Why coaching works, what separates training from coaching, and why the quality of the coach decides the outcome. Peer-reviewed sources, plain-language numbers, and what the research does not show.
A note on the percentages used on this site: most studies report their results as an effect size (a statistic such as Hedges' g or Cohen's d) rather than a percentage. To keep the pages readable, this site converts each effect size into a plain statement, "the average coached person did better than X percent of the comparison group," using Cohen's U3, the standard conversion from Jacob Cohen's Statistical Power Analysis for the Behavioral Sciences (1988). The original effect sizes and study details are listed in each piece's "research behind this" section.
Coached people outperform about two in three of the uncoached, and three in four on behavior. Practice with feedback nearly doubles the effect of practice alone.
ReadTraining gives people the knowledge. Coaching works with one person on how they use it with another. Both matter; they are not the same thing.
ReadWhy coaching has been expensive, scarce and uneven, what that costs, and what the research says separates the coaching that works.
ReadWhat the research found. A 2023 meta-analysis that pooled only randomized controlled trials of executive coaching found a moderate overall effect and a larger one on behavior, concluding that coaching "is a powerful instrument for organizations to support positive change" (Nicolau, Candel, Constantin and Kleingeld, 2023). Earlier meta-analyses reached the same place: "coaching is, overall, an effective intervention in organizations" (Theeboom, Beersma and van Vianen, 2014), with positive effects across skills, attitudes and individual results, and no loss of effect when coaching was delivered digitally (Jones, Woods and Guillaume, 2016). The mechanism is practice followed by specific feedback. Across 159 experiments, retrieval practice improved retention, and when feedback followed the practice the effect rose by nearly half again (Rowland, 2014). In a 2026 randomized experiment with 968 participants, people who practiced difficult conversations with an AI partner and received written feedback improved far more than those who watched a coaching video: 21.6 percent reliably improved, against 9.0 percent for the video and 2.9 percent for the control (Kumar et al., 2026, preprint).
In the workplace. In a 2,327-agent contact-center study, the coaching each person received each month "predicted objective performance improvements over time" (Liu and Batt, 2010). In one early field study, a management training program raised productivity 22.4 percent; the same training followed by eight weeks of one-on-one coaching raised it 88 percent (Olivero, Bane and Kopelman, 1997). It was 31 people in one agency, so it is a story rather than a law, but the pattern it describes has held up.
What this means for a sales or service team. Knowledge is the foundation, and training builds it. The research says the skill of using that knowledge with a person is built the same way every skill is: by doing it, with specific feedback, repeatedly. DAiVE is built to deliver exactly that loop to every person on the team, and publishes its own results as pilots report.
Training is how an organization gives people knowledge: the product, the process, the standard, delivered to a group through a classroom, a module, a video or a program that reinforces it over time. It is the foundation, and good training programs build it well and keep it in place.
Coaching does a different job. It works with one person on how they use that knowledge with another person: how they read what matters to the customer in front of them, choose what to say, handle the question that was not in the module, and carry the conversation through to a result. Knowledge can be taught to a room. The skill of using it is built one person at a time, through practice, specific feedback on that practice, and repetition.
The research on skill-building is consistent on that mechanism. Practice in simulations raised trainees' confidence by 20 percent, procedural knowledge by 14 percent and retention by 9 percent across 65 studies, and the gains held when the simulation "was a supplement to other instructional methods rather than stand-alone instruction" (Sitzmann, 2011). Practice with feedback nearly doubles the effect of practice alone (Rowland, 2014). And the feedback itself matters: "Feedback is more effective the more information it contains" (Wisniewski, Zierer and Hattie, 2020), with information-rich feedback about four times as effective as praise or correction.
So the question for an organization is not training or coaching. It is whether the people who have the knowledge also get the coaching. That is what DAiVE adds: a practice partner for every person, scored against the organization's own standard, with written coaching after every completed session, so the knowledge the training built is the knowledge the customer experiences.
Why coaching has been so hard to deliver. A professional coaching session costs an average of US$244 an hour worldwide, and only 27 percent of organizations make professional coaching available to all employees; it reaches managers and executives first (ICF, 2023). The fallback is the manager. Managers spend about 13 percent of their time developing the people who work for them (Deloitte, 2025), and 45 percent say they do not feel confident developing the skills their employees need (Gartner, 2019). What is left is coaching that is improvised, inconsistent, and rarely measured.
What uneven coaching costs. The largest meta-analysis of feedback research, covering 607 effect sizes and 23,663 observations, found that feedback improved performance on average "but that over 1/3 of the FIs decreased performance" (Kluger and DeNisi, 1996). The difference was not whether feedback was given but what it contained. Gartner's study of more than 7,000 employees and managers found "Always-On" managers who comment on everything reduce performance by up to 8 percent, while managers who give fewer, better-targeted interventions improve it by up to 26 percent; "The 'Always-on' manager often provides misguided, irrelevant or too much feedback" (Gartner, 2018). Only 26 percent of employees strongly agree the feedback they receive helps them do better work (Gallup, 2019).
The same pattern in sales. In a one-year study of 1,246 sales representatives in 136 teams, the manager's coaching skill predicted annual sales goal attainment, and "coaching frequency had a negative effect on goal attainment when skill was low" (Dahling, Taylor, Chau and Dwight, 2016). A 2024 peer-reviewed experiment with 374 participants found targeted, mistake-specific feedback after a practice negotiation improved the next one's outcome, while generic feedback was statistically no better than none (Shea et al., 2024). A 2025 meta-analysis of 41 studies found no statistically significant difference in learning performance between AI-generated and human feedback (Kaliisa et al., 2025). What moves performance is the specificity and consistency of the feedback, whoever or whatever delivers it.
Where DAiVE fits. DAiVE was built to remove the three constraints. Every person is coached, not the few a budget can reach. Every conversation is measured against the same standard, so the quality of the coaching does not depend on who is free that afternoon. The written coaching names the specific answer, the specific step and the specific fact, which is the information-rich feedback the research rewards. And the results are visible, by person and by location.
The research does not show that any AI coach outperforms a skilled human coach; the most recent working-adult trial found the reverse (de Haan, Terblanche and Nowack, 2026). It does not show that "sales reps forget 70 percent of training in a week"; that figure has no primary source and DAiVE does not use it. It does not show that three-quarters of feedback is harmful; the meta-analysis says over one-third. It does not show that coaching "quadruples" results; that was 31 people in 1997. What it does show is that coaching changes behavior, that practice with specific feedback is the mechanism, and that the quality and consistency of the coaching decides the outcome. DAiVE publishes its own client results as pilots report.