Executive search

A search-firm retrospective that produces one usable process change

Run an evidence-led executive-search retrospective: separate process from candidate judgment, choose one testable change and give the follow-up an owner. Includes an agenda and example.

By LaisyPublished

A search retrospective can end with an attractive list of lessons and no change in the next assignment. “Improve communication” sounds reasonable, but it does not tell a consultant what to do differently, who will check it or how to recognize an improvement.

The useful output is a specific process change that someone accepts responsibility for testing. The discussion should establish the event behind the proposed lesson, distinguish competing explanations and choose a practical follow-up.

This guide is for reviewing how the firm worked: the instruction, research handoff, client communication or review process. It is not a second candidate-assessment meeting, a blame exercise or a way to turn one assignment's outcome into a universal rule.

TL;DR

  • Bring a small set of relevant events and source records, with confidentiality boundaries intact.
  • Separate what happened, its effect on the work and the team's interpretation of why it happened.
  • Choose one usable change with an accepted owner, a future test and a review point.
  • Treat research on debriefing as background evidence, not a promise of a search-firm performance gain.

Copyable retrospective agenda

Use the sequence as a starting point and adjust the time to the actual discussion. Do not force the conversation through a fixed duration when a material issue needs separate resolution.

StepQuestionWorking output
Establish purposeWhich part of our process are we reviewing?A bounded question, distinct from individual performance judgments
Reconstruct eventsWhat happened and which record supports it?A short dated account with disagreements visible
Examine effectsWhat did the event change for the assignment?Observed consequences separated from estimates
Compare explanationsWhat might explain the problem or useful practice?Alternatives and evidence gaps
Choose a changeWhat can we do differently on the next suitable task?One specific proposal and an accepted owner
Define the testWhat would show that the change helped or created friction?A review criterion, required evidence and a stopping condition
Return to itWhen and where will we examine the result?A named review event and record location

The agenda is an original adaptation for search operations. It does not certify a method or promise a particular result.

Pick the process question before inviting the meeting

Choose a question the participants can investigate. “Why did the assignment fail?” may combine market conditions, client decisions and internal process into a debate too broad to resolve. “Why did the research team continue using a superseded criterion?” is narrower and points toward records the team can inspect.

Identify which participants understand the relevant work and which voices are missing. If a person cannot attend, avoid treating speculation about their intent as their account. Collect an appropriate factual clarification or keep the gap explicit.

The GOV.UK guidance on agile tools and techniques describes retrospectives as a way for teams to review what is working and what is not. That guidance concerns digital-service delivery. The method here adapts the reflective practice to a search process; it is not a search-industry requirement.

Separate urgent corrective action from learning. A live confidentiality issue or an unfulfilled commitment may need attention through the appropriate process immediately. It should not wait for a convenient retrospective slot.

Bring evidence without building a second archive

Select the few records needed to understand the question. These might include the current instruction, an earlier request, the accepted research handoff and a reviewed client update. Retain audience restrictions and avoid copying personal information into a general learning document.

Use the mandate kickoff record to establish the original instruction. Compare it with later confirmed changes. The point is to identify the working state at each event, not to assume today's understanding was available to everyone earlier.

If the issue concerns evidence quality, check the relevant candidate research notes. Keep the discussion focused on how sources and uncertainty were handled. Do not turn the retrospective into unsupported judgments about the people described in those notes.

The Government Data Quality Framework's guidance emphasizes critical data and communicating known limitations. Applied here, that means bringing the evidence needed for the process question and stating where it falls short. It is an adaptation of public-sector guidance, not proof that the team's records meet a standard.

Separate observation, consequence and explanation

Write these as distinct statements. An observation might be that a revised instruction was confirmed after a research request had been sent. A consequence might be that part of the research required review against the new instruction. An explanation might be that the revision had no visible acknowledgement step.

The explanation is a hypothesis until the team checks it. Perhaps the acknowledgement existed but the recipient missed it. Perhaps the instruction was ambiguous. Perhaps the researcher saw the change but did not know which existing records it affected.

Ask what evidence would distinguish those explanations. If the team cannot establish the answer, choose a test that helps next time rather than rewriting uncertainty as a settled cause.

Avoid treating the final assignment outcome as proof of every process choice. A successful outcome can coexist with a fragile handoff. An unsuccessful outcome can follow careful work under difficult external conditions. Review the specific practice on its own evidence.

Use a discussion structure that produces actions

An established facilitation format can help colleagues contribute without forcing immediate agreement. Atlassian's 4Ls retrospective play uses Loved, Loathed, Longed for and Learned. Its board includes six columns: milestones, the four categories and actions.

