Employee Retention Strategies: What the Evidence Supports

The employee retention strategies with the best evidence behind them work on four things at once: rewards beyond the paycheck, the quality of supervision, how well a person fits the job, and the ties that make leaving costly. Raising pay, the move most managers reach for first, is rarely the strongest lever on its own. In what its authors called the most comprehensive meta-analysis of voluntary turnover to date, published in 2018, pay was more weakly linked to who quit than rewards beyond pay or job embeddedness.¹

In practice, start with a diagnosis rather than a program. Find out which of your people are leaving, not just how many; ask the people who stay as well as those who go; write down what happened before each resignation; and try one change on one team before rolling it out. The five-step plan near the end of this page turns those moves into a checklist.

Engagement surveys are often read as an early warning of turnover, but what an engagement score measures is a separate question. This page sticks to who leaves, who stays and what a manager can change.

Which employee retention strategies have the strongest evidence?

The best evidence on what keeps employees comes from meta-analyses of studies that measured conditions at work first and then tracked who later quit. The broadest, by Alex Rubenstein and colleagues in 2018, found larger links to staying for rewards beyond pay, fit, leadership and job embeddedness than for pay itself, with the gap clearest for rewards beyond pay and embeddedness. Every study it included measured its predictors before anyone left, so the answers were given before anyone knew who would go.¹

These reviews rank predictors, not programs. A meta-analysismeta-analysis: A study that combines the results of earlier studies on the same question into one overall estimate. Pooling makes the estimate more precise, but it cannot repair the studies it pools: a meta-analysis of surveys is still survey evidence.Full entry in the glossary of this kind asks, in effect: among employees surveyed in January, did those who rated their supervisor poorly leave more often by the following year? It cannot say that fixing the supervisor would have kept them, only that the two go together.

The study

Moderate evidence

Who quits, across four decades of studies (Rubenstein and colleagues, 2018)

The authors pooled correlations between dozens of predictors and later voluntary quitting. Rewards beyond pay, such as benefits, growth opportunities and training, were linked to quitting at −.28, job embeddedness at −.26 and leadership at −.24, against −.17 for pay; the confidence intervals for rewards beyond pay and for embeddedness did not overlap with pay’s. Thoughts and intentions of quitting (.56) and job search (.40) were the strongest predictors of all.¹

Apart from intentions to leave and job search, signs of a decision already forming, every link in the review is small to moderate and none dominates, so no single lever is likely to explain most departures on its own. The authors flag their own caveats: most samples came from white-collar or hospital work, and relationships varied widely from one setting to the next.¹

Bar chart with 4 bars; the largest is Rewards beyond pay at .28.

• Rewards beyond pay.28
• Job embeddedness.26
• Leadership.24
• Pay.17

The pattern matches the earlier benchmark. Rodger Griffeth, Peter Hom and Stefan Gaertner’s 2000 meta-analysis found the best predictors were attitudes and steps close to the exit, such as satisfaction, commitment and job search, with smaller links for the work environment, including leadership, co-worker cohesion and promotion chances.²

Treat retention as a portfolio and pick the two or three levers your own data point to, rather than the one that is easiest to announce.

Pay matters, but less than most managers expect

Pay predicts quitting, but modestly. A 2010 evidence review by David Allen, Phillip Bryant and James Vardaman, partly supported by the SHRM Foundation, lists the belief that people quit because of pay as a common misconception: pay level and pay satisfaction are relatively weak predictors of individual decisions to leave.³

Part of the reason may be how pay is studied. Griffeth and colleagues noted that turnover studies mostly left out benefits and other forms of pay, and argued that this, together with the narrow spread of pay within samples, understated how much money deters quits.²

Ask people who have already quit, and pay comes first. In a February 2022 Pew Research Center survey of non-retired US adults, 63 percent of those who had left a job by choice in 2021 said low pay was a reason, the same share cited no chance to advance, and 57 percent said they had felt disrespected at work.⁴

The two findings fit once you notice who is asked: a survey of leavers cannot show whether stayers had the same complaint. Allen and colleagues give a hypothetical case in which three quarters of leavers tell exit interviewers they are unhappy with pay, and point out that most of the people who stayed may be just as unhappy. If exit interviews and focus groups show that pay is a particular grievance in one hard-to-fill role, they suggest reworking pay for that role, which may be more cost-effective than a company-wide rise.³

So before approving an across-the-board raise, compare leavers’ reasons with a survey of people who stayed. Matching a rival’s offer at the door is a different tactic, with its own evidence, and deserves to be judged separately.

