A case I sat with for a couple of weeks in my HR course. I’m not going to retell all seventeen pages of it. I want to keep the part that has stayed with me since — because it turned out to be about something I do for a living.
Anyone who has shipped something into a market they don’t live in has met this problem. The thing works. It passed every test. Then it lands somewhere with different assumptions underneath it, and the same artifact means something completely different than it did at home.
The setup
In 1995 a new managing director arrived at Merck’s Argentine subsidiary with a mandate to modernize an insular company, and a five-year horizon to do it in. Within about a year he had done a genuinely impressive amount. He scrapped a four-director pyramid and replaced it with twelve directors organized around business units, on the theory that silos were the thing actually preventing the company from moving. He hired an HR director from outside the pharmaceutical industry entirely — unheard of in a market that only ever hired from within its own ranks — and she was the first woman at that level in the industry there. He took a woman who had started at the company as a secretary and made her the training and development director, which meant she now sat as an equal beside the man she used to work for. He put managers through two hundred hours of training. He introduced 360-degree feedback. He refused to pay physicians for prescriptions, which was ordinary practice in that market.
And he posted the open jobs on a board.
That last one is the detail I keep coming back to, because it’s the smallest thing on the list and it may have been the loudest. In a system where jobs were handed out as favours through a web of personal connections, a bulletin board isn’t a convenience. It takes patronage — the actual currency of the place — out of the hands of every manager who had been spending it. You cannot post that board and also be neutral.
So a year in, the change was real. Roughly a quarter of the workforce was new. The company had grown from about 175 people to over 200. And then the new culture got its first bill.
The Friday
The company had built a young professionals intern programme and run it properly. An outside consulting firm screened about 1,200 applicants down to roughly 70 interviews, then 30 finalists, who spent two weeks in exercises scored on a point system. The ranking was the output. Fifteen would be hired. The rules had been explained to the candidates in advance.
Two days before the names were announced, a middle manager phoned the training and development director to mention that one of the candidates was the son of a senior official in the country’s largest government health organization — the body that decided which drugs made it onto the national formulary. Having him inside the company, the caller said, would be an excellent way to get Merck’s products approved. When she pointed out that this candidate had come sixteenth, the caller got agitated and threatened to go over her head.
She called the managing director from the auditorium where the announcement was about to happen. He told her he’d get back to her. What he did instead was summon three people to his office — her, the HR director, and the sales director — and ask each of them, in front of the others, to make a recommendation and defend it.
That meeting is where the case stops and hands you the decision.
Why sixteen is genuinely tempting
I want to be honest about the pull, because I think a case is only useful if you feel it.
The obvious move isn’t to demote anyone. It’s to make the class sixteen. Nobody loses a seat. The fifteen who earned their places keep them. The candidate in question is objectively good — he beat fourteen other finalists — and the company gains a relationship with the single largest purchaser of pharmaceuticals in the country, worth something in the region of five million dollars a year in access.
And the business context was not comfortable. Local manufacturers held over half the market, sold pirated copies of Merck’s own drugs at thirty to fifty per cent less, and sometimes fielded three sales reps for every one of Merck’s. The country had just gone into recession. One of the company’s own reps summed it up: everything in the environment was against us. Meanwhile there was an aggressive growth target and a managing director whose job depended on hitting it.
Somebody in my class put the temptation better than I can. He said that when he first answered the poll, he’d rationalized his way to accepting the candidate without much trouble — expanding to sixteen compromised nothing, everyone wins. Then he went back and reread the company’s own list of values, and found he couldn’t make the rationalization survive contact with them. He described the whole thing as a reality check on his own practice: he had spent years telling leaders that some values are red lines never compromised for anything, and the first time he met a scenario that actually tested it, he found himself fine with compromising.
What expanding to 16 buys
Nobody visibly loses
The original fifteen keep their places. A strong candidate gets a shot. Roughly $5M a year in formulary access. A growth target that stops looking impossible in a recession.
What it actually costs
The cut line stops being a cut line
The cohort size changes after the results are known, for exactly one candidate, for a reason that has nothing to do with his score. Everyone learns the criteria are negotiable if you know who to call.
