A sticky-note character at a whiteboard of small bar charts, each chart tagged with a paper source label

Agile and retrospective statistics: 178 figures, every one linked to its primary source

178 agile, scrum, retrospective, team health and meeting statistics. Each one is a single sentence carrying its data year, its sample and a link to the primary source it came from.

83% of organizations run retrospectives, second only to the daily standup at 87%, and the teams that strongly agree they hold them measure 24% more Responsiveness and 42% higher Quality than the teams that do not. Those two figures come from different studies a decade apart, and both are quoted constantly without the conditions that make them true. This page puts the conditions back.

It is written for the person who needs one number for a blog post, a paper, a deck or a model, and needs it to survive a fact-check. Every statistic below is a single sentence carrying its own data year and its own source, linked to the document that published it rather than to another roundup. Where a famous number turned out to be untraceable, we say so instead of passing it on.

Last updated: September 2026. 178 statistics, each fetched from its primary source on 21 September 2026.

In this guide

  1. Key statistics
  2. Sprint retrospectives and continuous improvement
  3. Team health, engagement and psychological safety
  4. Meetings and the cost of collaboration
  5. Remote, hybrid and async teams
  6. Agile adoption and scale
  7. Agile outcomes and failure rates
  8. Scrum
  9. Kanban and flow
  10. Developer productivity, DevOps and DORA
  11. Agile beyond software
  12. Agile tooling landscape
  13. Scrum master and agile coach careers
  14. What we have seen
  15. How we compiled this
  16. Cite this
  17. Keep reading
  18. Frequently asked questions

Key statistics

The twelve figures below are the ones most likely to be useful on their own. Each is a complete sentence, so it can be lifted as it stands, and each links to the section where the rest of that study’s numbers sit.

Sprint retrospectives and continuous improvement

Retrospectives are one of the few agile practices with both broad adoption data and a large measured performance study behind them. The adoption numbers come from Digital.ai’s annual State of Agile survey. The performance numbers come from a single dataset built by CA Technologies out of Rally Software’s product telemetry. Both are heavily quoted and both are heavily mangled, so this section states each one with its condition attached.

The most-quoted retrospective statistic, quoted correctly

The line that circulates as “teams that have regular sprint retrospectives have 24% more responsiveness and 42% higher quality” is real, but the source does not say “regular”. It says this:

Teams that strongly agree that they have sprint retrospectives have 24% more Responsiveness and have 42% higher Quality with less variability.

The condition is self-reported strong agreement on a 1 to 5 scale, which is a narrower and more demanding claim than holding retrospectives on a cadence. The comparison set is every team that sits below that on the same scale.

Measured performance differences alongside sprint retrospectives: 24 percent more Responsiveness and 42 percent higher Quality for teams that strongly agree they hold them, and 20 percent higher balanced performance where retrospectives are effective Teams that strongly agree they hold sprint retrospectives Teams with effective retrospectives Responsiveness Quality Balanced performance 24% 42% 20% 0 10 20 30 40 50 Higher than the comparison group (%) Source: CA Technologies, The Impact of Agile. Quantified., 2014 to 2015 data, published by Broadcom.
The condition is what makes these figures true. Responsiveness and Quality compare teams that strongly agree they hold sprint retrospectives against every team below that on the same 1 to 5 scale; balanced performance compares teams with effective retrospectives against teams that hold none. Source: CA Technologies, The Impact of Agile. Quantified., 2014 to 2015 data, published by Broadcom.

How many teams run retrospectives

Adoption of the four Scrum events: daily standups 87 percent, retrospectives 83 percent, sprint planning 83 percent and sprint reviews 81 percent of organizations Daily standup Sprint planning Sprint review Retrospective 87% 83% 81% 83% 0 25 50 75 100 Organizations running the practice (%) Source: Digital.ai, 15th State of Agile report, 2021, 1,382 complete responses.
Adoption of the four Scrum events among the organizations surveyed. The retrospective sits second, level with sprint planning. Source: Digital.ai, 15th State of Agile report, 2021.

Retrospectives as a source of improvement data

AI in the improvement loop

Where improvement work stalls

Team health, engagement and psychological safety

Engagement is the most heavily sampled number on this page and the one most often cited at the wrong edition. Gallup’s current figure is 20%, from 2025 data published in the 2026 edition; the 21% and $438 billion figures still being republished belong to the edition before it. The Q12 meta-analysis is a separate piece of work again, and it is the one to cite when someone asks what team health is actually worth.