That is Atlassian's format, not a proprietary Laisy method. For a search team, use a format only if it helps expose concrete events and possible changes. An unstructured account may be more suitable for a narrow issue.

Ask what “confusing client communication” means in the record: an unclear question, the wrong recipient or an absent decision request? Each needs a different remedy.

Retain useful disagreements. The team does not need to agree on every interpretation before trying a bounded improvement. It does need to agree what it is testing and which evidence will be reviewed afterward.

Fictional example: test an acknowledgement step

This example is invented. Its counts describe a possible exercise, not customer evidence, a statistically valid trial or a recommended firm-wide quota.

A fictional search team reviews three occasions on which revised instructions reached the researcher after work had begun. The records show the same uncertainty each time: nobody explicitly confirmed which existing work needed rechecking.

The team proposes an acknowledgement step for material brief revisions. Before expanding research, the researcher will confirm the affected criterion, the work requiring review and any question still unresolved. The consultant will confirm or correct that readback.

Test elementAgreed working plan
ChangeAdd a short readback when a material research instruction changes
OwnerResearch lead coordinates the test; consultant reviews each readback
Initial sampleUse the next two suitable changes, if they occur in permitted assignments
EvidenceRevised instruction, readback, correction and any affected-work note
Review questionDid the step expose ambiguity before additional work, and what effort did it add?
Stop or reviseReconsider if it duplicates an existing effective control or adds confusion

Two cases can expose usability problems, but they cannot establish a universal improvement rate. If no suitable change occurs, the team records that the test has not run. It does not count the absence of an event as proof of success.

Use debriefing research with its scope intact

There is published research on structured reflection, but it should not become a made-up search-firm forecast. The primary abstract of Tannenbaum and Cerasoli's 2013 meta-analysis reports 46 samples involving 2,136 participants across varied debriefing settings.

That is broader evidence about debriefing, not an evaluation of this template, Laisy or a particular executive-search workflow. The abstract is the source used here; this guide does not claim a full review of every study's methods or applicability.

The practical implication is modest: take the structure and follow-through of reflection seriously, then inspect the result in your own context. Do not promise clients or colleagues a percentage performance gain from holding a meeting.

Keep any local comparison equally careful. If the team observes fewer corrections in a small sample, record the sample and circumstances. A descriptive count may justify another test; it does not establish that the retrospective caused a firm-wide outcome.

Make the follow-up small enough to inspect

Choose a change the team can actually apply. “Improve knowledge management” has no useful completion test. “Include the confirmed source date and unresolved question in the next suitable research handoff” is observable.

Name the owner, the context where the change applies and the person reviewing the result. Identify the old practice it replaces so the experiment does not quietly add duplicate work forever.

Use the client-update workflow if the change affects how a decision request is prepared. Use the consultant handover if it concerns transferring ongoing context. Link the experiment to the existing work rather than creating a parallel process no one maintains.

At the review point, decide to adopt, revise, stop or gather more evidence. Record the reason and tell the people whose work the decision changes. “We discussed it” is not the same as a decision about what the next assignment should do differently.

Atlassian’s DACI play distinguishes driving a decision from approving it. Apply that distinction when assigning the experiment: coordinating a trial does not automatically authorize a firm-wide process change.

If the learning note is maintained in Google Docs, version history can help identify revisions. Keep the reviewed result and its evidence identifiable; a later edit does not by itself show that the experiment succeeded.

Keep the learning note useful to the next colleague

Capture the process question, supporting events, chosen change and follow-up result. Keep unnecessary personal detail out of the general learning record. Preserve any relevant access boundary for the underlying evidence.

State when the lesson applies. A process useful for a multi-stakeholder assignment may be excessive for a simple clarification.

Include abandoned hypotheses where they prevent a recurring misunderstanding. For example, the team may discover that a notification was sent correctly but its wording did not explain the decision required. That distinction guides a different remedy than adding another notification channel.

Frequently asked questions

When should a search team hold a retrospective?

Choose a meaningful review point after relevant work or a recurring process problem. Urgent issues need their own immediate handling. A calendar ritual without evidence or follow-through is unlikely to produce a useful change.

Should the retrospective evaluate individual consultants?

Keep this method focused on the process question and observable work. Handle individual performance matters through the firm's appropriate process. Do not turn a shared learning note into an automatic ranking of colleagues or candidates.

How many improvements should the team choose?

Start with a change the team can own and inspect. The single-change approach here is a working suggestion, not a research-derived universal optimum. Additional proposals can remain a visible backlog.

Can the team claim the new process has improved performance?

Describe only what the evidence supports. Explain the sample, comparison and limits. A useful local observation is different from a causal claim or a promised result for every future assignment.