Supervisors and the climate they create

Supervision and workplace climate are among the stronger predictors of staying that a manager can change. Rubenstein and colleagues write that their results back the saying that employees quit bosses, not jobs, and that turnover can stem at least as much from toxic climates or from feeling unsupported by the organization. They suggest leadership development may be especially valuable.¹

The 2010 review is more concrete: train supervisors in how to lead and build relationships, evaluate them partly on retention, and identify and remove abusive supervisors.³

Picture two teams in one company with the same pay bands. In one, the manager holds regular check-ins and spreads the interesting assignments around; in the other, favorites get the good work and problems go unheard. The research pattern predicts more departures from the second team, even though the pay data look identical. Well-run one-on-ones with each team member are one place where the better relationship gets built.

Company data from 2021 point the same way. Donald Sull, Charles Sull and Ben Zweig compared more than 1.4 million Glassdoor reviews of large US companies with attrition from April to September 2021. A toxic culture, marked by disrespect, unethical behavior and a failure to promote diversity, equity and inclusion, predicted a company’s attrition relative to its industry about 10 times as strongly as what employees wrote about their pay. Read that ratio as a pointer, not a measurement. The analysis appeared in a management magazine rather than a peer-reviewed journal, and the authors’ own companies supplied the tools: the Sulls co-founded CultureX, whose platform scored the reviews, and Zweig runs Revelio Labs, which estimated the attrition. The attrition measure also counts every departure, including layoffs.⁵

Job embeddedness: the forces that keep people in place

Job embeddedness is the pull of everything a person would leave behind: links to other people, fit with the job and the place they live, and what they would sacrifice by going. Terence Mitchell and colleagues introduced the idea in 2001, surveying staff at a regional grocery chain and a community hospital in 1998, when local unemployment was very low, and then tracking who left. Embeddedness predicted both intent to leave and actual quitting, over and above job satisfactionjob satisfaction: How positively a person judges and feels about their job, as a whole or in parts such as pay, colleagues or the work itself. It is usually measured by questionnaire and is related to engagement, but it is not the same thing.Full entry in the glossary, commitment, job alternatives and job search. To build it, the authors suggest making people mentors and putting them on long-term projects, which adds links, and offering perks tied to length of service, which raises the sacrifice.⁶ The 2018 meta-analysis later placed embeddedness among the stronger predictors of quitting.¹

1. Links: connections to other people and groups: co-workers, work groups, mentors, friends and relatives
2. Fit: how compatible a person feels with their job, their organization and their community
3. Sacrifice: what leaving would give up: rewards tied to tenure, a good work environment, promotion prospects, standing in the community

Top row: on the job (the organization). Bottom row: off the job (the community).

The idea turns the usual question around. Satisfaction asks whether people like the job; embeddedness asks what leaving would pull apart. A nurse who mentors two new hires, sings in a local choir and has benefits that grow with each year of service has several reasons to stay that are unrelated to how she felt about last week’s schedule.

For building embeddedness, the 2010 review lists teamwork, mentoring, employee referrals, realistic information during hiring, clear talk about the organization’s values, and career paths that do not require moving house.³

Regular, structured conversations with current staff about why they stay are another way to learn what holds people; they deserve a guide of their own.

