Three meta-analyses have tested whether the number of coaching sessions predicts the outcome. None found that it does. Coaching is still bought, priced and negotiated by exactly that number: six or eight, eight or twelve, twelve with the option of six more.
We lined up the three moderator tests side by side. Different teams, different rules, the same null.
Three reviews, one null
Theeboom, Beersma and van Vianen, 2014. Eighteen studies. The number of coaching sessions was not related to the effectiveness of the intervention.
Jones, Woods and Guillaume, 2016. Seventeen studies, workplace coaching only. No moderation of effect size by the duration of coaching, measured either as the number of sessions or as the longevity of the intervention.
Cannon-Bowers and colleagues, 2023. Eleven studies, tight methodological criteria. Number of sessions was not a significant predictor (Z = 1.03, p = 0.30). Hours of coaching was not a significant predictor either (Z = 1.15, p = 0.25).
Three sets of inclusion rules. Three statistical approaches. Eleven years apart. The same result each time.
We found no moderation of effect size by coaching format (comparing face-to-face, with blended face-to face and e-coaching) or duration of coaching (number of sessions or longevity of intervention).
No finding in the coaching literature converges more strongly. The three reviews disagree on the size of the overall effect, from about 0.36 to about 0.6 depending on what they let in. On dose they agree.
The null covers six to twelve sessions, and averages
The finding is easy to overstate in both directions.
One session is not shown to work as well as ten. The studies in these meta-analyses mostly used engagements of roughly six to twelve sessions. Nobody randomised anyone to a single session. Within the range that has been studied, differences in dose produced no detectable differences in outcome. Below it, the reviews are silent.
Dose can still matter in an individual case. A meta-analytic null is a statement about averages across studies. It is entirely consistent with some people needing more sessions and others fewer, in ways that cancel out in the aggregate. That is the most likely explanation for the result.
The standard purchasing logic is unsupported. Coaching is sold in packages priced by session, and a longer package is sold as a better outcome. The evidence does not support paying a premium for the longer package on those grounds. It does not support blaming a disappointing result on a short engagement either.
Fit, the gaps and the coachee carry the variance that has been found
If dose is not where the variance is, the other candidates are where the purchasing conversation belongs.
Coach fit and the working alliance. Most of the variance in coaching outcomes that anyone has managed to locate sits in the relationship, not in the specification. That makes the matching process, and the ability to change coach without friction or embarrassment, the part of a platform contract to scrutinise.
What happens between sessions. If the change happens in the gaps, then the gaps are the product. A programme with a structured commitment, a way to record what was tried, and a manager who knows a coaching engagement is under way is buying something the session count does not capture.
Who is being coached. Junior leaders show significantly stronger behaviour change from development than senior ones, and coaching budgets are almost universally allocated the other way. Moving a fixed budget down a level is a bigger lever than adding two sessions to each engagement.
None of these fits a procurement template as neatly as a session count. That is most of why the session count survives.
Spacing mattered in leadership training, volume did not
Nobody has tested a mechanism for the coaching null. What follows is inference from adjacent findings, not established fact.
The most likely account is that the work happens between sessions. A coaching conversation produces an intention and a commitment. The change, if it comes, comes in the four weeks before the next one. On that account the number of conversations sets the number of cycles rather than the amount of change, and the constraint is how fast someone can change one thing about how they work. How often they are asked about it does not set that pace.
The leadership training literature offers indirect support. The 2017 meta-analysis of leadership training found that programmes spread across multiple sessions produced significantly stronger transfer and results than single-block programmes, with no difference in learning. Spacing mattered. What was spaced mattered less. The coaching finding has the same shape: the calendar structure of an intervention does more work than its total volume.
The AI coaching trials point the same way. Outcomes there were strongly predicted by the coachee's own initial self-efficacy and hope, not by the intervention's parameters, as set out in what the AI coaching trials found.
None of the three reviews could test whether the gaps between sessions are where coaching works. The null on dose is settled. The mechanism behind it has not been tested by anyone.
Common questions
How many coaching sessions are needed to be effective? The evidence does not support a specific number. Three meta-analyses tested session count against outcomes and none found a significant relationship. Most studied engagements ran roughly six to twelve sessions.
Is a twelve-session package better than a six-session one? No evidence supports paying more for the longer package on outcome grounds. Both fall inside the range where the meta-analyses found no detectable difference.
Does the total number of coaching hours matter? The 2023 meta-analysis tested hours of coaching separately from session count and found it was also not a significant predictor.
Does the length of the engagement in months matter? The 2016 meta-analysis tested longevity of the intervention alongside session count and found no moderation of effect size by either.
Would a single coaching session be enough then? Unknown, and unlikely. Almost no studies tested doses that low, so the null result says nothing about the very short end of the range.
What does predict coaching outcomes, if not dose? The evidence base is too thin to answer confidently. What has been found is that format does not matter, and that in the AI coaching trials the coachee's own initial self-efficacy and hope strongly predicted success.
Should we still specify a number of sessions in a contract? Practically yes, because someone has to be paid for something. The point is that the number is a budgeting convention rather than a quality specification, and it should not be the variable that the negotiation is about.
Sources
- Theeboom, Beersma & van Vianen (2014), Does coaching work? A meta-analysis on the effects of coaching on individual level outcomes in an organizational context, Journal of Positive Psychology 9(1), 1-18
- Jones, Woods & Guillaume (2016), The effectiveness of workplace coaching: A meta-analysis of learning and performance outcomes from coaching
- Cannon-Bowers, Bowers, Carlson, Doherty, Evans & Hall (2023), Workplace coaching: a meta-analysis and recommendations for advancing the science of coaching, Frontiers in Psychology 14:1204166
- Lacerenza, Reyes, Marlow, Joseph & Salas (2017), Leadership training design, delivery, and implementation: A meta-analysis, Journal of Applied Psychology 102(12)
