The Eighteen-Month Bounce

How an industry lost half its most consequential hires without ever measuring it

The number nobody checked

The figure everyone quotes is 40%: senior hires pushed out, failing or quitting within eighteen months. It comes from Kevin Kelly, then chief executive of Heidrick & Struggles, describing an internal review of roughly 20,000 searches to the Financial Times in March 2009.¹ There is no published methodology behind it, no independent replication, and nothing more recent in the seventeen years since.

Around it circulates a cloud of larger numbers, 50%, 70%, attributed to nobody in particular. Some sources now describe the original review as one of the largest datasets on executive hire outcomes ever assembled, and claim it is corroborated by the Center for Creative Leadership, The Conference Board, Egon Zehnder, Deloitte and McKinsey. Those claims travel without citations. I have not been able to trace any of them to a parent study.²

Notice what that means before moving on. An industry that loses something close to half of its most consequential hires has never produced a properly sourced measurement of it. Not a contested measurement, or a disputed one, or a range that different methodologies produce differently. There is one remark from one executive to one newspaper, and everything since is that remark being passed along.

That is not a gap in the literature. It is the first piece of evidence, and this piece hangs from it.

Most senior operators have lived through the version that doesn't need a statistic. A Director-level search runs, a shortlist comes through, the strongest candidate gets the offer, the offer gets accepted. The first six months look fine: the new hire is learning the function, building relationships, asking reasonable questions. The next six months look mostly fine, with a few small concerns. By month eighteen something is clearly wrong, and the wrongness is structural rather than fixable. The team has reshaped itself in the new Director's image. Two of the strongest people who could have been the function's succession bench have left or disengaged. Decisions made in months four through ten have hardened into commitments that will take years to unwind.

The pattern is common enough to be a cliché in executive search, and the cliché is close to all anyone can say about it with confidence. The frequency is folklore. The diagnosis doesn't exist.

A failure this expensive stays unmeasured only where nothing in the process is capable of detecting it. That is the claim this piece is making: the absence of a number and the persistence of the failure have the same cause, and it is not incompetence.

Why the failure is invisible

The cost of a bad executive hire is not an unstudied subject. Search firms publish estimates constantly: 30% of salary at entry level rising past 200% at executive level, total cost of two to three times annual salary, occasionally much wilder multiples. All of that treats the problem as a bigger number of the same kind. It isn't. It is a different shape.

A good Director hire produces good outcomes at a roughly linear rate: the function performs, the team grows, the strategy executes. A bad Director hire does not produce bad outcomes at the same rate in the opposite direction. It produces compounding, distributed damage. They hire in their own image for eighteen months before anyone notices the pattern. They lose the people on the team who would have been the actual succession bench. They make architectural or strategic commitments that take years to unwind. They set norms (how meetings run, how disagreement is handled, whether the function tells the truth upward) that persist long after they themselves have gone.

The shadow keeps falling after the source has left. That is the part that matters, because it is what breaks the measurement. By the time the damage is visible, the failed hire has been entangled with dozens of downstream choices that all carry their own surface explanations. Nobody experiences a discrete loss they can trace back to a search. They experience a general sense that the organisation has got worse. In the searches I have watched play out, the decisions outlast the decision-maker by two to three years, and by then the original choice is no longer reviewable as an event.

So the feedback loop is broken at the diagnosis step, not the data-collection step. The absence of a real number is not an oversight by an otherwise functional industry. It is what an undiagnosable failure looks like from the outside. Every party can see their own portion of the outcome. Nobody can see the whole of it, and the parties best positioned to assemble the picture are the ones with the least reason to.

None of which is a secret inside the profession. The gap between what recruitment measures (time to fill, cost per hire) and what it would need to measure (whether the hire was any good, a year and a half later) is discussed constantly, and quality of hire has been called the holy grail of recruiting for long enough that the phrase has worn out.³ But that discussion treats the gap as a measurement problem waiting for a better instrument. The measurement is not missing because the tools are inadequate. It is missing because nothing in the arrangement requires it.

One clarification before continuing. Director-level work is the most common case where this shows up, which is why this piece focuses on it. But the underlying variable is asymmetric impact, not job title. The same logic applies to any hire whose downside compounds rather than accumulates: the first hire into a newly formed function, the replacement of someone foundational to how a team operates, specialist hires whose decisions shape years of downstream architecture, senior individual contributors whose judgment sets the standard for a discipline. Anywhere the cost function bends sharply, throughput is the wrong thing to optimise for.

