How Even Works
A guide for clinicians to what the app collects, how it reaches its suggestions, and where its evidence stops.
What it is
- A quick mood and stress diary on the iPhone
- A personal comparison of those entries with sleep, activity, calendar and alarms
- Prompts timed before moments that have been harder for that person
What it is not
- Not a diagnostic tool or a treatment
- Not monitored by anyone, and not a crisis service
- Not yet studied for benefit. It is used by a small invited group.
The clinical idea
Two established ideas shape the app. Ecological momentary assessment records how someone feels in daily life, close to the moment, which reduces recall bias compared with asking at a later visit [1]. Just-in-time adaptive interventions try to offer support at the moment a person is likely to need it and able to use it, based on information that changes over time [2].
Even applies both at the scale of one person. It does not compare anyone with a population or a norm. Each suggestion comes from that person’s own earlier entries.
What it collects
| Signal | Detail | Source | Stored |
|---|---|---|---|
| Mood | Five levels, Rough to Great | Self-report | This iPhone, plus the account |
| Stress | Optional, 1 to 5 | Self-report | This iPhone, plus the account |
| Notes and tags | Free text and labels | Self-report | This iPhone, plus the account |
| Sleep | Hours asleep and bedtime, from recent nights, or up to a month on the first sync | Apple Health | Account and iPhone |
| Heart and activity | Resting heart rate, heart rate variability, steps, exercise minutes | Apple Health | Account and iPhone |
| Calendar | Events for today and the next 7 days: title, time, calendar name and a category. Titles are stored. The Shortcut skips calendars the person turns off, anything that still arrives is discarded, and saved events can be removed. | iPhone Calendar | Account and iPhone |
| Alarms | Name, time and repeat days | iPhone Clock | Account and iPhone |
| Notification plan | Times and what each is for, including when a harder moment is expected or the hardest part of the day, plus time zone and latest check-in time. These come from ratings and answers that otherwise stay on the device. | The app | Account and iPhone |
| Sign-in | Email address, name if Google gives one, and a description of the browser for each session | Sign-in, or Google if the person chooses it | The account. Codes are sent through an email provider. |
| Breathing exercises | When each one finished its first round | The app | This iPhone only |
| Weekly reflection | Four questions about the past 7 days: how often the person snapped or lost their temper, how often small things set them off, how easily they calmed down when upset, and how often they woke up feeling rested. Combined into a 0 to 100 score. Offered 7 days after the first check-in, then weekly. Written for the app and not a validated scale. | Self-report | This iPhone only |
People choose in the app which kinds of data and which calendars the Shortcut sends. Health values are read by an Apple Shortcut when it runs. The app can’t read Apple Health directly, and the iPhone must be unlocked for the Shortcut to read it.
How a pattern is found
Even uses simple comparisons of averages. There is no machine-learning model, and no statistical test is applied. The rules are short enough to list:
| Question | Rule |
|---|---|
| Is a weekday harder? | Last 8 weeks. At least 3 rated check-ins on that weekday, and an average at least 0.4 points above the overall average. |
| What goes with stress, mood or energy? | Last 8 weeks. Compares check-ins with and without each of these: an activity tag; during an event or up to 90 minutes after it; after less sleep than usual; the usual sleep or more; the usual steps or more; a workout; heart rate variability or resting heart rate above the person’s usual; a busier calendar than usual; a first alarm before 7 AM; and the 2 hours after a breathing exercise. Needs at least 5 check-ins each way and a 0.4 point difference. Each subject appears once, with its strongest result: a tag, a calendar category or a measure about the same thing, or comparisons sharing at least 80% of their check-ins, count as one, with the others in its detail. If they disagree, it is marked mixed. Events with no category are left out. “Usual” is the person’s median over the last 28 days, once there are 7 days. Until then, the app uses under 6.5 hours of sleep, 7 hours or more, and 6,000 or more steps, and says so. Short sleep means half an hour under their usual. Morning runs from 5 AM to 11 AM, midday from 11 AM to 2 PM, afternoon from 2 PM to 5 PM and evening from 5 PM to 5 AM. A day with no exercise value counts as no exercise when the sync covered that day and the person reported exercise in the last 28 days; otherwise it stays unknown. Busy time counts overlapping events once. A back-to-back stretch is timed events with gaps of 10 minutes or less, lasting at least 120 minutes, compared for the 2 hours after it. A suggestion marked done is compared for the 2 hours after, under the same rules, and ranks above a correlation about the same subject. A pattern that doesn’t hold in both the older and newer half of the check-ins is labeled an early sign. Heart measures are personal context, never a health assessment. |
| How does mood vary? | Insights also shows average mood by time of day, with at least 5 check-ins for each, and before and after calendar events, with at least 5 paired events and the same 0.4 point difference. |
| When is it shown? | A prompt needs at least 0.4 points, the same as a pattern. |
A fictional example. Suppose check-ins near work events average 3.8 and all other check-ins average 3.0. The difference of 0.8 passes the threshold if each group has at least 3 entries. The app would say that stress “tended to be” higher, with the counts available. It would not say work causes stress.
