Based on anonymized first-party data from 205 people · 696 tracking sessions · ~480 hours measured.
The headline finding
Poor posture more than quadruples after 30 minutes of sitting. Across 104,373 posture readings from 205 people, only 12.7% of readings in the first 10 minutes of a long session were poor-posture. By the last 10 minutes, that figure climbed to 52.4% — a 4.1× increase. Good posture nearly halved over the same window (55.4% → 28.4%). And it turns out we hold good posture surprisingly rarely: just 34.9% of all readings. The most-flagged problem isn't the “tech neck” you might expect — it's uneven shoulders.
Key Findings at a Glance
Every figure below is computed from real SitSense posture data between February 1 and June 25, 2026 (n = 104,373 readings, 205 people). Each stat is reported as “X% of N.”
Poor posture more than quadruples after 30 minutes. In long sitting sessions, poor-posture readings rose from 12.7% in the first 10 minutes to 52.4% in the last 10 — a 4.1× jump (n = 10,714 vs 6,484 readings).
People hold good posture only 34.9% of the time. Of 104,373 readings, 34.9% were good, 27.4% fair, and 37.7% poor — making poor posture the single most common state.
Uneven shoulders — not tech neck — is the #1 flagged problem. Shoulder slope was poor in 17.0% of readings and downward neck flexion in 13.5%, while classic forward-head “tech neck” (CVA) was flagged in just 4.8%.
Sitting longer than an hour wrecks posture. Sessions under 30 minutes held good posture ~46–47% of the time; sessions over an hour dropped to 27–33% good and rose to 45–49% poor (n = 668 sessions).
Pain is the #1 reason people track posture. Among 379 surveyed users, 57.0% came for pain relief and 34.3% for appearance. The neck was the most common problem area (50.4%); average self-reported pain was 4.3/10 (n = 150).
Most people already know their posture is poor. Of the 217 users who rated their own posture, 79% called it “needs work” or “terrible” — only 5.5% said “great.”
The measurement base: 104,373 readings, not a survey. These findings come from continuous, frame-by-frame webcam measurement across ~480 hours of real seated work — not a five-point opinion poll.
How Often Do People Actually Have Good Posture?
People hold genuinely good posture far less often than they think. Across 104,373 posture readings from 205 SitSense users, only 34.9% were classified as good. Fair posture made up 27.4%, and poor posture 37.7% — meaning poor posture was the single most common state. In other words, the average person spends barely a third of their seated time in good alignment.
| Posture Classification | Readings | Percentage |
|---|---|---|
| Good | 36,448 | 34.9% |
| Fair | 28,577 | 27.4% |
| Poor | 39,348 | 37.7% |
Source: SitSense first-party posture data, Feb 1 – Jun 25, 2026. n = 104,373 readings from 205 people. Classification is the stored real-time score (the worst of the scored biomechanical metrics), measured relative to each user's calibrated baseline.
This is consistent with decades of clinical epidemiology suggesting that up to 80% of people experience back pain at some point in their lives[5]. What first-party sensor data adds is the live dimension: these aren't recalled survey answers, but continuous, frame-by-frame measurements of real seated posture during actual work.
How Quickly Does Posture Deteriorate While Sitting?
Posture degrades fast, and the slide is remarkably steady. In sessions longer than 30 minutes, poor-posture readings rose from 12.7% in the first 10 minutes to 52.4% in the last 10 minutes — a 4.1× increase, while good posture nearly halved (55.4% → 28.4%). Broken into 10-minute blocks, the deterioration is almost linear, with the sharpest break occurring right around the 30-minute mark.
Posture quality by minutes into a session
| Time Into Session | Good | Fair | Poor | Readings |
|---|---|---|---|---|
| 0–10 minutes | 55.4% | 31.9% | 12.7% | 10,714 |
| 10–20 minutes | 47.0% | 34.2% | 18.8% | 8,316 |
| 20–30 minutes | 38.7% | 32.5% | 28.8% | 7,658 |
| 30+ minutes | 28.5% | 25.4% | 46.1% | 32,696 |
Source: SitSense data. Readings grouped by elapsed time within sessions lasting 30 minutes to 4 hours (n = 59,384 readings).
The practical takeaway: roughly 30 minutes is the inflection point where seated posture starts to collapse. That aligns closely with WHO guidance to limit and regularly break up prolonged sedentary time[1], and with research recommending short movement breaks every 20–30 minutes of sitting[2]. Note this is an association across readings, not a controlled trial — but the pattern is consistent and monotonic across more than 59,000 readings.
