10 Metrics to Evaluate Executive Coaching Programs

Choose metrics to evaluate executive coaching programs by starting with the decision the evidence must inform. You might need to improve delivery, understand workplace application or decide whether a pilot warrants expansion. Those questions require different information, and none is answered by collecting the largest possible dashboard.

This guide develops ten measures, an original planning worksheet and a fictional four-leader pilot. It separates attempts from quality, observed change from coaching attribution, and financial arithmetic from a credible monetary benefit.

We build Coachful. Coachful is our #1 recommendation for organizing agreed coaching preparation, actions and reviews. That is an editorial workflow recommendation. A completed task or submitted Form is not automatically a validated leadership score, proof of financial return or permission to disclose private reflections.

Use ten options, rather than ten compulsory targets

Choose only measures that serve the actual review. Give each a definition, source, owner, observation period and audience. If nobody can explain what a number would change, question whether collecting it is proportionate.

Use evaluation frameworks without assuming a causal chain

Kirkpatrick Partners describes the historical four categories as Reaction, Learning, Behavior and Results. Its current account emphasizes starting with intended results, a nonlinear evaluation process and the performance environment. The framework helps distinguish evidence; it does not make favorable feedback automatically prove workplace or business impact.

ROI Institute’s separate methodology includes a financial-return level, alongside other evaluation data. Its account calls for isolating effects, valuing benefits and including fully loaded costs, while recognizing intangible benefits. This is not simply a fifth Kirkpatrick level, and not every engagement needs a monetary estimate.

Ask what could enable or block application. A leader may understand a decision process but lack authority, capacity or a relevant opportunity. Report those conditions rather than classifying every absent outcome as either resistance or a failed coaching technique.

1. Multi-rater behavior ratings

Ask observers about behavior they could actually see. “Names the decision owner before committing a date” is more interpretable than “shows executive presence.” Define the period and permit a not-observed response. A colleague who never attended the relevant meeting should not be forced to invent a rating.

Compare respondent groups, items and conditions before interpreting movement. A new manager or substantially changed role can alter the evidence. Report missing responses and avoid presenting a rise from three to four as a precise percentage increase in leadership. A licensed assessment also needs its authorized administration and interpretation process.

2. Goal attainment

Write the expected outcome before reviewing progress. Describe what would count as below expectation, the agreed target and a stronger outcome, with observable support. Keep the difficulty, relevant opportunities and any changed criteria visible.

Kiresuk and Sherman’s 1968 Goal Attainment Scaling paper concerned community mental-health program evaluation. Its abstract describes predetermined individual scales, standardized scores and comparison of treatment modes. An original coaching checklist is not automatically that formal method or a validated clinical measure.

For a practical review, say exactly what you used. A target of ten qualifying handoffs out of twelve can guide a discussion without being called a standardized GAS score. If the target changes midway, preserve the original and explain the new basis.

3. Retention

Name the group and period: for example, all eligible pilot participants employed at the agreed start, followed through a specified date. Distinguish an internal transfer from leaving the organization if that matters to the decision. Do not quietly remove leavers from the starting denominator.

Staying does not prove coaching prevented departure. Pay, management, family circumstances and labor-market conditions may matter. Leaving an unsuitable role may also be a reasoned client choice. If financial valuation is needed, use an appropriate organizational cost estimate and show its assumptions, rather than applying a universal salary multiplier.

4. Team experience

An existing survey or structured discussion may illuminate clarity, feedback or workload. Use a suitable question and explain who is invited, how responses will be used and whether participation is optional. Avoid adding a second survey merely because a coaching dashboard has an empty field.

Small groups and distinctive comments can identify people even when names are removed. Aggregation is not an anonymity guarantee. A higher team score should be reported with response counts, relevant workforce changes and other initiatives, not converted into proof of a culture-wide coaching effect.

5. Role or succession readiness

Use the organization’s actual role criteria and responsible assessment process. Ask what work examples support the responsibilities, what remains untested and which support is available. Keep coaching observations separate from a formal selection or promotion recommendation unless that role and method are explicitly appropriate.

A person can develop relevant capabilities without a vacancy opening. A promotion can also occur for reasons beyond coaching. Record the event without treating it as a complete causal evaluation or exposing private session material to justify the talent decision.

6. Financial ROI

Financial ROI compares net attributed monetary benefits with the full cost: subtract cost from attributed benefit, divide by cost and multiply by 100. The benefit-cost ratio divides benefit by cost without that subtraction. Label the currency, period, cost boundary and attribution method.

These fictional figures demonstrate arithmetic, not the pilot’s return or a benchmark. Provider fees alone may omit participant time, administration and evaluation. A stakeholder’s estimated benefit remains an estimate after a confidence discount; an arbitrary haircut does not create an experiment.