Many quits begin with a single event

Not every resignation follows months of growing unhappiness. Allen and colleagues’ review, drawing on the unfolding model of turnover, reports that job dissatisfaction may be the driving force in fewer than half of individual decisions to leave. Many exits begin with a shock: an event that makes someone think about quitting.³

Shocks can be good, neutral or bad: an unsolicited job offer, a merger announcement, being passed over for promotion, a spouse’s new job in another city. Someone who quits right after being passed over, without even searching first, makes an impulsive exit that no satisfaction survey would have flagged. For these exits the review suggests realistic job previews and clear communication, training managers to give negative feedback well and to prevent harassment and unfair treatment, and support routes such as grievance procedures, flexible work and employee assistance programsemployee assistance program: A confidential service an employer pays for that gives staff free assessment, short-term counseling and referral for personal or work problems. It is a route to help, not a replacement for a doctor or for fixing problems in the work itself.Full entry in the glossary.³

Entry is a risky stretch. The same review links realistic job previews, employee referrals and selection for fit to lower turnover, and onboarding that connects new hires with experienced colleagues to better retention in the critical first year, which is why the first months are where a structured onboarding plan earns its keep.³

A simple habit follows. For each resignation you learn about, write down the event that preceded it, whether a reorganization, a missed promotion or a new manager. Patterns in those events are often easier to act on than a satisfaction score.

Building a retention plan from your own data

Allen and colleagues’ framework puts diagnosis before any program: analyze who is leaving and at what cost, read those numbers in context, and collect data on causes from several sources; only then choose targeted or organization-wide fixes. The authors stress that some turnover does little harm, or even helps, such as the exit of poor performers; it becomes a problem when the wrong people leave or the rate grows costly.³

1. 1Analyzewho leaves, how often, at what cost
2. 2Read in contextyour industry, your plans, the job market
3. 3Diagnose causesleavers and stayers, several sources
4. 4Acttargeted or organization-wide, then check

1. Find out who is leaving

Break departures down by team, role, tenure and performance, and mark which leavers you would have fought to keep. A rising total can hide a healthy pattern, and a flat one can hide the loss of your strongest people. Put a rough price on the departures that hurt, using a step-by-step estimate of turnover costs, so any fix can be weighed against the loss.

2. Compare your rate with the market

In the US, quits averaged 2.0 percent of employment a month in 2025, down from 2.8 percent in 2022, in the Bureau of Labor Statistics’ JOLTS survey, with rates in accommodation and food services roughly five times those in government.⁷ Outside the US, look for comparable figures from your own national statistics office or industry body. Compare yourself with your own industry rather than the national figure, and read a change against the market: a rise that every employer in your sector shares calls for a different response from one that is yours alone.

3. Ask the people who stayed as well

Combine exit interviews with surveys, focus groups or conversations with current staff. Intentions to leave predict quitting better than anything else, but people may hide them, so Rubenstein and colleagues suggest anonymous surveys, neutral administrators and a climate where honesty is not punished.¹

4. Match the lever to the cause

The table sums up where each common lever stands.

A cause found in one division calls for a fix in that division. If high performers in one unit report poor relationships with their supervisors, Allen and colleagues suggest that supervisor training aimed at that unit may be more cost-effective than a company-wide program.³

5. Test before you scale

Because nearly every study behind this page is correlational, treat each change as an experiment. Try it on one team or site, keep a similar group for comparison, and track actual departures for about a year, the most common follow-up in the studies the 2018 review pooled.¹ Randomized tests of retention practices are uncommon among the sources behind this article. One, in 28 Gap stores in two US cities in 2015 and 2016, run by university researchers with funding from foundations and from Gap, made shifts more stable and predictable. Overall turnover did not change, though the researchers’ preliminary analysis found fewer experienced associates quitting, a subgroup result.⁸

Your retention plan, in order

• Break departures down by team, role, tenure and performance
• Compare the rate with your industry, not the national average
• Survey or talk to people who stayed as well as those who left
• Note the event that came before each resignation
• Check supervision, fairness and rewards beyond pay before pay itself
• Test one change on one team and track who leaves for a year

Download PDF ↓

Get this checklist as a one-page PDF. We'll email you the download link.

Your emailAlso send me the weekly newsletter (launching soon)Email me the PDF

We use your address to send this link and, only if you tick the box, the newsletter. Privacy policy.

The bottom line

No single lever is likely to explain most departures, so start by finding out who is leaving and why, then work on the levers the research ranks highest and your own data point to: rewards beyond pay, the quality of supervision, fit, and the ties that make a job hard to leave. Pay still matters, especially for a role the market is bidding up, but on its own it is rarely the whole answer. Change one thing on one team, and count who is still there a year later.