Skip to content
Analyst work

Data study · October 2026

Oura and Whoop

Question

Does paying for a wearable change how people live?

Headline finding

From 2020 to 2024 the share of US adults using a wearable rose from 30.2% to 41.1%, and the share meeting the aerobic activity guideline went from 47.9% to 48.0%. Oura keeps 85% of paid members for a year. I found no figure in the coverage of its IPO filing for whether members sleep or move more.

My recommendation

Turn Advisor into a weekly plan and publish the share of members whose sleep or activity improved after 90 days.

Tools
  • CSV tables with a source URL on every row
  • JavaScript on Node for the derived rates
  • D3.js for the first dashboard
  • React and SVG for the charts on this page

The question

The client

Oura’s leadership. Oura postponed its IPO on September 29, 2026, with 5.7 million paid members and 90% revenue growth expected for fiscal 2026. Whoop is the comparison, a band sold only with a membership, with more than 2.5 million members.

The decision they face

What should Oura build and report so that tracking turns into a lasting change in how members sleep and move?

The data, 2 tables from 39 sources

  • metrics.csv74 rows

    Every figure in the study, one per row. It covers growth at Oura, Whoop and Garmin, US surveys of wearable use and activity, trial results on activity and sleep, and three Duolingo rows that tie back to my streaks study.

    Where it came from
    Press coverage of Oura’s S-1, company press releases, Garmin’s SEC 8-K filings, Pew, the HINTS and NHIS surveys, and peer-reviewed meta-analyses.
    Why I chose it
    Each figure has an ID and a source URL, so the code and a reader can go to the exact row.
    What it cannot tell us
    What happens to one person. The survey rows are population totals, and the trial rows are results I took from published papers. The Oura rows cite press coverage of the S-1, and none cites the filing directly.
  • variables.csv6 rows

    Six product features scored 0 to 2 for Oura, Whoop, Apple Watch, Fitbit and Garmin.

    Where it came from
    My own coding from product pages and press releases as of October 2026, with a link to the research behind each feature.
    Why I chose it
    It puts Oura next to Whoop feature by feature. Oura totals 7 of 12 and Whoop 11, and the options aim at the gaps.
    What it cannot tell us
    Stand as measurement. The scores are judgment, and each row links its research so a reader can disagree. The table holds no URL for the product pages I read.

The 39 sources behind metrics.csv

metrics.csv holds figures from 39 sources. Each one is listed with the rows it supplies and what they cover. 19 of them carry figures a company reported about itself.

Show the 39 sources

The process

  1. One row per figure

    Each figure went into a row of its own with a source name, a URL, the date I retrieved it and a flag for whether the company reported it. That came to 74 rows in metrics.csv from 39 source pages.

  2. Parse

    51 of the 74 rows have a comma inside a field, such as “Met aerobic physical activity guideline, age-adjusted”. Both files also use CRLF line endings. The parser reads quoted fields cell by cell and drops the carriage returns.

  3. Join

    Wearable use comes from HINTS and the activity guideline comes from NHIS. I joined the two on year. Both cover 2020 and 2024, and only HINTS has a 2022 figure.

  4. Assumptions

    Some sources gave a figure in words. Whoop’s member count was “more than 2.5M” and went in as 2.5. Oura’s share of first-time wearable buyers was “about a third” and went in as 33. Rows M01 to M03, the umbrella review’s extra steps and minutes a day, are marked approximate. The HINTS abstract gives daily use without naming its base. The table reads it as a share of all adults, so daily use stays out of the findings on this page.

  5. Check

    Oura’s nine-month revenue shows up in several press reports, so I checked it two ways. Hardware plus membership comes to $1,214.5 million against $1,210 million reported. Fiscal 2025 plus the nine-month change gives $1,420.3 million for the trailing year against $1,430 million reported. The larger gap is 0.7%, which fits rounding in the coverage.

One year before and after the join.

Before, three rows of metrics.csv (six of 12 columns shown)
metric_id,metric,value,unit,period,source_name
S06,Use a wearable device to track health,41.1,%,2024,HINTS via Pedroso et al. 2026
S08,Use a wearable every day (share of all adults),18.5,%,2024,HINTS via Pedroso et al. 2026
N02,"Met aerobic physical activity guideline, age-adjusted",48.0,%,2024,NHIS via Healthy People 2030 PA-02
After, one row per year
{
  "year": "2024",
  "use": 41.1,
  "daily": 18.5,
  "aerobic": 48
}

The first two rows are HINTS and the third is NHIS. The comma inside the third metric name is why the parser reads quoted fields.

