PIP success rates: why none of the numbers is a statistic

Nobody has measured it, and this page is what exists instead

The search that brought you here has a plain answer: no one has published a measurement of how performance improvement plans end. Not an employer, not an HR body, not a researcher. Every number you will find is one of five things — a poll of whoever answered, somebody's recollection, a count of how many people went in, a quota, or fiction. This page goes through each of them, says where it came from and what it measures, and then says what a real answer would take. It turns none of them into a figure of its own.

That holds whichever way you are hoping this goes. A reader working to pass a plan and a reader preparing to leave on good terms are being sold the same numbers, and both would be deciding something on the strength of a thing nobody counted.

41 percent, a poll nobody who quotes it can show you

The figure travels as a Blind poll found that 41 percent of people on a PIP passed it. The earliest write-up we could find is a management-software vendor's blog post titled Do People Clear Performance Improvement Plans?, which said that Blind, the anonymous workplace app, had run a poll in which "an astonishing 41 percent of respondents said they passed a PIP", and called that a success rate. A staffing firm's blog repeated it in 2024 as a "recent poll from Blind", an executive-search consultant's newsletter repeated it in 2025 as "a survey by Blind", and a software vendor's guide rounded it to "approximately 41 percent of employees successfully complete their PIPs". None of them links the poll, and the post that started the chain no longer resolves.

What Blind itself has published on the subject is different. Blind's own 2022 article on PIPs reports a user-created poll of 2,390 verified professionals in which 36 percent said they would contest a plan they believed was unfair, and "the remaining 64% of professionals would seemingly accept their fate". That is a poll about what people would do, not about how plans end. The polls on Blind that do ask the question are created by users and show how many people took part — 187 on have you or anyone you know ever cleared a PIP, 81 on how long did you stay after clearing one, 2,956 on were you ever on a PIP — and not, to a reader on the outside, how they answered. On the first of those, one commenter writes: there must be a survey statistic somewhere.

Suppose the poll is real and the number is exactly as quoted. It would still be 41 percent of the people who use one anonymous workplace app, saw one post, chose to answer, and read passed however they chose to, in answer to a question that nobody who quotes the figure can show. That is not 41 percent of plans. Blind states a method when it runs a survey of its own; its 2025 survey on chief executives names 3,545 verified employees across 42 companies and the week they were asked. Nothing that carries the 41 percent attaches any of that to it.

70 to 90 percent, an estimate that became a range

This one can be traced step by step. A May 2025 newsletter by an executive-search consultant wrote: "it's estimated that 70–80% of employees placed on a PIP eventually exit the company." No source, and two lines away, the candid observation that "there's a reason HR departments don't publish annual PIP success rates". In August 2026 a leadership newsletter cited that post for "somewhere between 70 and 90 percent of people who are put on a Performance Improvement Plan either resign or get fired", then added that "PIPs are only successful between a measly 10 and 30 percent of the time". The second number is the first subtracted from a hundred. It is not a second measurement, and the range grew by ten points in one retelling.

The plain 90 percent has a shorter chain and a dead end. A coaching company's guide from October 2022 says "an estimated 90% of PIPs result in the employee leaving" and cites a technology news site that closed the following month. The article is offline, and nothing that repeats the figure says what the estimate rested on.

The 80 to 90 percent version is the most honest of the family, because it says what it is. An article from May 2026 quotes three Indian HR practitioners: two put the share of people who fail a plan or resign during it at 90 percent, the third at 80. These are recollections, presented as recollections, and the same article says that "most companies do not document PIP data" with any seriousness. One of the three had written years earlier about asking his own team how many people were on plans: near silence, then vague answers, then "when a single-digit number emerged after some pushing and prodding, I realised that it was never measured." That is the state of the data at the source of the estimate.

A recollection is directional. An experienced practitioner telling you that most plans they saw ended in the person leaving is telling you something true about the plans they saw, and this guide takes such accounts seriously. It is not a rate, and the moment it is written as one it starts being cited as one.

The rest of the family, and what each one measures

• "About 50 percent firings, 30 percent leave before they can be fired, 15 percent stay and do unremarkable work, 5 percent turn it around." A commenter with management experience, under an Ask A Manager post from 2023 that asks exactly this page's question. A former vice-president writing a management newsletter says plans "almost always end in you being fired" and that they have "rarely to never" seen someone survive one; a reader replies that in their experience the pass rate was about 50 percent. Three people, three memories, disagreeing by a factor of ten.
• "6 percent of employees placed on a PIP remain a year later" and "92 percent of managers admitted to using PIPs after deciding to fire someone." Both from one newsletter post in 2025. The first has no source. The second is attributed to a 2022 Harvard Business Review survey, and no such article or survey could be found on that publication's site.
• "Only 15 percent complete the plan, 60 percent are terminated, 25 percent go on stress-related medical leave." These sit in the summary of a Harvard Business Review case study from late 2025, which is where a citation or a search result lifts them from. The company in the case does not exist, its chief human resources officer does not exist, and the internal audit that "revealed" the split belongs to a scenario the publication itself describes as fictionalized, as the whole series is. A real journal, a fictional audit, and a set of figures that will be quoted without the word case.
• "A 2021 SHRM survey found 60 percent of organisations had to let people go after a PIP failed", "75 percent of employees on a plan improve", "plans with comprehensive support have 60 percent higher success rates". Aggregator and vendor pages, citing each other or nothing. The survey could not be found on SHRM's own site. One 2026 page titled Performance Improvement Plan Statistics lists more than twenty figures, and not one is about how a plan ended.