The room did not agree, and that was the useful part
I want to spend a section on this, because the most valuable thing about working a case with forty other people is discovering how many defensible answers there are to a question you thought was obvious. The discussion around this one ran for weeks, and the positions did not sort neatly into right and wrong.
The strictest reading was that the ranking is the answer and there is nothing further to discuss — fifteenth and sixteenth were close, but close is not a category the process recognises, and moving him up borders on corruption regardless of how it’s dressed. A softer version of the same conclusion argued for the original fifteen on grounds of consistency rather than purity: sticking to a published cut-off is the more defensible position both ethically and legally, and you don’t want to be explaining the exception later.
Then came the expansions, and they were not lazy. If the programme genuinely had the capacity to supervise one more intern, several people argued, growing the cohort to sixteen is the reasonable move — nobody is displaced, the candidate still has to meet every standard everyone else meets, and the organisation gets to be responsive without punishing anyone. One person went further and proposed expanding to twenty, on the logic that this dilutes the favour rather than granting it: the well-connected candidate stops being the sole beneficiary, four other near-miss candidates get a place they’d otherwise have lost, and the extra revenue would comfortably fund all five stipends. He was candid that it still sat in a grey zone. He also thought it was a win with nobody corrupted.
There was a reading I found genuinely hard to dismiss, from someone who does organisational change for a living, which refused to answer in the abstract at all and made the decision contingent on the numbers. If you can show a credible path to the growth target without the government relationship, hold at fifteen and call it Integrity. If missing that target would do real damage — and the company had just absorbed an enormous amount of leadership churn and needed time to reach a steady state — then expand to sixteen and call it Persistence. Both were on the company’s own list of values. Her closing point is the one that stuck: a healthy culture on a leadership team makes no difference if the business doesn’t survive, or if the managing director is fired for missing his numbers.
And somebody made the smallest, most deflating argument of all, which is that sixteenth out of thirty in a brutal two-week assessment means you did extremely well and missed by a hair — and if any of the fifteen had declined, he’d have been called anyway. From that angle the whole thing is a nothingburger being treated as a morality play.
Positions argued in the room — all of them by people with reasons
Hold at 15
The ranking is the answer. Close isn't a category the process recognises. Moving him borders on the exact thing the company is trying to leave behind.
Expand to 16
If there's real capacity to supervise one more, nobody is displaced and he still meets every standard. Responsive without being punitive.
Expand to 20
Dilute the favour instead of granting it. Four other near-misses benefit too, and the new revenue pays all five stipends. Grey zone, admitted.
It depends on the numbers
Hit the target without them? Hold, and call it Integrity. Miss it damagingly? Expand, and call it Persistence. Both words were on the company's own list.
It's a nothingburger
16th of 30 means he missed by a hair. If one of the fifteen declined, he'd be called anyway. We are staging a morality play about a waiting list.
I honestly don't know
Said out loud by more than one person, including some who had made this exact call at work. The most credible answer in the room, arguably.
Two arguments cut across all of those and changed how I think about the decision rather than just where I landed on it.
The first was a test rather than a position: would we be comfortable if this decision appeared on the front page of a major newspaper? It’s a blunt instrument, and it does no philosophical work at all, but it has the advantage of being answerable in about four seconds — and it quietly relocates the question from what we can justify to what we’d be willing to have seen.
The second was about procedural justice, and it’s the one that dismantled the expand-to-sixteen option for me even though I’d found that option reasonable. Changing a process after the outcomes are known undermines the perception of neutrality even when every standard stays identical. Consistent application of a rule turns out to matter more for trust than a favourable outcome does. Which means the expansion doesn’t fail because it’s unfair to anyone in particular — it may genuinely be unfair to no one — it fails because of what it teaches everyone watching about when the rules apply.
And people were watching. Somebody made the point that on the surface this is one internship slot, but in practice it is the moment when a company that has spent a year talking about transparency finds out whether the new culture survives contact with real pressure. She told a story about pushing back on a hire at her own firm — a candidate a colleague went skiing with, who simply didn’t meet the bar — and how quietly trust would have eroded had she not. That is the actual stake. Not the intern. The audience.
The part my job made obvious
Here is where I stopped reading it as an HR problem.