If there is no credible attribution or monetary value, report that gap. Preserve useful nonfinancial evidence rather than inventing money to fill a report. See the coaching ROI research guide for the distinctions behind headline figures.

7. Agreed action attempts

Count actions against the opportunities that were relevant. Ten preparation briefs out of twelve planned handoffs is different from ten activities of any kind. Decide what counts as an attempt and how canceled opportunities, absent records and uncompleted work are handled.

Then inspect quality. A brief can exist while omitting the accepted decision owner. A completed task can record a useful attempt that did not produce the desired result. Keep those states separate instead of calling completion a percentage improvement in leadership.

8. Sponsor alignment

Ask the client, sponsor and coach what outcome matters, what support the organization will provide and what information may be shared. Differences can reveal an unworkable arrangement before anyone measures it. Record accepted decisions and outstanding conditions, rather than treating an unanswered request as agreement.

Alignment does not mean opening the coaching sessions to the sponsor. The current ICF Code of Ethics addresses clear agreements and information boundaries. Agree the reporting purpose and recipients, then check the actual summary against that agreement.

9. Cross-functional coordination

Define the process boundary. A handoff measure might require a decision question, an accepted decision owner and an agreed recipient deadline at transmission. Timeliness could separately mean receipt by that accepted deadline. A message sent quickly is not necessarily a complete or usable request.

Keep the cases behind the average. Simpler requests, a redesigned form, extra staffing or changed approval authority may explain part of a difference. Describe the observed change and plausible contributors without claiming that coaching alone shortened every decision.

10. Client learning account

Invite a concrete example: what did the client understand differently, try, reject or still find difficult? Ask what did not help and what a different observer might say. A specific contradiction is usually more useful than a required statement that the program was transformative.

Preserve the client’s meaning and audience choice. A learning account can complement work records, but it is not automatically evidence of self-authorship, cognitive complexity or emotional maturity. Those claims require an appropriate method; an articulate closing reflection does not supply one.

Worked pilot: choose evidence for a continuation decision

In this fictional case, Tamsin sponsors a six-week pilot from 2 November through 11 December 2026. Owen coaches four leaders: Ari, Bela, Cleo and Drew. Nell owns the operational evaluation, and Kit reviews any financial claim. The decision is whether the delivery model and handoff practice justify expansion.

Before starting, all four leaders voluntarily agree to the coaching arrangement and the exact reporting boundary. Tamsin accepts a summary of defined process counts, method limitations and support needs, with no private coaching narratives or raw reflections. The small cohort is not promised anonymity.

They choose action attempts, coordination quality and an alignment review, with an original goal of ten qualifying handoffs out of twelve. They do not administer a formal GAS instrument, commission a new 360 or claim to evaluate retention over six weeks. The unselected measures remain options, not fabricated zeros.

Nell identifies ten eligible baseline handoffs from 19–30 October. Six contain all required fields at transmission. For the pilot, twelve planned handoffs have recipient-accepted deadlines, three per leader. No eligible handoff is silently removed because its preparation is missing.

Owen delivers three forty-five-minute coaching sessions to each leader, all held by 1 December. Twelve sessions total nine coach-live hours. Preparation, participant work, operational administration, reporting and receiving time are outside that narrow total. The pilot’s whole labor and monetary cost have not been established.

Count attempts, quality and timeliness separately

By 11 December, ten of the twelve handoffs have a preparation brief. Nine meet all three quality requirements at transmission. Eight of the twelve arrive by their accepted recipient deadline, and seven are both qualifying and timely.

The two absent briefs count as nonqualifying handoffs, not proof that the leaders did nothing elsewhere. The remaining nonqualifying brief, C08, omits an accepted decision owner. It is a completed preparation attempt, but it cannot satisfy the quality definition.

For C08, the receiving team had accepted the handoff deadline, and the message arrived within it. The proposed decision owner had not accepted that role. Those are different agreements, so timeliness cannot repair the missing quality field. C09 has the opposite pattern: the decision owner and other required fields are supported, but receipt is late.

Nell keeps a row for each eligible handoff, its source and the reason for classification. She does not change a historical row to qualifying because someone might accept the owner role later. A later repair belongs in a new dated field, preserving what the recipient actually had at transmission.

The qualifying proportion is six of ten before the pilot, or 60%, and nine of twelve during it, or 75%. That is a fifteen-percentage-point difference across two different sets of requests. It is not a fifteen-point rise on a validated leadership scale or a controlled estimate of coaching impact.

Correct the report rather than the target

Nell’s 11 December draft mistakenly calls all ten submitted briefs qualifying. During the actual 14 December evidence review, Bela points to C08’s absent owner acceptance. Nell checks the source, acknowledges the classification error and corrects the count to nine on 15 December. She retains the earlier draft in the evaluation history.