The code for the key steps

The parser, which handles commas inside quoted fields and CRLF line endings. derive.mjs
export function parseCsv(text) {
  const records = [];
  let cells = [];
  let cell = "";
  let quoted = false;
  for (let i = 0; i < text.length; i++) {
    const ch = text[i];
    if (quoted) {
      if (ch === '"' && text[i + 1] === '"') {
        cell += '"';
        i++;
      } else if (ch === '"') quoted = false;
      else cell += ch;
    } else if (ch === '"') quoted = true;
    else if (ch === ",") {
      cells.push(cell);
      cell = "";
    } else if (ch === "\n") {
      cells.push(cell);
      records.push(cells);
      cells = [];
      cell = "";
    } else if (ch !== "\r") cell += ch;
  }
  if (cell !== "" || cells.length) records.push([...cells, cell]);
  const [cols, ...rows] = records;
  return rows.map((row) => {
    if (row.length !== cols.length) throw new Error(`Expected ${cols.length} cells, got ${row.length}: ${row.join(",")}`);
    return Object.fromEntries(cols.map((col, i) => [col, row[i]]));
  });
}
The join between the HINTS rows and the NHIS rows, by year. derive.mjs
export function trend(metrics) {
  const v = pick(metrics);
  return [
    { year: "2020", use: v("S04"), daily: v("S07"), aerobic: v("N01") },
    { year: "2022", use: v("S05"), daily: null, aerobic: null },
    { year: "2024", use: v("S06"), daily: v("S08"), aerobic: v("N02") },
  ];
}
The headline changes and shares. derive.mjs
export function headline(metrics) {
  const v = pick(metrics);
  return {
    aerobic_change_pts: Math.round((v("N02") - v("N01")) * 10) / 10,
    walk_retained_pct: (v("M09") / v("M07")) * 100,
    daily_share_2020: (v("S07") / v("S04")) * 100,
    daily_share_2024: (v("S08") / v("S06")) * 100,
    shared_adults_2024: (v("S10") / 100) * v("S06"),
    whoop_per_member: v("W02") / v("W01"),
    duo_dau_mau: (v("D01") / v("D02")) * 100,
  };
}

Dead ends

  • Google Trends

    I left Google Trends out because the research workspace could not reach it.

  • No 2022 point for activity

    The Healthy People page has no 2022 value for the aerobic guideline. The 2022 wearable figure of 36.36% comes from a separate analysis of the same HINTS survey, so the comparison runs from 2020 to 2024.

The 3 findings that change the decision

  1. Finding 1

    +0.1 points

    Change in adults meeting the aerobic guideline, 2020 to 2024

    From 2020 to 2024 the share of US adults using a wearable rose from 30.2% to 41.1%, and the share meeting the aerobic activity guideline went from 47.9% to 48.0%.

    Dataset metrics.csv
    Query derive.mjs, trend()

    Wearable use rose from 30.2% to 41.1%. The share meeting the activity guideline moved 0.1 points

    Wearable use is from HINTS. Meeting the aerobic guideline is from NHIS, age-adjusted. They are two different surveys, and population totals cannot show what happens to one user. NHIS has no 2022 value on the Healthy People page, so each line has two points.

    • Meet the aerobic guideline
    • Use a wearable

    Source: metrics.csv, derive.mjs, trend().

    Show the numbers as a table
    YearMeet the aerobic guidelineUse a wearable
    202047.930.2
    202448.041.1
  2. Finding 2

    54%

    Share of the end-of-program walking effect left at long follow-up

    In a meta-analysis of mobile health activity trials, only 54% of the walking effect was left at follow-ups more than six months after a program ended.

    Dataset metrics.csv
    Query derive.mjs, headline()

    Only 54% of the walking effect is left after a program ends

    Standardized mean difference (SMD) against control, pooled in one meta-analysis (Monninghoff et al. 2021, JMIR). Around 0.2 is small and 0.5 is medium. Few trials report long follow-up, so the later bars rest on less data.

    Walking effect against control (SMD)

    Source: metrics.csv, derive.mjs, fade().

    Show the numbers as a table
    When the effect was measuredWalking effect against control (SMD)Moderate-to-vigorous activity effect (SMD)Trials behind it
    End of program0.460.28117 mHealth trials
    Up to 6 months after0.260.2033 trials with short follow-up
    More than 6 months after0.250.198 trials with long follow-up
  3. Finding 3

    1.9x

    Activity gain with a program around the tracker over the tracker alone, 0.33 against 0.17 SMD

    Trials that put a program around the tracker raised moderate-to-vigorous activity about twice as much as the tracker alone. The tracker-alone result was too small to rule out chance.