What the document selects for

If the failure can't be diagnosed after the fact, the only place to look is before it. Which means looking at the document at the top of the process, because that document does more selection work than anything downstream of it.

Job descriptions at Director level are almost universally written as person descriptions: skills, experience requirements, years in role, technical competencies, responsibilities. They describe the company's ideal candidate. They do not describe the ideal candidate's ideal role. The distinction between a job description (an internal artefact of role documentation) and a job advertisement (an external artefact of candidate attraction) is well established in the recruiting literature, and so is the advice that follows from it. What is not examined is the consequence at Director level, where the distinction stops being a question of marketing craft and becomes a question of which candidates the company is structurally selecting for.

A person description is read by two kinds of candidates in two completely different ways.

The candidate who likes their current job scans it to see whether this role is better than what they have. They are looking for the dimensions that would justify a move: scope, scale, the shape of the problem, where doing it well would take them next.

The candidate who dislikes their current job scans it to see whether their skills and experience map to the requirements. They are looking for an exit hatch.

Person descriptions speak almost exclusively to the second reader. A list of requirements is exactly what you scan when you are looking for somewhere that will have you. It is close to useless to someone deciding whether to give up a job they are doing well. The first reader, who is the candidate the company actually wants, closes the tab.

This is the most important point in the diagnosis and it deserves to be said directly. There is nothing wrong with the candidates who dislike their current job. Many of them are excellent. They will accept the offer if it is extended, and they may well perform. But their intent at the moment of decision is to leave their current situation rather than to take on this one, and that intent does not sustain through the hard parts of a Director role at month fourteen. When the function turns out to be more broken than advertised, when the political problem surfaces, when the first real decision has to be made against resistance, the person who moved to get away from something has nothing to draw on. They did not choose this. They chose not-that.

So the company has selected, without meaning to, for people whose primary motivation at offer was escape. Nothing in the interview process detects this, because the escape-motivated candidate answers "why are you interested in this role" perfectly well. They have to. Everyone has learned to. The motivation is not concealed so much as unasked-about, and the document that produced the shortlist already did the filtering before anyone got in a room.

That is the eighteen-month bounce, a great deal of the time. Not a capability failure. A motivation that ran out.

The same error has three other surfaces, and in most companies a senior executive has signed off on all of them personally.

The first is the employer brand: the function-level brand, the EVP, the mission statement recited in offer conversations. Most of this material declares the company's significance, its culture, and its impact on the world, in a register designed to make the candidate feel they are joining something meaningful. Strong candidates read it the way they read job descriptions, as evidence about whether real work has happened. A company that has genuinely worked out what it is, what it isn't, and what it is like to operate inside doesn't declare those things in the abstract. It shows them in the specificity of the role being offered, in the questions the hiring principal asks, and in the questions they encourage the candidate to ask back. Brand language floating free of that texture reads as performance, and performance is a reliable signal that the underlying work hasn't been done.

The second is the assumption that the role is static. Most job descriptions describe a snapshot: what the role looks like today, what the responsibilities are, how the team is structured. Director-level roles are not snapshots. The function is growing, the strategic context is moving, and a serious Director will reshape the role around their own judgment within the first nine months. Candidates worth hiring know this and are checking whether there is room for them to do it. A static description tells them there isn't. Agency language is either missing or, more deceptively, present in the form of "what you will achieve in your first six months", which is not an offer of agency at all but an instruction with a deadline attached. The grammar mimics empowerment; the semantics deliver compliance. Strong candidates read past the surface form. The document is telling them that someone has already decided what success looks like here, and their job is to deliver against that decision. That is a step backwards for a Director who would expect to define success themselves in the first ninety days. The candidates who respond happily to that framing are the ones without the instinct to push the role past its initial scope, which is the instinct the company most needed when they wrote the spec.

The third is the application form, where the filtering carries on after the document has finished. One of the better-funded AI companies asks four things of anyone applying to any role: a link to a LinkedIn profile, what makes them excited about the company's mission, a hard problem from their past, and a recent achievement. Read that against the two readers. The candidate looking for an exit completes it without difficulty, because enthusiasm for a mission is available to anyone who has read the homepage and costs nothing to supply. The candidate deciding whether this is better than what they already have is being asked to write down why they would move, before being told anything that would make the answer true. One of them writes three paragraphs. The other has nothing to put in the box, and it is not the box that is at fault. What the person description began, the form completes, and both are done before anyone has met.

Four surfaces, one error. The company is declaring what it is and filtering for who matches, when the work requires articulating what a particular kind of person could do with the role.