Engagement and wellbeing worldwide

Employee engagement worldwide: 20 percent engaged, 64 percent not engaged and 16 percent actively disengaged 20% engaged Engaged Not engaged Actively disengaged 20% 64% 16% Source: Gallup, State of the Global Workplace 2026, 263,810 respondents surveyed across 2025.
Engagement worldwide splits three ways, and the engaged share is the smallest of them. Source: Gallup, State of the Global Workplace 2026, 2025 data.

What engagement is worth

Gallup’s Q12 meta-analysis is the largest evidence base on this page: 736 studies across 347 organizations, 183,806 business units, 3,354,784 employees, 53 industries and 90 countries. Every figure below is a median difference between top-quartile and bottom-quartile business units, not a before-and-after.

Median differences between top-quartile and bottom-quartile business units on engagement: 78 percent lower absenteeism, a 70 percent wellbeing gap, 32 percent fewer quality defects, 23 percent more profitability and 18 percent higher sales productivity Lower absenteeism Wellbeing gap Fewer quality defects More profitable Sales productivity 78% 70% 32% 23% 18% 0 25 50 75 100 Median difference between quartiles (%) Source: Gallup, Q12 meta-analysis, 11th edition, 2024, 183,806 business units and 3,354,784 employees.
Each bar is a median difference between top-quartile and bottom-quartile business units, not a before-and-after measurement of the same teams. Source: Gallup, Q12 meta-analysis, 11th edition, 2024.

Psychological safety and how meetings are run

These three figures come from one case study inside a wider survey of nearly 200 senior executives, so treat them as a single intervention rather than a population estimate. They are the closest thing in the literature to a measured link between changing how a team meets and how safe people feel speaking up.

Culture and performance

Meetings and the cost of collaboration

If you are building a case for changing how a team meets, this is the densest section on the page. The meeting-cost study most often credited to Otter.ai was designed and analyzed by Steven G. Rogelberg of UNC Charlotte, and attributing it to the researcher is both more accurate and more useful in a footnote. The Microsoft figures come from product telemetry rather than a survey, which makes them a different kind of evidence: very large, very specific, and limited to organizations running Microsoft 365.

How many meetings, and how many are unnecessary

What unnecessary meetings cost

Why people go anyway

Employees accept 83 percent of meeting invitations, want to decline 31 percent and actually decline only 14 percent Accept the invitation Want to decline Actually decline 83% 31% 14% 0 25 50 75 100 Share of meeting invitations (%) Source: Rogelberg (UNC Charlotte) with Otter.ai, The Cost of Unnecessary Meeting Attendance, 2022, 632 US employees.
The gap between wanting to decline a meeting and actually declining it. Source: Rogelberg with Otter.ai, The Cost of Unnecessary Meeting Attendance, 2022.

What executives say about their own meetings

When meetings happen, and what they do to focus

Remote, hybrid and async teams

Two of the most-cited remote-work datasets now point in different directions, and the reason is sampling rather than disagreement. Owl Labs surveys US workers across all working patterns, and its in-office share has sat at around six in ten since 2023. Buffer surveys people who already work remotely, so its numbers describe life inside that group. Read them as answers to different questions.

Where people actually work

Where US full-time workers work: 63 percent fully in the office, 28 percent hybrid and 9 percent fully remote 63% in the office Fully in the office Hybrid Fully remote 63% 28% 9% Source: Owl Labs, State of Hybrid Work 2025, 2,000 US full-time workers, fielded July 2025.
Where US full-time workers work, sampled across all working patterns rather than among remote workers only. Source: Owl Labs, State of Hybrid Work 2025.

What hybrid meetings cost

Stress, retention and quiet quitting

What remote workers report

Agile adoption and scale

Adoption figures and depth figures tell opposite stories, and quoting one without the other is the most common mistake in this topic. Nearly everyone reports using agile somewhere; barely anyone reports it being embedded everywhere. The 2025 edition of the State of Agile survey is also much smaller and much more enterprise-weighted than the 2021 one, so the two are not a clean time series.

How widely agile is used

Where it has spread beyond engineering

How deep the adoption goes

How deep agile adoption goes: 13 percent deeply embedded, 25 percent in some departments inconsistently, 22 percent transitioning and 20 percent at team level only 13% 25% 22% 20% Deeply embedded Some departments, inconsistently Transitioning from traditional Team level only 0 25 50 75 100 Share of respondents (%) Source: Digital.ai, 18th State of Agile report, 2025. The four named depth levels do not add up to 100%.
The four depth levels the report names do not add up to 100%, because the survey offers further options beyond them. Source: Digital.ai, 18th State of Agile report, 2025.

