The Written Recommendation as an Owned Claim

The best test of a recommendation is whether the client can argue with it.

A chairman once called me two days after receiving a recommendation memo and opened, without pleasantries: "I disagree with your third reason." The third reason had claimed that the candidate's verified pattern of building successor depth was decisive for this mandate, given a leadership bench the memo had described, diplomatically, as thin. The chairman's objection was that the bench was thinner than we knew: several more senior departures were coming, unannounced, which meant the reason was more true than written, but the conditions were wrong: the candidate would need eighteen months of protection from exactly the succession pressure we had listed as her strength to deploy. We argued for forty minutes. The memo changed. The offer changed, a guarantee was added, a board expectation was reset, and the placement, now years old, held through them precisely because the disagreement had been possible.

Now imagine the alternative artifact, the one most of the industry ships: a deck, a spider chart, a match score of 87 percent. What, exactly, does a chairman call to disagree with? You cannot argue with an 87. You can only doubt it, silently and unfalsifiably and usually correctly. A score forecloses the conversation that makes placements durable; a reason invites it. That asymmetry is this chapter, and it is the last conversion this book performs: months of evidence, gates, and arithmetic become, at the end, a page of prose that a named person signs, this person, for this organization, under these conditions, for these reasons, with these uncertainties. Everything before was instrument. This is the claim.

The five elements

This person. Singularized, per Chapter 20's closing discipline: not "a strong transformation CFO profile" (the profile is what she is not) but this person, with this fork pattern, this risk map and its documented management, this optimizer meeting this world. The memo's first paragraph should be unwritable about anyone else on the shortlist; if it could be reused with the name swapped, it is a category description wearing a recommendation's clothes, and the singularization failed somewhere upstream.

This organization. The two-place predicate, completed in writing at last: the world's documented demands, from Chapter 3's scoping and not from the drifted version the process's enthusiasms produced, restated as the claim's second term. The memo argues person-to-mandate, and the mandate it argues to is the one the client signed at the beginning, which keeps everyone honest in both directions: a recommendation against a drifted mandate is unanswerable, and an undrifted mandate occasionally forces the memo to say that the role the client now wants is not the role that was scoped, which is Chapter 3's expensive kindness, delivered at the moment it costs most.

Under these conditions. The element most recommendations omit, and the one that converts a verdict into engineering. Conditions on the candidate: the guardrails the identity file specified, the licensed challenger, the development the dark-side conversation surfaced, the first-ninety-days asks. Conditions on the client, and here the memo earns its keep or loses its nerve: the decision rights that were promised must actually be granted (Chapter 20's autonomy trader will leave over a violated non-negotiable on schedule, and the memo says so, with the schedule); the board conduct the placement requires (Chapter 12's evidence runs both directions); the protection the chairman's phone call added. Half of placement durability is the client's behavior. Chapter 1's failure data said the failures are mostly relational and post-arrival, and a recommendation that ends at "hire her" has abandoned the placement at its riskiest moment. Conditions are the memo refusing to abandon it.

For these reasons. Three to five, numbered, each citing its evidence, the verified episode, the converged reference pattern, the enactment score, the fork analysis, in prose built to be challenged: specific enough to be wrong, grounded enough to be checked. The computed ranking is attached, labeled as computed, honored as the instrument it is, and it is never the argument. This is where the book's arithmetic and its ownership doctrine meet without contradiction, and where an emerging legal standard happens to converge on the same design: an accountable human rationale that does not merely restate the model's output is precisely what meaningful-oversight regimes now ask of consequential decisions, which means the memo is simultaneously the craft's best practice and the compliance artifact of the Appendix. Reasons, not scores, was always the professional standard; it is now also the regulatory direction of travel.

With these uncertainties. Named, specific, and few: the surviving H-cells and their verification plans; the attribution residue Chapter 3 admitted; the identity file's constitutive limits (inference from pattern, person in motion); the honest sentence about what only year one can teach. Stated uncertainty is not weakness in a recommendation; it is the signature of one built by someone who knows the difference between evidence and hope, and clients, who have read many confident memos and buried several of their subjects, know exactly which kind of author to trust. Boilerplate caveats do not qualify: "no assessment is perfect" protects the author; "we could not verify the 2021 attribution and the audit-chair call is pending Thursday" protects the client.

Craft

The memo is one page, with appendices. The claim appears whole in the first sentence, we recommend X for Y, conditional on Z, because a recommendation that must be excavated from its own document is hedging by structure. The reasons follow, numbered. The risk paragraph presents the risk map with its management, together, per Chapter 22's page-seven doctrine: a risk stated without its documented mitigation is an alarm; with it, it is due diligence. The conditions section is concrete to the point of discomfort. The uncertainties close the argument. And then the signature, a name, not a letterhead. The firm stands behind the memo; a person signs it; and the difference is the entire doctrine of this Part in typographic form: institutions cannot own claims, they can only honor the people who do. The signed memo is also what makes Chapter 25 possible, since an unowned file teaches no one anything at its second anniversary, because there is no one it can teach.

