This article is for readers who want to go further than “studies show coaching works” — who want to understand what kinds of studies, measuring what kinds of outcomes, with what limitations, producing what level of confidence. It covers more methodological ground than the overview in Is Life Coaching Worth It? What the Research Says, and is written for readers comfortable with that level of detail.
The shape of the evidence base
Coaching research sits in an unusual position within the behavioural science literature. It’s not a young field — serious research on coaching effectiveness has been accumulating since the mid-1990s — but it’s substantially smaller and less methodologically mature than research on psychotherapy, which has had decades of large, well-controlled trials and rigorous meta-analysis behind it.
The evidence base as of the mid-2020s includes several dozen randomized controlled trials of varying quality, a larger number of quasi-experimental studies using pre-post designs without control groups, a growing set of meta-analyses that synthesize findings across studies, qualitative research on coaching processes and client experiences, and a substantial practitioner literature that varies widely in rigour.
The strongest evidence comes from randomized controlled trials and the meta-analyses built on them. The weakest — though most voluminous — comes from case studies, testimonials, and self-report surveys conducted by coaching organizations without independent verification.
The major meta-analyses and what they found
Theeboom, Beersma, and van Vianen (2014) published one of the earliest rigorous meta-analyses of coaching effectiveness in the Journal of Positive Psychology, reviewing 18 controlled studies. They found significant positive effects across five outcome categories: performance and skills, well-being, coping, work attitudes, and goal-directed self-regulation. Effect sizes were in the moderate range — meaningful and consistent, not dramatic. This meta-analysis is frequently cited because the authors restricted inclusion to controlled studies and used rigorous coding procedures.
Jones, Woods, and Guillaume (2016) published a meta-analysis in the Journal of Occupational and Organizational Psychology focused on workplace coaching, reviewing 17 controlled studies. They found significant positive effects on performance, well-being, and attitudes, with particularly strong effects for goal attainment. Notably, they found that longer coaching engagements and coaching by external rather than internal coaches produced stronger effects — a finding with practical implications for how coaching is structured.
Sonesh et al. (2015) reviewed 17 studies in Coaching: An International Journal of Theory, Research and Practice and found significant effects on learning and performance outcomes. Their analysis also identified coaching relationship quality as a significant moderator of outcomes — the better the working alliance between coach and client, the stronger the results. This finding parallels a large body of psychotherapy research showing that relationship quality predicts outcomes across therapeutic modalities.
Grover and Furnham (2016) conducted a broader review in Frontiers in Psychology and found consistent evidence for coaching effectiveness on well-being outcomes, while noting significant methodological heterogeneity across studies — meaning the variation in study design, outcome measures, and populations makes it difficult to draw strong general conclusions from the aggregate.
The coaching relationship quality appears to matter at least as much as the specific approach used. This is a finding coaching marketers tend not to emphasize — but it has real implications for how to choose a coach.
What the studies measure — and why it matters
Understanding what coaching research measures is essential to interpreting it correctly. Most studies measure outcomes through one or more of the following approaches, each with its own strengths and limitations.
Self-report questionnaires are the most common method. Participants rate their own goal attainment, well-being, confidence, or performance before and after coaching. Self-report measures capture subjective experience — which is often exactly what coaching targets. Their limitation is response bias: people who have invested time and money in coaching may rate their outcomes positively regardless of whether meaningful change occurred. Studies that rely exclusively on self-report should be read with this in mind.
Validated psychological scales are a stronger form of self-report. Studies that use validated instruments — the Warwick-Edinburgh Mental Wellbeing Scale, Goal Attainment Scaling methodology, or validated measures of resilience or self-efficacy — provide more defensible data than studies using ad hoc questionnaires. When a study specifies the instrument used, that’s worth noting when evaluating its findings.
Behavioural observation and 360-degree feedback are used more commonly in executive coaching research. Rather than asking clients how they’re doing, these methods ask the people around them — managers, direct reports, peers. This approach is less subject to self-report bias and captures outcomes that matter in organizational contexts. Its limitation is that behavioural change may lag psychological change, and organizational politics can contaminate 360 data.
Organizational performance metrics — productivity, retention, revenue — are the most compelling to skeptics but the hardest to isolate. Many other factors influence organizational performance, and attribution to coaching specifically is methodologically difficult.