Which Posture Problem Is Most Common?
The most-flagged problem isn't the one people expect. Uneven, collapsed shoulders (shoulder slope) was poor in 17.0% of all readings — the single most frequently failed metric — followed by downward neck flexion (looking down at the screen) at 13.5%. Classic forward-head “tech neck,” measured by the craniovertebral angle (CVA), was flagged far less often than its reputation suggests: just 4.8% of readings.
| Metric | % Flagged Poor | Avg. |Deviation| | What It Measures |
|---|---|---|---|
| Shoulder Slope | 17.0% | 12.4° | Uneven / collapsed shoulder line |
| Neck Angle | 13.5% | 7.6° | Neck flexion — looking down at the screen |
| Neck Turn | 9.1% | 3.0° | Head rotation (e.g. toward a second monitor)† |
| Trunk Angle | 7.5% | 8.2° | Torso lean / slouch relative to vertical |
| CVA Angle | 4.8% | 9.3° | Forward head position (clinical “tech neck”) |
| Head Tilt | 2.6% | 0.9° | Lateral head lean |
| Head Forward | 2.2% | 2.0° | 3D anterior head translation |
Source: SitSense data, n = 104,373 readings. “Poor” uses each metric's clinical threshold relative to the user's baseline (e.g. shoulder slope ≥ 8°, neck angle ≥ 15°, CVA ≥ 7° of forward-head deviation). †Neck turn is captured as an ergonomic signal but is not part of the overall good/fair/poor score.
The tech-neck surprise
“Tech neck” dominates posture headlines, yet in our objective data the neck-and-shoulder complex shows up mostly as uneven shoulders and looking down, rather than the textbook forward-head translation. Reduced craniovertebral angle is still clinically important — it is strongly linked to neck pain and headaches[3], and every inch of forward head load adds roughly 10 lbs of effective weight to the cervical spine[4]. But the most common day-to-day failure our sensors catch is the slow shoulder collapse you can't feel happening.
Does Sitting Longer Make Your Posture Worse?
Yes — and the effect is large. Sessions under 30 minutes held good posture about 46–47% of the time, but once a session ran past an hour, good posture fell to 27–33% and poor posture rose to 45–49%. The shortest sessions weren't just slightly better; they roughly doubled the share of good-posture time compared to the multi-hour marathons.
| Session Duration | Good Posture | Poor Posture | Sessions |
|---|---|---|---|
| 1–15 minutes | 45.9% | 22.5% | 452 |
| 15–30 minutes | 46.9% | 19.0% | 43 |
| 30–60 minutes | 42.5% | 27.9% | 47 |
| 1–2 hours | 26.7% | 48.8% | 48 |
| 2+ hours | 32.8% | 44.5% | 78 |
Source: SitSense data. Completed sessions 1 minute to 8 hours, n = 668 sessions / 99,830 readings. Buckets between 15 minutes and an hour have smaller session counts (43 and 47) and should be read as directional.
Read alongside the minute-by-minute deterioration curve above, the message is the same from two angles: shorter, broken-up sitting protects posture. The 15–30-minute session looks like a practical sweet spot — long enough to get work done, short enough to beat the 30-minute cliff.
Why Do People Start Tracking Their Posture?
Pain is the overwhelming driver, but looking better is a strong second. Among 379 users surveyed during onboarding, 57.0% said pain relief was their primary goal, 34.3% said appearance, and only 7.1% said productivity. When asked where the problem is, the neck dominated.
57.0%
Pain relief
34.3%
Appearance
7.1%
Productivity
1.6%
Other
Where does it hurt most?
| Focus Area | Users | Percentage |
|---|---|---|
| Neck | 191 | 50.4% |
| Back | 128 | 33.8% |
| Shoulders | 35 | 9.2% |
| Focus / concentration | 25 | 6.6% |
Source: SitSense onboarding survey, n = 379 respondents.
There's an interesting tension here: users overwhelmingly point to the neck as their pain center (50.4%), yet our objective tracking flags shoulder alignment most often. The two are biomechanically linked — a collapsed shoulder line drags the neck and upper back out of position — which is exactly why measuring the whole upper body, rather than a single joint, matters.
How Much Pain Do Desk Workers Report?
