Sleep Efficiency: The Number That Drives CBT-I

6:15 am. I am doing math. Not the catastrophic kind, not the 2am calculation of doom. This is the morning-after kind, with a sleep diary and a calculator, and the number I am looking for is the one that will determine whether my sleep window gets longer, shorter, or stays the same.

The number is sleep efficiency, and it is the steering mechanism of CBT-I. Before I started treatment, I had never heard the term. I thought the only number that mattered was total hours. Seven was acceptable. Eight was ideal. Fewer than six was an emergency. Nobody had told me that how you spend your hours in bed matters more than how many you have.

The calculation

Sleep efficiency is total sleep time divided by total time in bed, expressed as a percentage.

Last night: I got into bed at 11:30pm. I estimate I fell asleep around midnight. I woke once at 3am for about twenty minutes. I woke for the last time at 6am. I got out of bed at 6:15am.

Time in bed: 6 hours, 45 minutes (11:30pm to 6:15am). Estimated total sleep time: about 5 hours, 55 minutes (6:45 minus 30 minutes of onset latency, minus 20 minutes of middle-of-night waking). Sleep efficiency: 5.92 / 6.75 = 87.7%.

That number is above 85%, which means things are moving in the right direction.

The math relies on estimates, and the estimates are imperfect. I do not know exactly when I fell asleep, because the moment of sleep onset is invisible to the person experiencing it. The diary captures perception, not polysomnographic data. My therapist told me this was fine: perception is what we are working with, and perception tracked over days produces averages that are clinically useful even if individual nights are imprecise. You are not trying to know exactly when you fell asleep on Tuesday. You are trying to know, roughly, what your sleep looks like across a week.

What the thresholds mean

Spielman’s original sleep restriction protocol established the thresholds that most CBT-I programs still use [1]:

Above 85-90%. Sleep efficiency is in the target range. The sleep window can be expanded by 15 to 20 minutes, typically by moving the bedtime earlier while keeping the wake time fixed. More time in bed is now warranted because the brain is using the existing time efficiently.

Between 80% and 85%. Hold. The current window stays the same. Sleep is consolidating but has not reached the efficiency target yet. This range feels like limbo: you are not bad enough to compress and not good enough to expand. Patience is the only instruction.

Below 80%. The sleep window may be compressed by 15 to 20 minutes. The brain is not yet using the current time in bed efficiently, and reducing the window increases homeostatic sleep pressure.

The floor is typically five hours. Sleep restriction does not go below five hours of time in bed, regardless of how low sleep efficiency is. Below five hours, the daytime impairment from sleep deprivation outweighs the consolidation benefit [4]. This floor exists because the goal is to consolidate sleep, not to eliminate it. The treatment is uncomfortable; it is not supposed to be dangerous.

Why efficiency matters more than duration

It is tempting to focus on total sleep time: how many hours did I get. I focused on it for years. I would calculate each morning whether I had reached six hours, and anything below that felt like a failed night. But total sleep time without context is misleading.

Seven hours of sleep in seven and a half hours in bed is excellent sleep. Seven hours of sleep in ten hours in bed is fragmented, inefficient sleep that probably includes extended waking periods and a bed that has become associated with wakefulness. Both people slept seven hours. One of them has a problem.

Sleep efficiency captures the relationship between the bed and sleep. A low number means the bed is functioning poorly as a sleep environment: too much time awake in it. CBT-I’s primary goal is to improve this ratio by compressing time in bed until the brain has no choice but to use it for sleep [5].

Trauer et al.’s meta-analysis found that CBT-I improves sleep efficiency by an average of 10 percentage points [3]. Someone starting at 68% might reach 78% after treatment, with continued improvement. The improvement comes from both reduced time awake in bed and, eventually, increased total sleep time as the window expands. The 10-point gain is an average; some people gain more, some less, and the trajectory is not linear. My own efficiency went from 62% in the baseline week to 72% after week two and 88% after week six. The jump from 62 to 72 felt like nothing, because I was still tired. The jump from 72 to 88 felt like a different life.

