnia, agreement drops to 39 percent. Read that second pair of numbers again, because it is the uncomfortable one. The tool performs worst for the person most motivated to consult it. Anyone sleeping badly enough to study a score every morning is exactly the person the score is least equipped to describe. The devices also run consistently optimistic under the hood. They overestimate how efficiently you slept and underestimate how long you were actually awake in the middle of the night, in some cases by as much as half an hour. What The Fake Score Research Actually Measured In one trial, 63 adults with diagnosed insomnia were told at wake time that their sleep efficiency the night before had been good or poor. The reports were invented. The group handed the poor report showed reduced alert thinking by evening at an effect size of 0.79, with fatigue up at 0.55. Effect sizes near 0.8 are large. That is a made up number producing a real, measurable change in how well someone thought for the rest of the day. An earlier version ran on 164 people with no sleep complaints at all. Those told they had below average REM scored 22.13 on a working memory test. Those told the opposite scored 34.81. Their own sense of how they had slept predicted nothing. The only variable that moved was the figure they were shown that morning. Where The Other Side Is Right None of this makes hours irrelevant, and nobody should read it as permission to run on five. A county-level analysis across more than 3,100 US counties, using survey data from 2019 through 2025, found insufficient sleep ranked second only to smoking as a predictor of how long people in a community live. It came in ahead of physical inactivity, ahead of diet quality, ahead of social connection. Chronic short sleep is its own serious problem and no amount of consistency fixes it. The timing finding is for people already sleeping a reasonable amount who want to know where the next lever is. For that group, the lever is when. Trackers earn their keep here too. Duration and timing tracked across weeks is precisely the thing a device does better than memory. The failure mode is the morning verdict, not the hardware. Two Moves Anchor the wake time. One alarm, seven days a week, weekends included, with bedtime free to move as much as the day requires. The hours tend to sort themselves out within a few weeks once the wake time stops drifting. Move the score from a daily verdict to a weekly trend. One night's reading is mostly noise, and a low one takes real performance out of your morning whether or not it was accurate. Read the two to four week average, and if you are going to look daily anyway, form your own read before you open the app. A clock that runs five minutes fast is still a useful clock, as long as you know it runs five minutes fast. New to The Second Act? Start with 2Yr Anniversary Parting Advice Recap.