Among 150 users who reported a pain score during onboarding, the average was 4.3 out of 10 (median 5.0). That describes moderate, persistent discomfort — not disabling, but constant enough to motivate action. Self-reported sitting time averaged 7.3 hours per day (median 8.0), across 367 respondents.
4.3/10
Average pain score (n = 150)
7.3 hrs
Avg. daily sitting time (n = 367)
8.0 hrs
Median daily sitting time
Pain also tracked with care-seeking: users who had seen a doctor, physical therapist, or chiropractor in the previous 12 months reported higher average pain (4.9/10, n = 29) than those who hadn't (4.1/10, n = 121) — about 19% higher among those seeking care.
For context, the Global Burden of Disease Study ranks low back pain as the leading cause of disability worldwide[6]. Our survey suggests desk workers live a milder, chronic version of this — discomfort that accumulates across roughly 7–8 hours of daily sitting and is concentrated, by their own report, in the neck and upper back.
How Do People Rate Their Own Posture?
People are not in denial about their posture — if anything, they're harsh on themselves. Of the 217 users who rated their own posture, 79% called it “needs work” or “terrible,” and only 5.5% said “great.” That low self-assessment lines up with what the cameras independently measure: good posture only about a third of the time.
| Self-Rated Posture | Users | Share of Raters |
|---|---|---|
| Terrible | 51 | 23.5% |
| Needs work | 121 | 55.8% |
| Decent | 33 | 15.2% |
| Great | 12 | 5.5% |
Source: SitSense onboarding self-assessment, n = 217 users who answered the question (162 skipped it).
A separate, free-standing posture quiz tells a similar story. Across 45 people who completed it in this window, the average self-scored posture rating was just 35 out of 100 (median 34), and not one scored 70 or above. The most common self-identified pattern was forward-head posture (51.1%), followed by upper-back collapse (40.0%).
A small but telling mismatch: people self-identify forward-head as their top problem (51.1% of quiz takers), yet objective tracking flags forward-head CVA least among the neck-and-shoulder metrics (4.8%). The quiz sample is small (n = 45), so we treat it as directional — but it hints that the posture problems we feel aren't always the ones a sensor sees.
Methodology & Honesty Notes
What this is
Exploratory first-party data from SitSense users — not a representative population study. Read every figure as “among SitSense users,” not “among people in general.”
Data source & window
Anonymized posture readings captured between February 1 and June 25, 2026. We deliberately use a recent ~5-month window so the data reflects the current product and avoids stale early-cohort noise.
Sample size
104,373 posture readings, 696 tracking sessions, 205 people, and roughly 480 hours of measured sitting time. Onboarding survey: 379 respondents (150 gave a pain score, 367 gave sitting hours, 217 self-rated their posture). Standalone posture quiz: 45 completions.
How posture is measured
Browser-based computer vision using Google's MediaPipe Pose framework. SitSense scores seven upper-body angles — neck angle, CVA, trunk angle, shoulder slope, head forward, head tilt, and neck turn — many times per minute, relative to a personal baseline the user calibrates at the start of each session. All processing happens locally; no video is recorded, stored, or transmitted.
Classification
Each reading's overall good/fair/poor label is the stored real-time score, set by the worst of the scored metrics. Per-metric “poor” rates are recomputed here from the raw angles using the same clinical thresholds the app applies. Neck turn is measured but excluded from the overall score, so per-metric rates don't sum to the overall poor rate.
What we deliberately did not report
- Improvement over time. We compared good posture across users' first 3 sessions (32.4%, 199 users), sessions 4–10 (39.0%, 57 users), and 10+ sessions (34.7%, 12 users). The trend was not reliably monotonic and the 10+ cohort is too small to claim a durable training effect, so we are not publishing an improvement figure this edition.
- Time of day. We do not store each user's local timezone, so a UTC timestamp can't be interpreted as “morning vs afternoon.” We computed it and dropped it rather than mislabel it.
Known limitations
- Self-selected sample: users chose to track posture, so they may differ from the general population.
- Webcam, upper body only: the lower back, hips, and legs are not measured.
- Camera framing can affect the shoulder-line measurement, so we read the shoulder-slope finding as “uneven or collapsed shoulders” rather than a precise clinical asymmetry.
- Observational data: we describe associations (e.g. posture vs minutes sitting), never causation.
Reproducibility
Every figure here comes from one script in the SitSense repo (scripts/analytics/state-of-posture.mjs), and the aggregate dataset is published for download.
