Recent workplace coaching studies: moderate effect, but no study measured business results
A meta-analysis of 11 workplace coaching studies published after 2014 finds g = 0.44 overall, larger for skills than for attitudes, and notes that no included study measured results-level outcomes.
Finding
What the study measuredAcross 11 studies, workplace coaching improved work outcomes with g = 0.44 (95% CI 0.31 to 0.57) after one extreme study was removed. Effects on skills (g = 0.72) were larger than on affective outcomes (g = 0.41), but the difference was not statistically significant. Face-to-face (g = 0.48) and virtual (g = 0.35) coaching did not differ significantly, and neither the number of sessions nor total hours predicted the effect. No included study measured results-level (organizational performance) outcomes.
Authors' interpretation
What the authors concludeThe authors conclude that workplace coaching is effective but that the science is immature. They ask researchers to describe coaching approaches and coach credentials in detail, to measure results-level outcomes, to study real coaching clients, and to use longitudinal designs.
| Question type | effect |
|---|---|
| Population and setting | Workplace coaching by trained coaches, published after 2014, individual-level work outcomes; Workplaces; face-to-face and virtual coaching |
| Intervention | Workplace coaching by trained coaches (process-oriented or outcome-oriented) |
| Comparison | Varies by study; several controlled designs, one randomized |
| Follow-up | Varies by study; not pooled |
| Outcomes | Skill outcomes; Affective outcomes |
| Measures | Self-ratings (most common); Manager or supervisor ratings; Subordinate ratings (one study) |
| Effect size and uncertainty | Overall g = 0.44, 95% CI 0.31 to 0.57, k = 11, after removing one outlier study (g = 3.52); random-effects prediction interval 0.170 to 0.689. Skill outcomes g = 0.72 (95% CI 0.49 to 0.95, k = 5); affective g = 0.41 (95% CI 0.26 to 0.58, k = 8). |
| Limitations | Small number of studies; several moderators rest on one study; thin reporting of coaches, coachees and session content in the primary studies; few controlled studies; publication bias assessed only visually with a funnel plot; some statements in the text do not match the tables (see commentary). |
| Funding and conflicts | No funding statement found. Conflict of interest: the first author was employed by Cannon-Bowers Consulting; the other authors declared none. |
| Replication | Not applicable (synthesis of prior studies). |
| Source reuse terms | CC BY (open access) |
Figures reported, with their location in the source
- g = 0.44 (95% CI 0.31–0.57), k = 11: Overall effect after removing the outlier (Table 3)
- 0.170–0.689: Prediction interval (Table 2)
- g = 3.52: Outlier study removed (Onyishi et al., 2021) (Section 4.1)
- I² 91.3 → 20.09: Heterogeneity before and after removing the outlier (Section 4.1)
- g = 0.72 (0.49–0.95), k = 5: Skill outcomes (Table 3)
- g = 0.41 (0.26–0.58), k = 8: Affective outcomes (Table 3)
- g = 0.48 vs 0.35; Q(1) = 0.67, p > .05: Face-to-face vs virtual (Table 3; Section 4.6)
- manager g = 0.5; self g = 0.41; subordinate g = 0.24 (k = 1); Q(2) = 0.39, p > .05: Effect by rating source (Table 3; Section 4.5)
- sessions p = 0.30; hours p = 0.25: Meta-regressions on dose (Sections 4.7 and 4.8)
Read the source
Cannon-Bowers, J. A., Bowers, C. A., Carlson, C. E., Doherty, S. L., Evans, J., & Hall, J. (2023). Workplace coaching: A meta-analysis and recommendations for advancing the science of coaching. Frontiers in Psychology, 14, 1204166. https://doi.org/10.3389/fpsyg.2023.1204166
Verification record
- Status
- human-verified
- Access level
- full-text
- Date accessed
- 2026-10-24
- Tool-checked
- 2026-10-24
- Automated retraction check
- 2026-10-24: Crossref record has no update, correction or retraction notices (Crossref carries Retraction Watch data). Owner to confirm.
- Human-verified
- 2026-10-24
- Retraction status
- none-found (checked 2026-10-24); Owner check: "Retraction check: r-0003, none-found"
- Last reviewed
- 2026-10-24
How to cite this entry
consultants.coach. (2026). Recent workplace coaching studies: moderate effect, but no study measured business results (Entry r-0003, version 2026-10-24 00:00:00 -0400 EDT). https://consultants.coach/research/r-0003/
Original source: Cannon-Bowers, J. A., Bowers, C. A., Carlson, C. E., Doherty, S. L., Evans, J., & Hall, J. (2023). Workplace coaching: A meta-analysis and recommendations for advancing the science of coaching. Frontiers in Psychology, 14, 1204166. https://doi.org/10.3389/fpsyg.2023.1204166
Change log
- : First published after human verification.
Related entries
Randomized trials of coaching show a moderate average effect, with signs of publication bias
A meta-analysis of 37 randomized controlled trials (39 samples, n = 2,528) finds coaching improves leadership and personal outcomes, with a best estimate of g = .59, and reports significant publication bias.
Workplace coaching meta-analysis: positive effects overall; the largest effects rest on very few studies
A meta-analysis of 17 workplace coaching studies (2,267 people) finds a positive overall effect (δ = 0.36), larger without multisource feedback and with internal coaches, but the biggest effects come from subgroups of only three studies.
Psychologically informed coaching: strongest effects on goal attainment and self-efficacy
A meta-analysis of 20 studies (957 participants) of psychologically informed workplace coaching finds g = 0.51 overall, largest for goal attainment (g = 1.29), with no significant difference between coaching approaches.
Our commentary
consultants.coach editorial viewFor a practitioner answering a sponsor, the most useful finding is the gap: none of the recent studies measured business results, so claims of organizational ROI cannot lean on this evidence. Ratings by managers (g = 0.50) and by the coachee (g = 0.41) were similar; the single study with subordinate ratings is too small to interpret. Read the details with care: the text calls the subordinate result significantly lower and negative, while the table shows a positive, non-significant estimate. The first author's consulting employment is disclosed.