How to De-risk Management Promotions with YouMight

1. We promote based on the wrong data, and it’s been proven.

The Peter Principle “every employee tends to rise to their level of incompetence” was long treated as a coffee-machine joke. Benson, Li & Shue (2019) tested it using real-world data: 38,843 salespeople, 131 companies, and 1,553 promotions into management positions.

Two findings:

• First, companies do indeed promote based on individual performance. Being the top salesperson on your team roughly triples your probability of being promoted the following month.

• Second, individual performance negatively predicts managerial performance. Promoting a high-performing employee into a management role reduces the overall performance of their subordinates.

The best expert is, on average, a worse manager.

Not because they are bad, but because excelling at a job does not mean being good at managing in that same field.

And yet we keep doing it. Not out of stupidity, but because of a lack of evidence. We make decisions based on the data we have (individual performance, measured and indisputable in a promotion committee), because we lack the data we actually need (managerial behavior, which by definition does not yet exist).

A management promotion is a decision made using data about a different role that is not representative of the new one.

2. But how much does a failed promotion cost?

Take a frontline manager: seven people on the team, ~€65k in fully loaded salary, with a team averaging ~€55k fully loaded per employee. The promotion doesn't work out. Over 12 to 18 months, we can add up:

• Manager departure and replacement (recruitment, vacancy, onboarding and ramp-up of the replacement): 6 to 9 months of fully loaded salary

→ €33k to €49k

• Team underperformance during the period of disruption: 7 people × €55k × 10 to 20% × 9 months

→ €29k to €58k

• Induced turnover: 1 to 2 departures, each replaced at a cost of ~6 months of fully loaded salary (recruitment, onboarding, inefficiency before leaving, customer impact, etc.)

→ €28k to €55k

That gives us an order of magnitude of €90k to €160k per failed promotion.

Now think about how many management promotions over the past three years have ended in someone leaving, a quiet demotion, or a damaged team. Multiply that by €100k (a reasonable average cost to use).

This cost doesn't appear in any budget because it doesn't have anyone's name on it.

3. Why “selecting better” isn't enough.

The natural reaction: if the problem comes from selection, let's improve selection. Assessment centers, personality tests, longer and longer interviews.

Except that the ceiling is known, and it's low. Sackett, Zhang, Berry & Lievens (2022) revisited all estimates of the validity of selection methods (correcting a statistical error that had inflated the figures for twenty years).

Result: the best-known selection method explains less than 20% of the variance in future performance. No one knows how to predict the remaining four-fifths.

Simulation is no exception. No tool can reliably tell you who will be a good manager. This is not a technological ceiling that AI can break through; it is a statistical ceiling.

But we need to look carefully at what this result actually says. It doesn't say that assessment is useless. It says that using assessment to predict performance and rank candidates has a ceiling.

But assessing someone to understand what they need to work on : producing a map of strengths and weaknesses to know where to focus, requires no predictive validity at all: you don't need to know whether someone will succeed eighteen months from now to observe that they avoided the difficult topic three times in a row, and train them on that topic starting Monday.

De-risking isn't predicting failure. It's reducing its frequency and its cost.

That's exactly YouMight's mission: map potential problems before the promotion, and support each talent in a personalized way.

Step 1: Before the promotion, assess to understand.

Let's go back to the problem from paragraph 1. When the committee decides on a promotion, the person's managerial behavior doesn't exist yet. So it can't be observed. Unless we make it exist.

That's exactly what our simulations do: they artificially create the one thing that is missing: observable managerial behavior. By having someone manage a simulated conflict, a difficult project alignment, a customer crisis, or even a sexual harassment case, we can effectively understand their strengths and weaknesses.

Important: The goal of this phase is not to get a grade, and certainly not a potential score: the goal is to build a map. Here is where this person is already strong; here are the two or three areas that could prove costly during their first six months.

We've worked extensively on this map:

A traditional HR framework, such as the one used in annual performance reviews, with broad labels (“demonstrates leadership,” “communicates effectively”) rated from 1 to 5 by someone who remembers the past year, does not do this job. For two main reasons:

• It isn't observable. “Demonstrates leadership” is not a behavior; it is a conclusion. A simulation framework needs to go one level deeper, down to what the evidence actually contains: did they consult others before making the announcement, or afterward? Did they address the issue through a channel appropriate to its stakes? Did they check that the message was understood, or simply assume it was? This is the principle behind behaviorally anchored rating scales, formalized by Smith & Kendall as early as 1963.

A score without justification cannot be challenged, and therefore cannot be acted upon — and frankly, it is not acceptable in a promotion context. Every point in our framework points to dated excerpts from simulations.

• It isn't contextualized. The same sentence does not have the same value depending on who hears it, when, and what happened beforehand. A simulation framework evaluates behavior in context, not a trait in isolation.

That's why we created our own framework of contextualized and evolving assessment frameworks.

Step 2: Before taking the role, work on your weaknesses.

Once the promotion has been decided, the usual reflex is to send the person through a standard program: three days of “management fundamentals,” the same for everyone, regardless of their starting point.

Here, the journey doesn't start from scratch. It targets the two or three problems the assessment has already identified. That's what “personalized” means in a precise sense: it doesn't mean “we call you by your first name.” It means scenarios generated from your organization, your role, and your measured weaknesses. Everyone trains on their own blind spots, within their own company, with the types of people they will actually interact with.

And because the marginal cost of a repetition is just a few cents, you can repeat the exercise until you find the keys to success and turn those keys into genuine reflexes.

Step 3: During the transition, provide support to de-risk it.

The critical moment isn't Day 0; it's Day 90. Theory holds for three weeks. Then come the first real conflict, the first unpopular decision, the former colleague who has become your subordinate and comes to test where the boundaries lie.

That's why our support doesn't stop at the promotion:

The new manager continues to practice on what is actually happening to them (rather than trying to remember the training they received three months too early, when they didn't yet know what questions to ask). They continue to improve, but at an accelerated pace, with each simulation providing a condensed dose of experience.

As everything moves faster, junior managers no longer have the time to learn on the job the way their predecessors did.

Conclusion

The Peter Principle is not a cultural inevitability.

Today, we give people managerial responsibilities based on the only data available to us, and that data concerns the wrong job, the one they held before the promotion. Then we hope everything will work out.

We will never be able to predict with 100% certainty whether someone will be a good manager. The best-known selection tool explains less than one-fifth of the variance in performance, and this ceiling is statistical, not technological.

What we can now do, however, is something that was out of reach three years ago: produce observable managerial behavior. We can therefore understand each person's weaknesses and simulate accelerated experience to reduce them.

Practice. Analyze. Repeat.

A methodology that allows everyone to push back their level of incompetence. Grow faster while limiting risk.

At YouMight, we are building a platform to de-risk management transitions: a simulated organization that puts future managers into real, personalized situations, where they can learn by doing and work on their weaknesses before they impact your company.

Sources

• Benson, Li & Shue (2019), Quarterly Journal of Economics — empirical test of the Peter Principle on 38,843 salespeople
• Sackett, Zhang, Berry & Lievens (2022), Journal of Applied Psychology — re-estimation of the validity of selection methods
• SIOP — accessible summary of the re-ranking of predictors following Sackett et al.
• Smith & Kendall (1963), Journal of Applied Psychology — Behaviorally Anchored Rating Scales (BARS)
• Regulation (EU) 2016/679, Article 22 — automated individual decision-making