Agile outcomes and failure rates

There is no credible primary figure for “agile succeeds x% of the time against waterfall’s y%”, which is why this section does not carry one. What there is instead: what organizations say they wanted from agile, what they report getting, and where it stops working. The satisfaction drop between 2022 and 2023 is the sharpest single movement in the whole State of Agile series.

Why organizations adopt agile

What they report getting

Satisfaction, and the pressure to prove it

Satisfaction with agile fell from 71 percent in 2022 to 59 percent in 2023 2022 2023 71% 59% 0 25 50 75 100 Satisfied with agile (%) Source: Digital.ai, 17th State of Agile report, 2023.
The sharpest single movement in the State of Agile series. Source: Digital.ai, 17th State of Agile report, 2023.

What blocks it

Scrum

Scrum’s adoption figures come from surveys; its performance figures come from the same Rally dataset behind the retrospective numbers, which is the only large source here that measured teams rather than asking them. That makes the estimation and iteration-length findings unusually concrete, and unusually old: the data is from 2014 and 2015.

How many teams run Scrum

The ceremonies

Sprint length and estimation, measured

Iteration length across agile teams: 59.1 percent run two-week iterations, 23.4 percent three weeks, 9.8 percent four weeks and 6.2 percent one week Three weeks Four weeks One week Two weeks 59.1% 23.4% 9.8% 6.2% 0 25 50 75 100 Share of agile teams (%) Source: The Impact of Agile. Quantified. (Broadcom), 160,000+ projects.
Iteration length measured from product telemetry rather than self-reported, which is rare in this subject. Source: The Impact of Agile. Quantified., CA Technologies, 2014 to 2015 data, published by Broadcom.

Kanban and flow

Kanban is the quiet winner of the State of Agile series: board use went from a rounding error to a majority practice over the survey’s lifetime. The flow findings underneath come from the Rally dataset again, and the most useful one is the least quoted, because it cuts against the usual advice: driving work in process to an extreme low costs productivity rather than adding it.

Kanban adoption

What flow data shows about work in process and team shape

Developer productivity, DevOps and DORA

Adoption of AI tooling is now close to universal among developers, and trust in it is not. That gap is the most interesting thing in this year’s research, and both DORA and Stack Overflow found it independently with very different samples. DORA’s 2025 edition also adds a cluster analysis that sorts teams into seven profiles.

AI in the development workflow

Developer trust in AI accuracy: 46 percent actively distrust it, 33 percent trust it and only 3.1 percent highly trust it Actively distrust accuracy Trust accuracy Highly trust accuracy 46% 33% 3.1% 0 25 50 75 100 Share of developers (%) Source: Stack Overflow, 2025 Developer Survey, 49,000+ developers across 177 countries.
Use of AI tools is close to universal among developers; trust in their accuracy is not. Source: Stack Overflow, 2025 Developer Survey.

Platforms, priorities and performance

Who answered, and how they feel

Agile beyond software

Agile marketing is the best-documented non-software adoption, and its numbers are strikingly more positive than software’s, which is worth a caveat rather than a celebration: the survey reaches people who already chose agile marketing. The more sober finding sits underneath it, and it matches the enterprise picture in software: leaders are largely not involved in shaping how agile is practiced.

Agile marketing

Leadership and the shape of the agile role

Agile tooling landscape

Two things stand out in the tooling data, and neither is about which vendor wins. Spreadsheets are still used by more organizations than dedicated agile tools, and the analytics layer most vendors sell against is barely in place: only a fraction of organizations move agile data into BI at any meaningful scale. Note that the tool-recommendation figures come from a survey run by a tool vendor.

What agile teams run on

Data and AI in the tool stack

Scrum master and agile coach careers

Career and community figures are the hardest to source cleanly, because the bodies that hold the data also sell training. Scrum Alliance’s membership and course numbers are its own operational data, which is the right kind of figure to quote. Its salary and skills claims are quoted from a separate 2023 study, Skills in the New World of Work, which we have not read, so neither is published here second-hand.

The Scrum Alliance community

Who does agile work

What we have seen

Everything above is other people’s research. This short section is the one place where our own data appears, and we keep it deliberately narrow: three figures we have already published in full, with their method and limits, in the Follow-Through Index.

  • About 73% of retrospective action items are eventually completed, across hundreds of thousands of action items created in real retrospectives by teams using TeamRetro.
  • About 90% are completed when an action carries a named owner and a due date.
  • About 55% are completed by teams that retrospect only occasionally, against roughly three in four for teams with a regular cadence.

Those are a ceiling rather than an industry average, because they describe teams who already run their retrospectives in a dedicated tool. The methodology, the exclusions and the five things we would not claim from the data are all on that page.