Adopted, credited, declined

The memo's lineage is acknowledged where it is owed. Success-profile-before-names — the discipline of specifying the role before meeting candidates, so the final argument runs against a stable target — is the Spencer Stuart tradition's contribution, installed in this book at Chapter 3 and closed here. Thesis-linked assessment — the private-equity habit of deriving leadership requirements from the deal thesis and arguing candidate-to-thesis — is adapted as mandate-thesis linkage: the memo's reasons trace to what this world must accomplish, not to general excellence. The scorecard-to-memo spine owes the ghSMART tradition its credit. And two industry conventions are declined by name. The match percentage — the single-number pseudo-precision that decorates so many search decks — is banned as exactly the score-without-grounds this chapter exists to replace. And the hedged non-recommendation — "all three finalists would be excellent choices" — is named for what it is: a courage failure invoiced as diplomacy. A shortlist without an argued first choice is not a judgment; it is a menu with a fee attached. When the evidence genuinely ties, and it happens, the memo says that, as a finding with reasons: "the capability evidence cannot separate A and B; the choice turns on which risk the board prefers to manage, as follows", which is an owned claim about a tie, and a different artifact entirely from three laminated maybes.

The hardest memo

Sometimes the owned claim is no. Do not hire from this shortlist; the market has been honestly searched and the mandate as scoped cannot be filled; the role must change, or the compensation, or the chairman's expectations, before this hire can succeed, and here is what must change, and here is what we found that proves it. The economics of contingent search punish this memo, which is precisely why writing it is the ownership test's final form: a firm that has never sent one has a fee model where its judgment should be. They are, in our experience, the memos clients remember longest and quote most often, the search that told them the truth at the cost of its own invoice, and they generate more mandates than any placement announcement the firm will ever publish. The willingness to write them is not a marketing strategy. It just, over a decade, behaves like one.

Proof-First, structurally

One paragraph on the institutional form, because it completes the doctrine rather than advertising it. Everything this chapter demands, the challengeable reasons and the named signature and the conditions and the stated uncertainties, presupposes that a recommendation is the kind of thing that can be tested. The Proof-First structure this book's publisher operates simply takes that presupposition commercially seriously: the validated shortlist and its argued recommendation are delivered before the main placement fee, so the client tests the claim, meets the evidence, probes the reasons, calls the chairman-shaped objections, while the testing still has commercial teeth. The fee follows the validation. Whatever firm a reader works with, the principle travels as one question, and it is the right question to ask any assessor, this book's authors included: what happens to you if this recommendation is wrong, and when would you find out? The first half of the answer reveals the incentive structure. The second half reveals whether there is a Chapter 25.

Red flags — in the memo itself

Adjectives without citations, "impressive," "seasoned," "high-caliber": Chapter 2's four tests apply to the final document most of all. The balanced-to-death memo, equal ink to every finalist, no argument anywhere: balance is a virtue in evidence-gathering and an abdication in recommendation. The naked yes, no conditions: a durability claim with its engineering omitted. Uncertainty theater, generic caveats in place of named unknowns. And the committee memo, the unsigned artifact of collective non-ownership, drafted so that no one is wrong if it is: the flag that all the others grow from.

Instrument: the memo template

Page one: ¶1 The claim, whole — we recommend [person] for [mandate], conditional on [headline conditions]. ¶2–4 The reasons, numbered (3–5), each citing file evidence. ¶5 The risk map with its management, together. ¶6 Conditions — on candidate, on client, on the first ninety days, concrete. ¶7 Uncertainties — named, specific, with verification plans. Signature block: name, date, role in the process. Appendices: the two-layer output (rank labeled as computed); the scorecard profiles; the identity file summary in register grammar; the verification log; the override log extract, if any.

Where the rules run out

The honesty paragraph. There is no validity literature on recommendation memos, since no meta-analysis compares memo formats to placement outcomes, so this chapter is craft doctrine, marked as such: built on Chapter 23's combination evidence, the accountable-rationale convergence, and years of watching which artifacts produced arguments and which produced silence. The courage economics are real and named: an argued first choice can be visibly wrong, which is the point and the price, since visible error is what Chapter 25 turns into calibration, while the hedged menu is never wrong and never learns. The conditions-on-client doctrine is commercially delicate, because some clients hear conditions as impertinence, and the memo's tone must earn the right the evidence gives it. And the memo inherits every limit upstream of it: it is the best claim the file supports, not more, which is exactly why it states its uncertainties, and exactly why it is signed. The signature does not certify that the author is right. It certifies that the author can be found, and asked, and held to reasons, which is everything this book has meant, from its first chapter, by judgment.

Notes and sources

Evidence grades: [M] meta-analytic/systematic; [L] peer-reviewed primary; [S] credible practitioner/survey; [T] flagged craft doctrine.

• Reasons-over-scores as accountable rationale; convergence with meaningful-oversight standards (rationale that does not restate the model): the regulatory synthesis underlying the Appendix (EU AI Act Art. 14 oversight posture; GDPR Art. 22 jurisprudence direction). [S]
• Success-profile-before-names: Spencer Stuart tradition; thesis-linked leadership assessment: PE management due-diligence practice; scorecard lineage: ghSMART/Who — all credited per the practitioner landscape research. [S]
• The combination arithmetic beneath the memo: Chapter 23's sources (Meehl; Grove; Kuncel; Dawes). [M]
• Post-placement relational failure concentration (why conditions exist): Chapter 1's sources (CCL; transition literature; CEO-departure analyses). [L/S]
• Stated uncertainty and advisor credibility: the calibration and trust literature adjacent to Chapter 14's sources. [L/T]
• The five-element form, the conditions-on-client doctrine, the no-memo, the signature doctrine: this book's synthesis from practice. [T]