    Dataset metrics.csv
    Query derive.mjs, design()

    A program around the tracker moved activity about twice as much as the tracker alone

    Subgroups of one review of consumer trackers (Brickwood et al. 2019, JMIR mHealth). Each group was compared with a control group, and the tracker-alone result was not statistically significant. SMD is the standardized mean difference, where around 0.2 is small and 0.5 is medium. In the review a program meant added support such as counseling or financial incentives. Trackers that came with app messages or emails counted as tracker alone.

    Effect against control (SMD)

    Source: metrics.csv, derive.mjs, design().

    Show the numbers as a table
    Outcome and designEffect against control (SMD)Source rowNote in the table
    MVPA, tracker plus program0.33M16None
    MVPA, tracker alone0.17M15Not significant (P = .05)
    Daily steps, tracker plus program0.26M14None
    Daily steps, tracker alone0.20M1328 RCTs, 3,646 adults

My recommendation

Turn Advisor into a weekly plan, and publish how many members sleep or move more after 90 days.

  1. A

    Turn Advisor into a weekly plan

    Each week Advisor picks one behavior from the member’s own sleep and Readiness data. It sets a target for that week and checks in midway.

    Evidence
    Findings 2 and 3. Trials that put a program around the tracker raised moderate-to-vigorous activity about twice as much as the tracker alone, and the walking effect fades once a program ends. On my 0-to-2 scale for coaching, Oura scores 1 and Whoop and Fitbit score 2.
    Main risk
    That result comes from subgroups of one review, and the tracker-alone result beside it was not significant. The programs in that review added support such as counseling or financial incentives. App messages counted as tracker alone, so an Advisor plan may land nearer the tracker-alone result.
  2. B

    Publish a behavior-change number

    Report the share of members whose sleep time or active minutes improved after 90 days, beside the 85% retention figure.

    Evidence
    Finding 1 and Oura’s own funnel. Oura reports 3.6 million rings sold in a year, 94% of activations turning into paid members, 85% of paid members kept for a year and a median 23 hours a day of wear. Its closest figure to an outcome is a survey where 56% of Advisor users said it helped them act.
    Main risk
    Oura would report the number itself. 19 of the 22 Oura rows in this study already come from the company, and the 56% survey has no outcome data behind it.
  • Two findings back the weekly plan. A program around the tracker moved activity about twice as much as the tracker alone, and the effect faded once the program ended, so the plan has to keep running.
  • The 90-day number tells Oura and its members whether the plan works. Today the closest thing Oura has is a survey where members said Advisor helped.
  • That number should go out with its method, since Oura would be grading itself.

Limits and next steps

What the data cannot support

  • Population totals cannot show what happens to one user, and wearable use and the activity guideline come from two different surveys.
  • No row here compares people who pay a subscription with people who do not, and none of the trial rows names Oura or Whoop.
  • The Oura rows cite press coverage of the S-1. I found no behavior outcome in that coverage, which is a weaker claim than saying the filing has none.
  • 31 of the 74 rows in metrics.csv are flagged as reported by the company they describe, including every Oura engagement figure.
  • Finding 2 compares different sets of trials. The end-of-program effect pools 117 mHealth trials, and the later points rest on 33 trials with short follow-up and 8 trials with long follow-up.
  • The feature scores are my own judgment.
  • Every row was retrieved on October 31, 2026, so nothing after that date is here.

What I would test with more access

  • With HINTS microdata and survey weights I would link wearable use and activity for the same person.
  • With Oura’s own data I would measure the number option B asks for, the share of members whose sleep time or active minutes improved after 90 days.
  • I would add Google Trends search interest, which I could not reach, and source the Oura rows from the S-1 itself.

TL;DR

  1. Studied

    Whether paying for a wearable changes how people live, read from US surveys and activity trials.

  2. Found

    From 2020 to 2024, wearable use rose from 30.2% to 41.1% and adults meeting the activity guideline went from 47.9% to 48.0%.

  3. By the numbers

    Change in adults meeting the guideline
    +0.1 points
    Of the walking effect left at long follow-up
    54%
    Paid Oura members kept for a year
    85%
  4. The catch

    Wearable use and the activity guideline come from two different surveys, so neither shows what one user did.

  5. Next

    With person-level HINTS survey data, link wearable use and activity for the same person.