The cohort was fifteen before anyone was scored. Changing it to sixteen after the ranking exists is not generosity. It is editing the acceptance criteria after seeing the results — and it’s being done for one candidate, on a basis that has nothing to do with the criteria.
I have watched that exact move in software, more than once, and it never announces itself as cheating. It arrives sounding reasonable. The threshold was always a bit arbitrary. We’re not weakening anything, we’re just widening slightly. This particular failure is a special case. And it is always, always cheaper in the moment than the alternative. A test you’re willing to rerun until it passes has not passed. It has just stopped being a test.
The clean version of the question isn’t does this candidate deserve a place. It’s what would have happened if he had ranked twenty-fourth? If the answer is that nobody would have called, then the call is the variable being tested, not the candidate. Which means the programme is no longer measuring what it says on the tin, and that’s true whether or not this particular young man would have made a fine intern.
And the reason consistency matters more than the outcome here is that nobody outside the room can see intentions. They can only see the rule and what happened to it. A process that bends once, for a good reason, in private, is indistinguishable from a process that was never real.
The thin line
Now the part I actually wrote about at the time, and the part I’d defend hardest.
The managing director in this case did something admirable and something worrying with the same instrument. He used supportive supervision to demolish an old boys’ club — he flattened the hierarchy, brought in outsiders, posted the jobs, and put serious money into training so that managers might become leaders rather than bosses. That support was real, and it made transparency feel like a practice rather than a poster.
But the same supervision that empowered people also frightened them. He fired the blockers. He put managers through unusually blunt 360-degree feedback. He asked his leadership team who is with me and who is against me. He moved fast enough that people described the experience as traumatic. He was, in effect, asking for candour while demonstrating what happened to people who were candid in the wrong direction — which is a hard thing to ask for and undercut in the same quarter.
And then the decisive detail. When the hard call finally arrived, it went to his office. Three senior people were summoned to give an opinion, and the decision still sat with him.
I was not alone in this reading, but I was also not obviously right, and the counter-arguments were good. Someone who runs change management professionally listed what she thought was missing: no stakeholder analysis, no identified change champions, no ownership handed to the likely resistors so they had a stake in the how, no clear case for change that answered what any of it meant for an individual contributor. Removing the leaders who would have poisoned the effort was, she thought, done well — but the equally important work of bringing the rest along wasn’t, and behavioural change of that magnitude carries a real risk of a productivity dip nobody had planned for. Her sharpest line was that a leader who champions open feedback while firing everyone who can’t get on board fast enough has not built a place where anyone feels safe doing the exact thing he’s asking for.
The rebuttal was just as strong, and came from someone who has run transformations. The company was chasing aggressive growth in an inflationary economy against competitors with three times the sales force. In that situation you either call for decisive action or you concede the objectives up front. The verbalisation — who is with me and who is against me — was clumsy. The urgency wasn’t wrong. As another classmate put it, the tree had to be shaken hard.
The disagreement I found most interesting was about the thing that wasn’t there at all. A classmate with fifteen years in pharma pointed out that for a company operating where corruption is normalised, the absence of any ethics and compliance function is the striking omission — an ethical decision pathway isn’t a nicety in that industry, it’s part of the licence to operate, and a robust one would have resolved this dilemma before it ever reached a managing director’s office on a Friday. That’s the argument I answered at the time, and I’d make my answer again, but three other people offered defences of the omission that I think are worth more than mine. One was that in a deliberately flat organisation, a compliance department cuts against the model — better to build ethics measures into functional KPIs and the performance review structure than to create a box that owns them. Another was that the omission may have been the point: make ethics everyone’s, embedded across every function, rather than a department other departments can outsource it to. The honest counter to both, which someone raised immediately, is that anything designated everyone’s job has a well-documented habit of becoming no one’s.
I don’t think that’s hypocrisy, and I’d defend him further than most of my classmates would. There’s a reasonable argument that standing up a formal ethics and compliance function in year one would have undermined his own authority at exactly the moment he needed it, slowed the transformation, and produced findings that neither the subsidiary nor the parent company could then afford to ignore. Sequencing is a real constraint. He had come from a regional role at headquarters, so he knew roughly what he was walking into; deferring the compliance apparatus may well have been deliberate rather than careless.