The agreed target stays ten of twelve. The final result is below it. At the closing cutoff, C08’s decision-owner acceptance remains unresolved; correcting the report has not repaired that operational condition.

Tamsin asks during the review to see the private coaching notes to explain the two missing briefs. Owen explains the agreed boundary and offers to include only permitted process facts and client-approved support needs. On 15 December Tamsin withdraws the notes request and accepts that limited route.

There are also competing explanations. A shared handoff form was introduced on 16 November and an additional analyst started on 23 November. The requests are not identical across periods, and no comparison group was established. Kit therefore rejects a proposed monetary-savings headline on 15 December: neither attributable hours nor an appropriate monetary benefit has been verified.

Make the evidence usable without opening private material

Owen deliberately grants each leader an original review Form through Clients > client > Worksheets & resources/Assign, sends the private client-access link and verifies that all four can open it before the review. The questions state the purpose, recipients and optional support comment. A licensed instrument would require its own authorized route.

A standalone assigned Form does not automatically appear in My Coaching’s next-action list. The pilot uses the granted client link. Confirm the actual response access available to authorized staff; neither a private-sounding title nor a broad administrative role is a confidentiality agreement.

A goal-linked task may help a leader remember the next preparation attempt. Its completion does not certify the accepted owner field or write the evaluation’s financial result. Nell maintains the approved organizational source records in their authorized setting, rather than attaching raw business files to a shared discussion.

Review automatic session recap and approval settings before relying on review or privacy. Keep personal journal and takeaway pages unshared when they contain private reflection. Where an enabled Share with coach control is used, the client intentionally exposes that page; it is not then an unshared personal record.

Secure the actual summary and close the service

All four leaders review and approve the exact limited summary by 17:00 on 15 December. Nell sends it to Tamsin at 09:00 on 16 December. Tamsin reads it at 11:00 and acknowledges the nine-of-twelve result, missed target and unverified financial return. Approval to send is distinct from this actual receipt.

The final summary’s original decision text reads:

Ten preparation briefs were completed across twelve eligible pilot handoffs. Nine met the agreed quality definition at transmission; eight were timely, with seven meeting both conditions. The target was ten qualifying handoffs, so it was not met.

The baseline qualifying proportion was six of ten. The pilot proportion is nine of twelve. The samples differ, a shared form was introduced and staffing changed. These observations do not isolate a coaching effect. Financial return has not been established, and one decision-owner acceptance remains unresolved.

The evidence supports reviewing the handoff and delivery arrangements. It does not establish readiness for promotion, a developmental-stage transition or permission to expand the service.

The sponsor receives an account relevant to the agreed decision, rather than an export of personal reflections. The report includes the miss and the unresolved condition in the same text as the favorable observed difference. It does not hide them in a separate appendix while leading with a claim that every leader improved.

The service allows one clarification per leader by noon on 14 December, Owen’s answer by 16 December at 17:00 and administration through 17 December at 17:00, Europe/London. Only Cleo asks, at 11:00 on 14 December, whether the activity evidence establishes a developmental-stage change.

Owen answers at 16:00 on 16 December, one hour early: the records describe particular attempts and handoffs, not a stage assessment. Cleo reads and acknowledges this on 17 December at 09:00. The other three submit no clarification; that absence is not interpreted as satisfaction or disengagement. Owen closes administration at 17:00.

On 17 December Tamsin declines expansion on the current evidence and asks for a revised proposal. No new scope, resources or date have been accepted. The observed improvement remains useful, but it has not produced a rollout authorization or resolved C08.

Use an original evaluation planning worksheet

For the broader evidence, read research on coaching effectiveness. Explore Coachful’s features for the agreed preparation and review workflow. The ten-measure menu and this worksheet are original practical applications, not a universal evaluation standard or a proprietary assessment.

Frequently asked questions

Must an executive-coaching program use all ten metrics?

No. Choose a proportionate set that answers the actual decision. Record why measures are selected or excluded, so an empty ROI field is not mistaken for a measured zero and an uncollected survey is not presented as evidence.

Does a better 360 score prove coaching worked?

It describes reported perceptions under a particular instrument and set of conditions. Check comparable raters, items, observation opportunities and other explanations. It does not alone identify the causal contribution of coaching.

Can completion rates replace behavior evidence?

No. Completion can identify an attempted activity. Check the agreed quality criteria and workplace consequence separately. In the case, ten briefs exist but only nine qualify; the difference changes the actual target decision.

Should every pilot publish financial ROI?

No. Credible attribution, monetary values and full costs are prerequisites for a useful financial claim. Keep hypothetical calculations separate from observed return, and report relevant nonfinancial findings without converting them into invented savings.

Does a summary guarantee anonymity?

No. Small samples and distinctive details can reveal identities. Agree the audience and content, inspect the actual summary and avoid promising confidentiality that the method or access arrangement cannot provide.