Do Manager Training Programs Work? What the Evidence Says

Evidence grade: Mixed. Credible causal studies exist and disagree, or the effect holds only under conditions the studies can name. This grade describes the published evidence. It is not a prediction about your workforce and not a result RetainScore has produced.

Two things are strongly evidenced. The course is not one of them

The case for manager training rests on a chain of three links. Managers affect whether people stay. Training changes what managers do. Therefore training managers keeps people. The first two links are about as well evidenced as anything in this field. The third is the one the budget is for, and it has been tested directly once, as half of a bundle. This page grades the third link, because it is the claim being made.

Managers move quits

Start with the meta-analytic literature the site's own diagnostic is weighted from. Across 42 samples and 28,637 employees, leadership quality correlates -.24 with voluntary turnover, the second-strongest association among the factors we score.

The sharpest single study is Hoffman and Tadelis at a large technology firm, using the firm's own personnel records and employee surveys about their managers. Their headline is that "survey-measured people management skills have a strong negative relation to employee turnover": a manager one standard deviation better at the people side of the job has about 28% less attrition on their team. What makes this more than a correlation is the design: "A causal interpretation is reinforced by research designs exploiting new workers joining the firm and managers moving jobs", which follows people assigned to managers rather than people who chose them. Retention is also where the skill shows up. Those same managers "do not consistently improve non-attrition outcomes" the firm tracked.

The cheapest experiment on record says managers can move quits without being trained at all. Friebel, Heinz and Zubanov ran a randomized trial across a large retail chain: "We conduct a large-scale (238 stores, 7,700 workers) and long-term (lasting 16 months) RCT in a retail firm". The intervention was a message from the chief executive that "sets new goals for the managers of the treated stores", asking them to do what they could about quits. "The treatment decreases the quit rate by a fifth to a quarter, lasting nine months before petering out, but reappearing after a reminder". The mechanism was time, not skill: "treated store managers spend more time on human resources (HR) and less on customer service", and sales did not fall.

Training moves managers

Lacerenza and colleagues pooled 335 independent samples of leadership training and concluded that "leadership training is substantially more effective than previously thought", with corrected effects of 0.63 on reactions, 0.73 on learning, 0.82 on transfer to the job and 0.72 on results. Training does change what managers know and do.

Read the results column carefully, though. Turnover is never estimated on its own. It sits inside "organizational results (e.g., turnover, absenteeism, ROI, profit)", pooled with profit and return on investment, at 0.75. The row that measures what the trained manager's staff actually experience is a different size: "Subordinate outcomes 30 2,507 .21 .22", a fifth of a standard deviation across 30 samples. The training reaches the manager. Much less of it reaches the people who report to them.

When trained managers meet a turnover outcome

One randomized experiment puts supervisor training in front of a measured exit rate. The Work, Family and Health Network randomized teams at a large IT firm into "a flexibility/supervisor support initiative called STAR", which trained supervisors in supporting employees' lives outside work and, at the same time, gave teams control over when and where they worked. Moen and colleagues report that being on a STAR team "lowered turnover intentions 12 months later and reduced the risk of voluntary turnover over almost three years", through a merger announcement that landed in the middle of the study. It is the best evidence there is for the third link, and the training was half of it. The trial cannot say what the course would have done without the schedule control that came with it.

The other randomized test is from Cambodian garment factories, where the IFC and ideas42 randomly assigned supervisors to a soft-skills course. The result on the people side was clear: "training the direct supervisors leads to higher personal job satisfaction for workers and more pleasant work relationships". On the business side, "Researchers did not find a significant impact of direct supervisor training on any of the productivity and production measures". Turnover was in the design, since "employee rosters were collected, before and after the training, to measure overall worker and supervisor turnover", but the abridged report the public can read gives no estimate of what happened to it. The finding that trained supervisors were liked more and moved nothing measurable is not nothing; it is the pattern the leadership meta-analysis predicts.

Why the grade is mixed

Mixed is the grade for credible studies that disagree, or for an effect that holds only under conditions the studies can name. The conditions here are named. Quits fell when managers were given the goal and the time (retail), and when training came with authority over schedules (STAR). Where training was delivered alone, relationships improved and the business numbers did not (Cambodia), and the meta-analytic effect on subordinates is small. The evidence does not say manager training fails. It says a course by itself has not been shown to keep anyone, and the things that have been shown to keep people are what the manager does with time and permission afterwards.

This page replaced one of ours

Until today this URL held a page titled Manager Training as a Retention Investment. It built the case from Gallup's claim that managers account for 70% of the variance in engagement, chained to a survey figure on turnover in low-engagement teams. Our own provenance trace grades the first figure as method-not-published, and the chain from engagement surveys to turnover is the one the engagement-surveys grade takes apart. The conclusion was roughly right and the evidence under it was not, so the page was replaced with the studies that carry the weight.

The honest boundaries

• No trial isolates the course. STAR bundled training with schedule control; the retail trial had no training; Cambodia reports no turnover estimate. The clean experiment, train managers and nothing else and count who leaves, has not been published.
• One source is read from its abstract. The STAR turnover paper sits behind a publisher paywall with no open copy; the claims drawn from it are the ones its abstract states.
• The tech-firm study is one firm. Its designs are stronger than most, and it is still a single employer with an unusually well-measured workforce.
• Effects fade. The retail effect lasted about nine months and needed a reminder. Whatever training does, expect it to decay on the same timescale unless something keeps it live.

If you are buying manager training

• Train the behaviours the evidence links to leaving. Support for people's lives outside work, clear expectations, workload, the things the retention diagnostic scores. Generic leadership content has the weakest subordinate effects in the meta-analysis.
• Give managers the time and the permission first. The two interventions that cut quits changed what managers were allowed to spend time on. A trained manager with no hours for their team is the Cambodia result.
• Try the cheapest version before the expensive one. Telling managers that retention is theirs, and reminding them, cut quits by a fifth in a 238-store trial. It costs a letter.
• Measure quits by manager, before and after. That is the number the studies moved and the one your payroll already holds. Course satisfaction scores are the reactions row, the least informative in the meta-analysis.

Read the companion page on bad managers as a cause of turnover, or the grade on whether mentorship programs work, where the same gap between changing the helper and changing the outcome appears.