But the cost of that sequencing shows up precisely here. The training and development director was doing everything the new culture asked of her. She had a documented process, a defensible ranking, and a manager applying pressure to override it. And there was nowhere for her to take that except upward, to one person, informally, by phone, on the morning of the announcement. A culture that has replaced ask your patron with ask the managing director has improved the answer without changing the shape of the question.
That’s the localization failure, stated properly. It isn’t that the policy was wrong for Argentina — the policy was right, and brave, and the ethical position on incentives was harder there than it would have been anywhere else. It’s that the policy was localized and the system around it wasn’t. A default shipped without a local escalation path isn’t localized. It’s translated.
To change a culture, change the systems
The line I kept returning to during the discussion was one I’d argue for outside this case entirely: culture cannot change on its own. How we work is made of systems, created or maintained, intentionally or otherwise — and those systems are what shape the culture, not the other way round. Which means “culture” is a poor target. It’s a lagging indicator. Aim at it directly and you get posters.
A classmate said the same thing from the other end, and better: culture change fails when it is announced rather than rebuilt through the systems that actually govern behaviour. Another added the specific mechanism — the change sticks only when hiring, promotion, incentives and performance evaluation all reinforce the same expectation, because people learn very quickly which behaviours are genuinely rewarded and adjust to that rather than to the stated values. She also made a point I hadn’t considered: framing ethics as the core of who the company is, rather than as rule enforcement, is what makes it safe to raise a concern. Enforcement produces compliance. Identity produces the person who speaks up.
So the practical version of my position isn’t “have better values.” It’s that the leadership should have been running periodic metrics against these systems the way you’d run a regression suite — because a system that isn’t measured has already started drifting, and you find out in the form of a phone call on a Friday morning. Otherwise you get compliance on paper and not commitment in practice, which is a distinction somebody in that thread drew and I have quietly stolen.
Where I land
You cannot create a culture by compromises.
That’s the sentence I’d keep if I had to throw the rest away. Not because integrity is its own reward — this case is a poor argument for that, since we know the reforms didn’t fully hit their targets — but for a duller structural reason. A culture is not made of its values statement. It’s made of what survived the first time honouring it was expensive. Every exception you grant quietly becomes the actual policy, and the written one becomes decoration. There is no version where you bend it once and it stays intact, because the bend is the information everybody else acts on afterwards.
So: hold at fifteen. Not because the sixteenth candidate is unworthy — he isn’t — but because the moment the size of the class depends on who a candidate’s father is, the programme is measuring something other than what it claims to, and every future participant is right to assume so.
What I argued at the time, and still would, is stronger than that: any proposal other than declining the sixteenth candidate deviates from the culture they were trying to build, and is a conflict of interest whichever number you land on. Sixteen and twenty are not compromises between the two positions. They are the same decision with better manners — because the reason the cohort grew is unchanged, and that reason is the phone call.
I’ll concede the strongest objection honestly, though, because it deserves better than a footnote. If the business genuinely could not survive without that relationship, then a pristine culture inside a company that fails is not a moral victory, and I don’t think the people arguing that were being cynical. My answer is only that this is an argument for having found the growth somewhere else, much earlier — not for spending the culture on the first occasion it became expensive. But I hold that one less tightly than the rest.
The part that isn’t abstract for me
There’s a reason this case didn’t stay academic for me, and I’d rather say it plainly than dress it up.
I have spent a lot of my working life as the person in the room without the local connections — without the shared school, the shared accent, the shared shorthand that gets you a phone call made on your behalf. If you are that person, a documented, published, scored process is not bureaucratic overhead. It is the entire door. It is the only mechanism in the building that doesn’t require you to already be known.
That’s why the phrase it’s just one exception lands so differently depending on where you’re standing. From inside the network, it’s a small courtesy. From outside it, it’s the discovery that the posted rules were the visible interface and the real system was running underneath the whole time.
Fifteen people earned places that week under rules that were written down in advance and applied to everyone. That is a rarer thing than it sounds, and it was worth more than five million dollars.
Thank you for reading this far. If you’ve been on either side of this — the person asked to make the small exception, or the person who found out afterwards that one had been made — I’d like to know what it did to your trust in the process, and whether it ever fully came back.
Comments