First who, then what — and the bill for replacing them

An electrical materials distributor in Brazil, 180 people, seven branches (figures in Brazilian reais). The purchasing coordinator resigned in March. She had eight years with the company, knew the real lead time of every supplier, knew which orders could wait and which ones would stall a client's job site. None of that was written down.

The position stayed open for 74 days. The person hired took four more months to negotiate without checking with someone else first. Along the way, two suppliers adjusted their terms for the worse because no one followed up on the right day, and management concluded the purchasing team 'was underperforming.' Payroll, during those 74 days, had gotten cheaper.

01The idea: first who, then what #

Jim Collins published Good to Great in 2001, based on a study comparing American companies that sustained superior performance for fifteen years with similar companies that did not. One of the findings he highlights is the order of decisions: the teams that stood out reportedly decided first who was on the team, and only then the direction to follow.

The bus metaphor — getting the right people on, the wrong people off, and each one in the right seat before choosing the destination — became the most quoted and most misused part of the book. It is often read as license to fire quickly. The reading truer to the argument is different: who decides is a prior and more decisive question than what to decide, and it deserves the same care given to a capital investment.

02The criticism the method received #

Good to Great is a well-known case of selecting on the dependent variable: the companies were chosen because they did well, and only then did researchers look for what they had in common. That design cannot establish causation — traits also present in companies that failed would not show up in the sample.

Phil Rosenzweig developed that argument in The Halo Effect (2007), showing how known performance contaminates the description of a company's attributes. The criticism became more concrete over time: some of the companies celebrated in the book went through severe difficulties in the following years. And the literature on manager effects — Henderson and colleagues, in 2012 — indicates individual contribution exists, but requires long series to separate from chance.

None of this invalidates the useful part, which doesn't depend on the study: the decision of who does the work carries large economic consequences and is made with less rigor than an equipment purchase of the same value. That part can be checked in your own company, with arithmetic.

03The math of one replacement #

Scale of the 180-person distributor, a coordination role with a R$ 12 thousand salary and R$ 19 thousand loaded cost per month. Stated assumptions for you to redo:

Want to see how this looks inside a real operation? Explore the platform.

The math, item by item: 74 days of an open position covered by colleagues, with an estimated 40% loss in the role's output, comes to around R$ 19 thousand; hiring, between management hours, interviews, and postings, R$ 28 thousand; a five-month ramp at 50% average productivity costs R$ 47.5 thousand; negotiation mistakes and missed deadlines during the period, conservatively estimated at R$ 48 thousand. Total close to R$ 143 thousand — roughly seven and a half times the role's monthly loaded cost.

Worth saying what this math is not: an argument against letting go of someone who needs to go. Keeping someone in the wrong seat has its own cost, and a bigger one. What the math does is give size to a decision usually treated as administrative.

04Where this breaks down in practice #

1. The cost shows up in the wrong place. The payroll savings go to the month's bottom line; the loss shows up scattered across delays and rework in another area.
2. The role only opens after the departure. Without visibility into overload, the company discovers the need on the day of the resignation letter.
3. Knowledge walks out the door. A supplier's real lead time, a client exception, the reason behind a rule: all of it lived in one person's head.
4. Urgency hires the next departure. A hire made to plug a hole tends to repeat the cycle within a year.
5. The ramp isn't planned. No one defines what the person needs to know in weeks 1, 4, and 12 — so they find out alone, through mistakes.

05What data-driven management answers #

The expensive part of a replacement isn't finding someone. It's the knowledge that only existed in one person's practice, and the time it takes the next person to rebuild it. That's a documentation problem before it's a recruiting problem — the same mechanism as in what only Paulo knows.

Retention also depends on what we discussed in a raise doesn't fix engagement: without context and without follow-up, a salary correction just postpones the problem. And it's worth looking at the bottleneck before hiring, as in improving everything is improving nothing — many open positions are a response to a stuck workflow, not a lack of people.

On the platform, this is the people, process, and knowledge modules, with ownership, workload, and decision history recorded within the work itself. Solutions by area show the specific angle, and a conversation helps estimate this with your own numbers.

Before the next hire, it's worth writing down the cost of the last one. It's usually bigger than the salary discussion that preceded it.