Proactive support
Notifications are optional, and the person can see each one and why it was planned. A short set of questions, offered after the first check-in, sets the first schedule, and every setting can be changed afterward. The plan covers the next 7 days, each time arrives within about 5 minutes, and it refreshes when the app is opened. If the app stays closed, the last week repeats, while nudges before specific events stop.
| Type | When | What a tap does |
|---|---|---|
| Morning brief | 15 minutes after the morning alarm, or 30 minutes into the day. Can name the weekday pattern, a short night, and the first event’s category with how full the day is, and says which sync it came from. It drops sleep and calendar facts when the last sync is over 20 hours old. It shows what the app last wrote, so it can trail a sync the app hasn’t opened since. | Opens a check-in |
| Check-in | Spread across the day, favoring usual check-in times from the last 30 days. A check-in time planned at least 8 times with under a quarter answered moves up to 60 minutes toward a time the person does check in, or 30 minutes later, at most once a week, and the plan says why. Bedtime, brief and harder-time notifications never move. This is learned on the device and can be turned off in Settings. A person can also ask for one pair of check-ins a day, 20 minutes before and 5 minutes after an event, for a category with 3 or 4 before-and-after pairs. These replace ordinary check-ins. Skipped after a check-in in the last 90 minutes. | Opens a check-in |
| Harder time of day | The part of the day the person named as hardest | Opens a check-in |
| Around calendar events | A check-in that would land during a timed event moves to 5 minutes after it ends and stays at least 45 minutes from other notifications. If it no longer fits the day, it’s left out. The wind-down plans for tomorrow’s first timed morning event as if an alarm rang an hour before it. All-day events are ignored. | The plan in Settings shows the same times |
| Before a harder moment | About 20 to 30 minutes ahead of an anticipated moment. At most 2 a day, 2 hours apart. | Opens the breathing exercise |
| Wind-down | 20 minutes before usual bedtime, or earlier when tomorrow’s alarm calls for it | Opens a check-in |
| Sync reminder | A chosen time, only if nothing has synced in 20 hours | Offers to run Health Sync |
What the person is asked to do
- A breathing exercise. Four rounds of four-second inhale, hold, exhale and hold, which is under 4 breaths a minute, and about a minute in all. Reviews link slow breathing with changes in heart rate variability and with relaxation, though studies are small and varied [3]. The app says to breathe normally and stop if it feels uncomfortable.
- A sleep target. In the evening, the app works back from tomorrow’s alarm using the person’s usual recent sleep, or 8 hours when there’s too little data. That sits within the 7 to 9 hours recommended for most adults [5]. Sleep loss has been shown experimentally to heighten negative responses to mild stressors [4], which is why sleep is a central signal here.
- Something from their own history. If workouts have tended to go with lower stress and there’s been none today, the app suggests some movement. A check-in with high stress, low mood or low energy in the last 6 hours steers the choice toward what has gone with relief from that. The wording stays at “tended to”, and the counts are one tap away.
- A plain next step. The first screen lists up to three things to do, such as “take a minute” or “run Health Sync”, each with its reason where there is one. Until setup is finished, it shows one setup step at a time instead: add Even to the Home Screen, sign in, add and run Health Sync, and, after the first check-in, choose notifications. Each step can be put off, and checking in never waits for setup.
Garmin, Peloton and other apps
Even reads Apple Health, so an app that shares with Apple Health needs no separate connection. Apple Health is the recommended route because it adds one consent screen and no new accounts. Direct connections add cost and more companies holding the data.
| App | Through Apple Health | Not available this way | Direct connection |
|---|---|---|---|
| Garmin | Writes data to Apple Health, reportedly including steps and heart rate. Other categories should be checked on the phone. | Garmin’s own scores, such as Body Battery and its stress score | Possible through its developer program, which reopened in October 2026. Needs approval and has limits and terms to review. |
| Peloton | Sends workouts to Apple Health when its write access is on | Advanced ride metrics such as power | No official public route found |
| Hume Health | Hume reads from Apple Health. Its help pages don’t say whether body measurements are written back. | Possibly body composition | No public route found |
Workouts recorded by another app may not count toward exercise minutes, which is the measure Even reads today. Until that’s confirmed on a phone, the Exercise tag on a check-in is a reliable way to record them. This assessment reflects published information in November 2026, and these services can change.