Three patterns we can see but will not put a number on, also from the Follow-Through Index:

  • Follow-through does not track how a team scores itself. We went looking for a correlation between completion and teams’ own health-check scores and found essentially none, which suggests following through is a discipline rather than a mood.
  • Retrospectives outrun their own actions, so most of what looks abandoned at the next retro is still moving rather than failed.
  • Some of the actions that are never completed were never the team’s to complete, and read better as escalations filed in the wrong place than as follow-through failures.

How we compiled this

This page exists because the same numbers keep circulating without their conditions, their editions or a working link. The rules below are what we used to decide what to publish, and what to leave out.

What was included. Every statistic here was checked against the publisher’s own document on 21 September 2026 rather than taken from a roundup, with two exceptions we want to be open about. Digital.ai’s 18th State of Agile report is only available behind a form, so its figures were read from a verbatim copy of the PDF and are linked to Digital.ai’s official report page. Harvard Business Review’s “Stop the Meeting Madness” sits behind a paywall, so its figures were checked against a reprint of the article and are linked to the original. Where a survey runs annually, we used the most recent edition we could actually read, and we name that edition in the sentence rather than just the organization. Where a figure is one we would rather not have published, such as falling agile satisfaction or AI’s negative correlation with delivery stability, it is here anyway. We do not cite research published by competing retrospective tools. Most of the organizations we do cite sell something in this space too, so read every vendor survey, including Digital.ai’s, as a vendor survey.

What was left out, and why.

  • The Standish Group CHAOS comparison of agile against waterfall. The “42% against 13%” figure is the most-quoted statistic in this subject and we could not trace it to a document. Standish sells CHAOS rather than publishing it, and every trace we followed ended at a blog post quoting another blog post. It is excluded rather than repeated.
  • Google’s Project Aristotle. The re:Work pages that held it now return a page shell with no readable guide text, so we could not verify the 180-teams and 250-attributes framing against a live primary source. It is omitted here. The DORA culture findings and the meeting-redesign study above cover similar ground with sources you can open.
  • Amy Edmondson’s 1999 psychological safety study. The publisher hosts returned errors rather than the paper, so it did not clear the same bar, despite being the foundational work in that area.
  • Gated or oversized reports, including Atlassian’s State of Teams multipliers, McKinsey’s Developer Velocity Index, LinearB’s benchmark metrics and Kanban University’s WIP-limit claim. Where a report’s headline numbers sit behind a form, only the figures its public pages state are used.
  • Market-size figures for agile tooling. The available estimates come from press-release-grade aggregators that disagree with each other by roughly a factor of two.
  • Two Scrum Alliance career figures, on salary premiums and expected agile skills, which its annual report quotes from a separate 2023 study, Skills in the New World of Work, that we have not read.
  • One vendor-favorable result in the Impact of Agile report, showing performance gains for teams using the vendor’s own tool. It is real, and it is also a vendor measuring its customers, so it is not on the page.

Corrections we made along the way. The 87% standup and 83% retrospective figures are routinely credited to the 17th State of Agile report; they are from the 15th, and the 17th dropped that question entirely. The retrospective performance figures are CA Technologies and Broadcom’s, not any aggregator’s, and the source says “strongly agree”, not “regular”. The meeting-cost study is Steven G. Rogelberg’s research in partnership with Otter.ai rather than Otter.ai’s own. Gallup’s current engagement figure is 20% from the 2026 edition, not the 21% and $438 billion that most pages still quote from the edition before. And Microsoft’s “275 interruptions” and its capacity-gap findings come from two different artifacts that share a year rather than a dataset.

Link checking. Every source URL on this page is re-checked by the external link audit that runs over this site, so a source that goes dark gets found and replaced rather than sitting here rotting. If you spot one before we do, tell us.

Changelog

  • September 2026: first published, 178 statistics.

Cite this

83% of organizations run retrospectives, second only to the daily standup at 87% (Digital.ai, 15th State of Agile report, 2021). Teams that strongly agree they hold sprint retrospectives show 24% more Responsiveness and 42% higher Quality with less variability, across 160,000+ projects and 50,000 agile teams (CA Technologies, The Impact of Agile. Quantified., 2014 to 2015 data, published by Broadcom). 20% of employees worldwide are engaged at work (Gallup, State of the Global Workplace 2026). Employees average 17.7 meetings a week and say about 30% of them could have been skipped (Rogelberg with Otter.ai, The Cost of Unnecessary Meeting Attendance, 2022). — Agile and retrospective statistics, TeamRetro (2026)

If you reuse a figure from this page, cite the primary source named beside it, and link this page as the compilation if it is what helped you find it. Using it in research or a talk? We would love a link back.