The titration process

Sleep restriction is not a static prescription. It is a titration, like adjusting medication dosage. The sleep diary provides weekly data. The sleep efficiency calculation determines the adjustment. The adjustment changes the sleep window. The new window generates new data. It is a feedback loop, and unlike the feedback loop that maintains insomnia, this one converges toward something functional.

This process continues over weeks. The typical trajectory: Week 1, sleep efficiency jumps because time in bed has been sharply reduced. Sleep is consolidated but short. Daytime sleepiness increases. You feel, in a word, terrible, and the efficiency number is the only indication that something is working.

Week 2 to 3, the brain adapts and sleep efficiency stabilizes above 85%. The window expands by 15 minutes. This expansion feels disproportionately rewarding. Fifteen minutes of additional sleep opportunity should not feel like a gift. After a week of going to bed at 1am, it does.

Week 4 to 6, gradual expansion continues as sleep efficiency remains high. Total sleep time increases while efficiency stays above target. The expansion decisions are mechanical: calculate the number, check the threshold, adjust or hold. The simplicity of the decision is part of its value. It removes judgment from the process and replaces it with arithmetic.

Kyle et al. documented the early costs honestly [4]. The first week of sleep restriction often involves reduced total sleep time and increased daytime somnolence. Participants in their study showed objectively impaired vigilance during the acute phase. This is not the treatment failing. It is the treatment creating the conditions for the efficiency gains that follow. The study also confirmed that these impairments were temporary, and that the efficiency improvements justified the short-term cost.

Doing the math yourself

If you are keeping a sleep diary, calculating sleep efficiency takes thirty seconds each morning. The weekly average is more useful than any single night. A bad night brings the average down, a good night brings it up, and the average over seven nights smooths out the noise.

What you are looking for is a trend. If your weekly average sleep efficiency is consistently above 85%, your sleep window is probably too narrow and you are ready for more time in bed. If it is consistently below 80%, your window may be too wide and you are spending too much time awake in bed.

A word of caution about precision. The diary is an estimate. The sleep onset time is a guess. The minutes of middle-of-night waking are approximate. Do not spend twenty minutes trying to get the numbers exact; the pattern across a week is what matters, not the decimals on any given night. I wasted the first few days trying to timestamp everything to the minute, which meant I was watching the clock, which is its own form of sleep-disrupting monitoring behavior. My therapist told me to round to the nearest fifteen minutes and move on. The math works fine with rough numbers.

What makes this hard

The hard part is trusting a number over a feeling. After a bad night, the feeling is: I need more time in bed, not less. The number may say the opposite. Sleep efficiency below 80% means the solution is less time in bed, not more, because the additional time is being spent awake and reinforcing the bed-wakefulness association.

This is counterintuitive in the way that most effective insomnia treatments are counterintuitive. More time in bed feels like the answer. Less time in bed is the answer, until the efficiency improves enough to warrant expansion. The number arbitrates between what feels right and what the data says, and in my experience, the data wins.

The other hard part is the mornings when the number goes the wrong direction. A week of 88% followed by a week of 76% is demoralizing. The instinct is to treat it as a failure. The protocol treats it as information: hold or compress, wait for the next week’s data, let the trend emerge. The number is not a grade. It is a gauge. Some weeks the gauge drops, and that tells you what to do next, not what you did wrong.

I check my number every morning. On good weeks, it is a small validation. On bad weeks, it is a guide. Either way, it is data, and data is more reliable than the 3am assessment of how things are going.

If your sleep is affecting daily life, talking to a GP or sleep specialist is a good first step. The American Academy of Sleep Medicine has a provider finder at sleepeducation.org. If insomnia is accompanied by depression, anxiety or other concerns, those are worth raising with your doctor as part of the same conversation.