Safety
- If a person rates stress 4 or 5, or mood Rough or Low, on at least 3 of the last 4 calendar days, counting today, the app suggests talking to someone and links to support.
- Settings lists crisis contacts, including 911 and the 988 Lifeline, and says plainly that the app can’t respond in a crisis.
- Nobody reviews entries. There is no clinician view and no escalation. High scores alert no one. Even should not be relied on for monitoring risk.
- The app does not screen for depression, anxiety or any other condition, and does not advise on medication.
What limits these results
| Limit | Why it matters |
|---|---|
| Very small samples | Three entries per group is enough to display a pattern, and far too few to be confident in it. Counts are shown so readers can judge. |
| Many comparisons | Tags, event types, weekdays, sleep and alarms are all checked with no correction for multiple comparisons, so some patterns will be chance. |
| Daytime sleep | A main sleep that starts between 10 AM and 6 PM is treated as a nap, so people who sleep during the day have less sleep data. |
| Time zones | Time of day uses the phone’s current zone, so history made while traveling is read in the home zone after returning. The plan follows the new zone the next time the app opens. |
| Confounding | Short sleep, early alarms and busy days often coincide. The comparisons consider one factor at a time. |
| Direction of effect | Exercise may lift mood, or feeling well may make exercise more likely. Daily totals can’t separate the two, or show whether a workout came before or after a check-in. |
| Who checks in, and when | People may record more on bad days, or stop when struggling. Prompted entries may also differ from unprompted ones. Self-monitoring can itself change mood [1]. |
| Expectation | Being told a hard moment is coming could raise anxiety in some people. This has not been measured in Even, and the prompts can be turned off. |
| Missing data | Sync needs the person’s iPhone, unlocked. Each run fills in up to 7 days of steps, exercise, resting heart rate and heart rate variability, so a longer gap stays empty. Calendar data covers today and the next 7 days, so its history builds up one sync at a time, and old alarms aren’t kept unless a sync recorded them that day. |
| Wearable accuracy | Sleep and heart measures come from consumer devices and vary by device. When two devices record the same night, overlapping time is counted once, but the devices can still disagree. Alarms that were skipped once may still appear. |
| Energy ratings | Energy is only counted when someone sets it. Older check-ins hold 3 for an untouched slider, which can hide a real difference. |
| Delivery | iPhone settings such as Focus can hold back or hide notifications, and the app can’t tell whether one was seen. |
| No outcome data | Even has not been evaluated for whether it reduces stress or improves wellbeing. |
Privacy
- Using Even requires an account, so mood, stress and notes are stored on the device and in the account as readable text. Erasing synced data removes them from the account and from the person’s devices. Reflections, breathing sessions, suggestions marked done and which planned check-ins were answered stay on the device. Check-ins saved while offline are lost if the app is deleted before they upload.
- Synced records are limited to the verified account that owns them, and each person can erase them or delete their account from the app. Removing a tester deletes their synced data, and devices unused for 90 days disconnect.
- Data travels over encrypted connections. It is not end-to-end encrypted, so the service’s operators could technically access synced records, including calendar titles, alarm names and any synced check-ins and notes.
- A first name can appear in notifications at the start and end of the day and at hard moments, so it can show on the Lock Screen.
- The app has no ads or analytics. It is not a covered clinical record system, and should not hold information a clinic would be required to protect.
Where a clinician’s view would help
- Are the thresholds too eager, and how should uncertainty be shown to a lay user?
- Could prompting before a predicted hard moment do harm, and for whom?
- Is a breathing exercise the right default, and for whom should it be avoided?
- Should the “talk to someone” suggestion appear sooner, or read differently?
- What would a small study need to measure to show real benefit?
References
- Shiffman S, Stone AA, Hufford MR. Ecological momentary assessment. Annu Rev Clin Psychol. 2008;4:1–32. PubMed
- Nahum-Shani I, Smith SN, Spring BJ, et al. Just-in-time adaptive interventions (JITAIs) in mobile health. Ann Behav Med. 2018;52(6):446–462. Article
- Zaccaro A, Piarulli A, Laurino M, et al. How breath-control can change your life: a systematic review on psycho-physiological correlates of slow breathing. Front Hum Neurosci. 2018;12:353. PubMed
- Minkel JD, Banks S, Htaik O, et al. Sleep deprivation and stressors: evidence for elevated negative affect in response to mild stressors when sleep deprived. Emotion. 2012;12(5):1015–1020. PubMed
- Hirshkowitz M, Whiton K, Albert SM, et al. National Sleep Foundation’s updated sleep duration recommendations: final report. Sleep Health. 2015;1(4):233–243. Record