Keep reading

Frequently asked questions

What percentage of teams run retrospectives?

83% of organizations run retrospectives, in Digital.ai’s 15th State of Agile report, which drew on 1,382 completed responses collected between February and April 2021. That makes the retrospective the second most widely used agile practice after the daily standup at 87%, level with sprint planning at 83% and ahead of the sprint review at 81%. Retrospective adoption stayed close to constant across 15 annual editions of that survey while the methodologies around it churned. One caution when you cite it: these figures are routinely attributed to the 17th edition, which dropped the practices question entirely.

Do retrospectives improve quality?

The one large measured study says yes, with a condition most write-ups drop. Teams that strongly agreed they held sprint retrospectives showed 24% more Responsiveness and 42% higher Quality with less variability, in the report The Impact of Agile. Quantified., produced by CA Technologies from Rally Software data and now published by Broadcom. The dataset covers more than 160,000 projects and 50,000 agile teams. The comparison is between teams that strongly agree on a 1 to 5 scale and the rest, not between teams that hold retrospectives regularly and teams that do not. The same report states that effective retrospectives result in 20% higher balanced performance, and warns in its own words that correlation does not necessarily mean causation.

How much time do meetings cost?

Employees average 17.7 meetings a week, totaling 18 hours, and say about 30% of them could be skipped if they were kept in the loop another way. That comes from a 2022 study of 632 US employees designed and analyzed by Steven G. Rogelberg of UNC Charlotte with Otter.ai. The unnecessary share works out at about $25,000 per employee per year, roughly 31% of an $80,000 annual meeting bill per professional, or nearly $2.5 million a year for a 100-person company. Executives are not happier about it: only 17% told Harvard Business Review their meetings were a generally productive use of group and individual time.

How many organizations use agile?

71% of practitioners use agile in their software development lifecycle, in Digital.ai’s 17th State of Agile report from 2023, and 86% of organizations reported agile adoption in software development in the 15th edition in 2021. Depth is the weaker number. In the 18th edition, published in 2025, only 13% said agile is deeply embedded across business and technology functions, while 25% practice it in some departments inconsistently and 20% keep it at team level only.

Is agile more successful than waterfall?

We could not verify a primary source for that comparison, so this page does not publish one. The figure usually quoted, that agile projects succeed about 42% of the time against 13% for waterfall, comes from the Standish Group’s CHAOS research, which is sold rather than published. Every trace we followed ended in a blog post quoting another blog post rather than a document from Standish. We left it out deliberately. What is measurable is narrower and less flattering: satisfaction with agile fell to 59% in 2023 from 71% the year before, and 47% name general organizational resistance to change as the top barrier to adoption, both in Digital.ai’s 17th State of Agile report.

What share of developers use AI tools?

84% of developers use AI tools and 51% of professional developers use them daily, in Stack Overflow’s 2025 Developer Survey of more than 49,000 developers across 177 countries. DORA’s 2025 report puts adoption among technology professionals at 90%, with more than 80% of AI users saying it increased their productivity. Trust runs the other way: 46% of developers actively distrust AI accuracy against 33% who trust it, 66% name almost-right AI answers as their top frustration, and DORA finds AI adoption still correlates negatively with delivery stability even though throughput now moves the other way.

What percentage of employees are engaged at work?

20% of employees worldwide are engaged at work, with 64% not engaged and 16% actively disengaged, in Gallup’s State of the Global Workplace 2026, which covers 263,810 respondents surveyed across 2025. Manager engagement sits at 22%, down nine points since 2022, and Gallup puts the cost of low engagement at about $10 trillion a year, roughly 9% of global GDP. Quote the edition rather than just Gallup: the widely republished 21% engaged and $438 billion figures belong to the previous edition.

How many teams use Scrum, and how many use Kanban?

63% of agile users run Scrum, still the most popular team-level methodology, in Digital.ai’s 17th State of Agile report from 2023. The 15th edition put it at 66%, with another 15% on Scrum derivatives such as ScrumBan and Scrum with XP. In that same edition 61% of organizations used Kanban as an agile technique, up from 6% in the first State of Agile survey, which is the biggest long-run practice shift the survey records, and 77% used Kanban boards. At enterprise scale SAFe leads at 26%, ahead of Scrum@Scale and Scrum of Scrums at 19%.

Can I cite these statistics?

Yes. Every figure on this page names the organization that published it, the edition, the data year and, where the source states it, the sample. Cite that primary source, and link this page as the compilation if it is what helped you find the figure. Every source link was fetched and confirmed live on 21 September 2026, and they are re-checked by the external link audit that runs over this site.