Sleep Regularity Index

Regularity is the key to healthy sleep. For many years now Sleep as Android makes your sleep regularity a central focus in your Sleep Score just beside the Duration measure.

In fact more evidence is emerging that regularity may play even be more important role in sleep than duration. See for instance Sleep regularity is a stronger predictor of mortality risk than sleep duration. You can get a similar message from the famous book “Why we sleep” by Matthew Walker.

Regularity is calculated as the variance of your mid sleep hour based on the longest sleep recorded for each day, the lower the better and is expressed in minutes.

Recently, another approach to regularity has become popular: Sleep Regularity Index (SRI) which has several advantages over variance.

SRI is the probability of you being in the same state (asleep or awake) at any two time points 24 hours apart. In other words, we are looking for the portion of your two consecutive days being in the same asleep/awake state at the same time.

Many thanks to Jeff HuangJi Won Chung, and Neil Xu (Brown HCI Research Group) for the idea of implementing SRI in Sleep as Android and for contributing the SRI calculation code.


Let say you get to bed at the same time but one day you wake up at 6:00 and the other at 7:00. Your SRI for that day would be 23 / 24  = SRI 96. The greater the difference between your days, the lower your score.

The SRI and the original Regularity score (Mid sleep hour variance) both have their PROS and CONS, so we will keep both approaches in Sleep as Android for now for your to experiment with. 


  • Used as a standard Regularity measure in new studies, so you can e.g. take your SRI score as calculated by Sleep as Android and see how this relates to various studies results. For instance see how your SRI relates to your mortality risk here and here, to cardiometabolic risk, academic performance or relapse in alcohol use disorder
  • Accounts for naps and awake periods withing sleep records, works well with poly-phasic sleep schedules where our variance measure only accounts for the largest sleep for each day ignoring any naps. 

Mid sleep hour variance PROS+

  • Better interpretability as Variance is measured in minutes where SRI is probability-based 
  • Even they may have important health impact, some irregular patterns may show smaller impact on SRI then on variance. For instance imagine a swift worker working more consecutive days at night and then switching for day work.
  • We can show Regularity of the very last record, where for final SRI we need to wait for the next sleep to get recorded. 

Philips et al. even states: “Sleep regularity was uncorrelated with sleep duration, suggesting that regularity captures another informative dimension of sleep. The SRI metric we used here captures a specific type of regularity–day-to-day differences in sleep patterns–and does not require a main daily sleep episode to be designated, which is advantageous in populations with highly irregular sleep patterns.

We have integrated SRI across various features in Sleep as Android. 


In charts you see actual daily SRI, long term trends for longer periods or how did individual days contribute to SRI.

And you can check if you have consistently lower/higher SRIs on certain days of the week:


SRI is now part of your Sleep advice. We will tell you if SRI trend is significantly negative/positive and even check for absolute SRI level in terms of published SRI-bases studies.

The absolute thresholds for a good / bad behavior are 80 / 60 respectively which is consistent with most of the studies. And we will alert you whenever your SRI increases / decreases by more than 5.  In a long term to short term comparison. 


Please let us know what you think about the SRI feature in Sleep as Android here



Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.