Frequently Asked Questions
How quickly does posture deteriorate while sitting?
Fast. In long sessions, poor-posture readings rose from 12.7% in the first 10 minutes to 52.4% in the last 10 — a 4.1× increase. By 10-minute block, poor posture climbed 12.7% → 18.8% → 28.8% → 46.1%, with the sharpest break near the 30-minute mark.
What percentage of the time do people sit with good posture?
Only 34.9% of 104,373 readings were good. Fair was 27.4% and poor 37.7% — so poor posture was actually the most common state, and people held good posture only about a third of their tracked sitting time.
What is the most common posture problem for desk workers?
Uneven shoulders (shoulder slope), flagged poor in 17.0% of readings, followed by downward neck flexion at 13.5%. Classic forward-head “tech neck” (CVA) was flagged in just 4.8% — far less than its reputation suggests.
Does sitting for longer make your posture worse?
Yes. Sessions under 30 minutes held good posture about 46–47% of the time; sessions over an hour dropped to 27–33% good and rose to 45–49% poor. The data points to roughly a 30-minute window before posture starts to collapse.
Is webcam posture tracking private?
Yes. SitSense runs Google's MediaPipe Pose entirely on-device in the browser. It analyzes the video frame to score posture, but no video is recorded, stored, or transmitted — only the resulting anonymized angle scores are saved.
Last updated: June 26, 2026 · Next update: September 2026
References
- World Health Organization (2020). WHO Guidelines on Physical Activity and Sedentary Behaviour. Geneva: WHO. Recommends limiting sedentary time and replacing it with activity of any intensity.
- Ekelund, U., et al. (2016). “Does physical activity attenuate, or even eliminate, the detrimental association of sitting time with mortality?” The Lancet, 388(10051), 1302–1310. Supports breaking up prolonged sitting; aligns with widely used guidance to move every 20–30 minutes.
- Yip, C.H.T., Chiu, T.T.W., & Poon, A.T.K. (2008). “The relationship between head posture and severity and disability of patients with neck pain.” Manual Therapy, 13(2), 148–154. A reduced craniovertebral angle is significantly associated with neck pain severity.
- Hansraj, K.K. (2014). “Assessment of stresses in the cervical spine caused by posture and position of the head.” Surgical Technology International, 25, 277–279. Forward head posture sharply increases effective load on the cervical spine.
- Andersson, G.B.J. (1999). “Epidemiological features of chronic low-back pain.” The Lancet, 354(9178), 581–585. Up to ~80% of people experience back pain at some point in life.
- GBD 2019 Diseases and Injuries Collaborators (2020). “Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019.” The Lancet, 396(10258), 1204–1222. Low back pain is the leading cause of disability worldwide.
Download the Data
The anonymized aggregate statistics behind this report are published for anyone to cite or reuse (aggregate rows only — no per-user records, zero PII):
- state-of-posture-2026.csv — tidy spreadsheet format
- state-of-posture-2026.json — structured / machine-readable
Licensed CC BY 4.0. Please attribute as: “Source: SitSense 2026 State of Posture Report (https://www.sitsense.app/blog/state-of-posture).”
Related reports & guides
Same measurements
Every number above came from a webcam like yours.
104,373 readings from 205 people at their own desks, taken by the same model that would read you.
Free to start. Nothing to install.
Read next
State of Remote Work Posture 2026: What 76,000+ Data Points Reveal
We analyzed 76,000+ real posture readings from 205 remote workers. Posture deteriorates 3.2× after 30 minutes. Forward head posture (tech neck) is the #1 problem. Full data, methodology, and findings.
Remote Work Is Destroying Your Posture: What the Data Revealed
We analyzed hundreds of hours of real posture data from remote workers using webcam AI. Here are the biggest findings about when, why, and how posture breaks down during the workday.
What 540+ Hours of Posture Data Revealed
Insights from analyzing over 540 hours of real posture tracking data. Common patterns, peak slouching times, and improvement strategies.
Is Your Posture Affecting Your Appearance? Here's What the Data Says
Research-backed insights on how posture impacts your appearance, confidence, and first impressions. Learn what 540+ hours of posture data revealed.
How to Fix Posture: The Complete Guide for Computer Users
Science-backed exercises, stretches, and desk setup tips to fix your posture. Includes video tutorials and a 10-minute daily routine for desk workers.