The methodological limitations that matter
Sample size. Many individual coaching studies have small samples — fewer than 50 participants in each condition. Small samples produce unstable estimates; an effect that appears significant in a study of 30 people may not replicate in a study of 300. Meta-analysis helps by aggregating across studies, but only to the extent that the underlying studies are sound.
Control conditions. A controlled study compares coached participants against a control group who receive no coaching. But what does the control group receive? In many coaching studies, the control condition is simply “no intervention.” This is a weak control because it doesn’t account for the effect of attention itself — being enrolled in a study, meeting regularly with someone, being asked to set goals. Studies with active control conditions, where the control group receives structured support that isn’t coaching, are more informative but substantially rarer.
Blinding. In clinical drug trials, the gold standard is a double-blind design where neither participants nor researchers know who received the treatment. This is essentially impossible in coaching research — participants know whether they’ve been coached. This limitation is inherent to the field and shared with psychotherapy research, but it’s worth knowing when evaluating effect size estimates.
Follow-up periods. Most coaching studies measure outcomes immediately after the engagement ends, or within a few months. Very few studies track participants at six months, a year, or longer. The research base tells us substantially more about short-term than long-term effects. Whether coaching produces durable change, and under what conditions, remains one of the more important open questions in the literature.
What the evidence doesn’t resolve
After two decades of accumulating research, some important questions remain genuinely open.
Which clients benefit most? The research consistently shows positive average effects, but averages obscure variation. Some clients will benefit substantially from coaching; some modestly; some may not benefit meaningfully at all. The literature has not yet produced reliable guidance on which individual characteristics predict the strongest outcomes — though motivation and readiness consistently emerge as important factors across studies.
Which coaching approaches are most effective? Most studies treat “coaching” as a single intervention without distinguishing between different theoretical approaches — cognitive-behavioural coaching, solution-focused coaching, co-active coaching, ontological coaching. Whether different approaches produce different outcomes for different clients is largely unstudied. The field’s tendency to study coaching as a category rather than a set of distinct methods is a significant gap.
How does coach quality affect outcomes? This is perhaps the most practically important open question. Studies credential coaches but don’t assess their skill. The variance in coaching quality within credentialed populations is almost certainly large, and its effect on outcomes is almost certainly significant — but the research hasn’t found a reliable way to measure it. This means that effect sizes from research studies may not translate directly to what you’ll experience with a specific coach.
An honest summary
The evidence for coaching is meaningful. Multiple rigorous meta-analyses, across different research teams and different outcome domains, find consistent positive effects. The effects are moderate in size — real, but not transformative in the dramatic sense that coaching marketing sometimes implies. Relationship quality appears to matter at least as much as specific approach. Client motivation is a consistent predictor of outcome.
The evidence also has genuine limitations. Samples are often small, follow-up periods are short, self-report bias is pervasive, and publication bias almost certainly inflates the published estimates. The evidence for coaching is not the evidence base that exists for established psychotherapies — it’s younger, smaller, and methodologically less mature.
The honest position: coaching, done well with motivated clients, produces real results. The confidence with which advocates cite the research sometimes exceeds what the research can actually support. The confidence with which critics dismiss it is also unwarranted. The honest read is somewhere between those two poles — which is, admittedly, where the honest read on most complex questions lives.
- The coaching evidence base includes multiple rigorous meta-analyses showing consistent moderate positive effects on performance, well-being, goal attainment, and self-regulation.
- The strongest meta-analyses — Theeboom et al. (2014), Jones et al. (2016) — restricted inclusion to controlled studies and found meaningful, consistent effects across outcome domains.
- Coaching relationship quality (the working alliance) is among the strongest predictors of outcome across studies — often stronger than the specific approach used.
- Key methodological limitations: small samples, weak control conditions, heavy reliance on self-report, short follow-up periods, and pervasive publication bias.
- Important open questions: which clients benefit most, whether different approaches produce different outcomes, and how coach quality (within credentialed populations) affects results.
- The honest summary: the evidence supports coaching as genuinely effective under the right conditions. The confidence of advocates slightly exceeds what the research warrants. The dismissiveness of critics is not warranted either.