Two numbers with a method behind them, which still do not answer

Both are real. Both count how many people go into a process, and neither says how anyone came out.

HR Acuity's Employee Relations Benchmark Study is the closest thing to an industry dataset. Its 2026 edition draws on 274 American organisations with at least 1,000 employees, 8.8 million employees in all, reporting the cases their employee-relations teams opened in 2025. Performance issues ran at 40.1 per 1,000 employees in 2022, 43.6 in 2023, 39.4 in 2024 and 50.1 in 2025, a rise of 27 percent in the last year. The study does not define a performance issue, and it counts cases opened, not plans passed. What it shows is the denominator moving: whatever share of plans ended one way in 2024, the population those plans came from grew by a quarter the next year.

The other is Amazon's own paperwork. Internal documents reported by Fortune in 2024 show fewer than 2,000 employees placed into Focus, the first stage of Amazon's process, in April 2022, and "upwards of 3,300 staffers a month" by the end of that year, with entries into the formal stage doubling over the same period, ahead of 27,000 role cuts between November 2022 and March 2023. Amazon's spokesperson said that to conflate the two "is simply wrong". Either way, these are entry counts, and they make the point about denominators sharper: a rate measured in April would have been a different number by December, at the same company, under the same process.

Then there are numbers that describe a policy rather than people. A coaching firm that works with Amazon employees repeats, from Blind, a fixed share of a team expected to be on Focus and on a plan each year, calls the figures roughly right at best, and says the targets are set at director or vice-president level and change each year. A technology career coach describes managers "expected to meet PIP quotas, whereby a certain percentage of their staff must be on a PIP at any given time". A share of leavers fixed before any plan is written is not a finding about the people on the plans. It is a decision, and the Amazon page and the Google page each carry the reported target it is known for, with the caveats it needs.

The one survey with a method, and the different question it answers

There is exactly one survey we could find that asked supervisors what happened to their poor performers and published how it was done. The US Merit Systems Protection Board's 2019 research brief draws on a 2016 survey that put questions to 13,058 federal supervisors, managers and executives. Sixty-three percent had at some point supervised an employee who failed a critical performance element. Asked what happened to the most recent one, 21 percent said the employee was still there and still failing, and another 21 percent said the employee was still there and performing acceptably or better. Among supervisors whose case was four to six years back, 70 percent said the employee was no longer with the organisation, whether resigned, removed, retired or transferred; 19 percent said the employee remained and was performing well; 11 percent said the employee remained and still was not. The same brief reports that 72 percent of supervisors agreed that supervisors wait longer than they should before starting a plan, and that only 26 percent were confident they could remove an employee who had failed after completing one.

Read carefully, it answers a different question. The unit is an employee who failed a critical element, which is not the same population as people on a plan. No longer with the organisation merges resignation, removal, retirement and transfer into one box. It is what supervisors remembered, not what records showed. The setting is the American federal civil service, where an opportunity to improve is a legal step and removal is procedurally hard, which is nothing like a technology employer with a target. And the horizon is up to six years. It is the best-documented number on this page, and it cannot be read across to your plan.

What a real answer would require

Nobody has produced one, and it is worth being precise about why, because the gaps are what the circulating numbers paper over.

• A denominator. Every plan issued across a defined set of employers, including the informal stages that come before a formal plan and that some employers do not tell the person about.
• An outcome with more than two values. Passed and still employed at six months and at twelve; passed and gone within the year; left during the plan; dismissed at the end; a negotiated exit; transferred. The employer's own playbook sets the end-of-plan test as sustained improvement rather than a single review, so "passed on the last day" is not an outcome even by the other side's standard.
• A population, stated by employer, country, level and tenure, because the rules differ. A statutory improvement period, an at-will employer, a Dutch precondition to dismissal and an Estonian warning are four different processes wearing one name.
• A period, because the number of people entering plans moves with the business cycle, as the two datasets above show.
• A data holder willing to publish. Employers' HR systems hold all of this. None has published it, and at least one practitioner reports that his company did not even count. Polls of app users cannot stand in for it, tribunal databases cannot — they hold the disputes that reached judgment, a small and selected group — and a case study cannot.

Until someone does this work, every number a reader meets is a poll of whoever answered, a recollection, an incidence count, a quota, or fiction. That is not a reason to distrust the people offering them, most of whom are describing real experience honestly. It is a reason not to treat the numbers as a fact about you.

What to do with no number

The absence of a rate is not the absence of information. It moves the information from somebody else's plans to yours, which is where it was always going to have to come from. What completing a plan consists of, and how to make your side of it provable, is the question that does have an answer.

The things that can be read from a plan are on the page that asks whether a PIP means you are being fired: whether the goals have been made specific, whether the support the document promises is arriving, whether anything was raised before the plan, and the record of dates. The people who have been through it and write about it afterwards say the same thing from experience, and it holds under every ending: work the plan as if it can be passed, and prepare for the other ending as if it cannot. A written plan of your own back, questions until the goals are concrete, a dated record, and a search run in parallel and in silence serve a reader hoping to stay exactly as well as a reader planning to leave.

The read on this site is the version of this question that is about your plan rather than somebody else's. It looks for the markers and says which way they point. It gives no percentage, because there is none to give, and you should distrust anyone who gives you one — this site included, if it ever does.

None of this bears on your legal position

Fairhanded is not a law firm. Nothing on this page bears on the legal position of a plan where you live; the jurisdiction pages say what the law attaches to, and a lawyer where you live says what applies to you. What this page rests on is listed below, with the date each source was read, and the provenance of every figure above is recorded in the research file the page is generated from.