Push and pull, meaning something else

The operational expression of that difference is usually called push and pull, and I use both words in something close to the opposite of their received sense. That is deliberate, not a mistake.

In common usage, push describes what the employer does outbound: job boards, campaigns, cold approaches. Pull describes candidates arriving of their own accord, drawn by brand and reputation. Both terms sit with the employer, and describe the direction of the employer's effort.

The useful version locates the force on the candidate instead, because that is where it actually operates. A candidate who responds to a generic job description and a templated approach is being pushed: out of their current job, by frustration or stagnation or a manager they can't stand, towards whatever exit will accept them. Person descriptions only interest people in that state. A candidate worth hiring at Director level is not being pushed by anything. They are doing well where they are. If they move, they have to be drawn: pulled towards a specific opportunity described in terms that make it recognisable as an upgrade for them in particular.

Read that way, push recruitment is not a less thoughtful version of pull. It is structurally only able to reach candidates whose own circumstances are generating their willingness to engage with a weak signal. The pipeline runs on the exhaust of bad management elsewhere. It depends on a steady supply of people being pushed out of their situations, and captures them as they leave.

That supply is real and roughly continuous, which is the crucial point. It means the model appears to work. The candidates arrive, the roles get filled, the invoices clear, and the bounce happens far enough downstream to be attributed to something else. A process can run on a structural error indefinitely as long as the error's cost lands on someone who wasn't watching.

Why nobody runs the other process

The obvious question is why, if the upstream work is what produces good outcomes, so little of it happens. The answer is economics rather than effort, and it is the same answer three times over.

Volume recruiting works well where errors are cheap and roughly symmetric, which describes most hiring. Throughput is the right objective when each decision carries a similar magnitude of consequence. But the unit economics (recruiters paid on placements, contingency or near-contingency fees, search velocity as the primary measure) are actively misaligned with asymmetric work. The marginal cost of "we don't have the right person, this search continues another six months" falls on the firm. A firm that runs a genuinely low-throughput process breaks its own business. This is not a failure of care. It is an incompatibility between the cost function of the work and the business model of the firm doing it.

Retained executive search has the opposite problem. The fee structures support patience, and the senior consultants involved often do excellent work at VP level and above, where the firms are positioned. Director-level roles fall into an awkward middle: too senior for internal TA or contingent agencies to serve well, not senior enough to justify the fee and process intensity of a retained firm whose partners are running five searches at a time. The retained model produces good outcomes when it engages. It mostly doesn't engage, and when it does, the cost structure pushes execution down to associates running the same person-spec process as everywhere else.

Internal TA is running the volume playbook for the same structural reason: the headcount, the tooling and the metrics are all built around throughput, which is right for IC and senior-manager hiring and wrong here. Asking that function to be more careful, or more diligent, or better at writing job descriptions will not change the outcome, because the constraint is the design rather than the effort. The people running these searches take most of the blame when the bounce happens, and almost none of it is theirs to take.

What is left, in practice, is internal TA running Director searches with the senior-manager playbook, contingent agencies hosing the hiring executive with CVs, and the CRO or CTO or CMO doing most of the real evaluation themselves on top of their day job.

The candidate-side failure and the client-side mispricing are therefore one fact with two faces. Volume firms can't price the work because they can't reach the people; they can't reach the people because they can't run the process that would price the work. A model that genuinely served either party would have to serve both, because the work that produces a good outcome for the company is the same work that produces a real opportunity for the candidate. The two sides are not in tension. They are being failed together, by the function meant to connect them.

This is not an unstudied structure, although it is a thinly studied one. The clearest treatment I have found analyses the client, the headhunter and the candidate as a triad governed by information asymmetry, moral hazard and adverse selection, and arrives at a version of the same puzzle from the other direction: that headhunters assess candidates reasonably well and match them to clients badly.⁴ It explains the gap through what each party knows and withholds. This piece explains it through what the document at the top of the process selects for. The two accounts are compatible. The striking thing is how few of them there are.

The work that isn't happening

It is worth describing what the missing work would consist of, if only to see how little of it the current arrangement can afford.

It starts with the hiring principal being able to say what the role is for rather than what tasks it covers: what scope it has, what scale of impact it carries, what doing it well will let the hire go on to do, what recognition follows from success, and what reward is on offer.

Those five are not arbitrary. They are what a candidate who is not unhappy where they are is scanning for, in roughly that order. They are the dimensions along which a move registers as an upgrade to someone who does not need to move.

Most companies cannot answer them on demand. They can answer the surface version (the title, the team size, the comp band) but not the substantive version of any of them without a conversation that surfaces what the role is really for, what the function is really becoming, and what someone doing this well for three years will be in a position to do that they cannot do today. That conversation takes a senior person, several hours, enough trust for the principal to say things that can't appear in a public document, and a willingness to discover in front of someone else that they hadn't worked out the answers.

Everything downstream depends on it. The materials come out of the conversation, the shortlist comes out of the materials, and the rest is execution. When it happens, the advertisement is written for the candidate who is doing well and open to better: it leads with scope, scale and impact rather than requirements, and it is specific enough that a candidate can tell whether this is an upgrade for them in particular rather than for an abstract Director-of-X. The first call can go somewhere, because whoever is making it knows what is hard about the role. The people who engage do so because the articulation drew them.

And the conversation itself produces something the company did not have before, independent of whether the search succeeds: a statement of what the role is for, at a level of specificity that survives contact with a sceptical candidate.

Now notice what that requires, and who is positioned to fund it. It is unbilled work performed before a single candidate is contacted. Under contingency, it is unbilled work that may never be billed at all. Under an internal TA function measured on time-to-fill, it is time spent not filling. Under a retained structure, it is partner hours on a role whose fee doesn't carry partner hours. Every arrangement in the market prices this work at close to zero, and every arrangement is right to, by its own logic. The conversation isn't skipped because people don't know it matters. It is skipped because nothing in the system pays for it, and the cost of skipping it arrives two years later, attributed to something else.

What this argument can't show

The honest position on all of this is that I am arguing from pattern rather than from data, and the piece would be weaker if it pretended otherwise.

I can't demonstrate that escape motivation is the largest cause of the eighteen-month bounce. I can say that it is a mechanism nothing in the standard process would detect, that it is consistent with what I have watched happen, and that its absence from the literature is at least partly explained by the same undiagnosability the rest of this piece describes. That is an argument for taking it seriously. It is not proof.

Nor can I show that better upstream work produces measurably better retention at eighteen months, because that study would require the thing the industry has never done: tracking senior hires against the process that produced them, over a horizon longer than anyone's commercial interest in the answer. The absence of that study is the piece's central complaint and also its central limitation. I am in the same position as everyone else, arguing about a failure rate nobody has measured properly.

What would change my mind is straightforward to specify, which is worth something. A dataset of Director-level hires tracked to twenty-four months, coded for how the role was articulated at the point of first contact and for what the hire said their reason for moving was, would settle most of this. If motivation at offer turned out not to predict survival at month eighteen, the argument fails. Nobody is collecting that data. That is the problem in miniature.

[Optional: one observed example here, framed as illustration rather than proof. Most useful if it is a search where the hiring principal could not articulate one of the five dimensions at intake, and where the gap in the document is visible alongside who did and didn't respond to it. Weakest if it reads as a success story.]

Why it persists

The failure survives because every party's local incentives are satisfied at the moment it occurs.

The company fills the role, which was the thing it set out to do. The recruiter gets paid, on a timeline that closes well before any outcome is known. The hiring executive gets a Director and stops carrying the function themselves. The candidate gets a title, a rise, and an exit from a situation they wanted out of. Every one of those is a good outcome as experienced by the person having it. Nobody in the transaction feels a failure, because at that moment there isn't one.

The failure arrives eighteen months later and lands on people who were not in the room: the two strong performers who left, the team that reshaped itself around the wrong judgment, the executive two levels up who inherits commitments they can't unwind and no way to trace them back. By then the person who made the decision has often moved on themselves, and the firm that ran the search has closed forty more since.

For this to change, someone would have to be accountable at month eighteen for a decision made at month zero, and nothing in the current arrangement creates that link. Not the fee structure, which resolves at offer. Not the internal metrics, which resolve at fill. Not the vendor relationship, which is typically re-tendered on price before the outcome of the last search is legible. The commercial memory of a senior search is shorter than the failure it produces, and until that changes the failure will keep being paid for and keep not being seen.

Which returns us to the number. Seventeen years of an industry quoting a single unmethodologied remark about its own failure rate is not laziness, and it is not a scandal either. It is what happens when nobody in a market has both the standing and the incentive to measure the thing. The figure survives because it is unfalsifiable in practice and useful in argument, which describes most of what circulates as knowledge in this field.

The eighteen-month bounce is not evidence that senior hiring is hard. Senior hiring is hard. It is evidence that a market can run for decades on a process that reliably destroys value, provided the destruction is slow enough, distributed enough, and arrives late enough to be blamed on something else.