How do you Solve a Problem like Potential?

Prediction, Pygmalion, and Performance Instability

In Ovid’s telling of the myth of Pygmalion in Metamorphoses, a sculptor carved a statue so divinely beautiful and so lifelike he could scarcely keep himself from believing she was real. He brought her gifts — flowers, shells, smooth stones — offerings you would bring someone you love, not something you made. When he finally kissed her, she turned from ivory to flesh. The story has become shorthand for the idea that people can grow into what others imagine they might become.

Pygmalion and Galatea. Auguste Rodin (1908).

In organizations, we tell ourselves a similar mystical story. We call it potential. We build training programs, leadership pipelines, succession plans, all premised on the belief that exceptional talent can be identified early and molded into something extraordinary.

It’s a seductive idea. It’s also, I would argue, one of the least rigorously scrutinized in the entire talent management canon. It’s not that potential doesn’t exist; it’s just that we treat it as a performance evaluation problem when it’s actually a prediction problem — and a surprisingly tricky one.

Measuring potential demands precision about something that is, by definition, not yet knowable. We are trying to assess who someone may become, for a role they have never held (and may not even exist yet). The label itself shapes outcomes that look like validation evidence. And worse, the whole premise is built on a criterion that is shifting beneath our feet.

How do you catch a cloud and pin it down?

Potential is prediction, not performance

There was a time earlier in my career when our talent management team had just finished a performance review cycle — managers trained, reviews written, ratings collected and calibrated — and I was pulling together some data for a leadership presentation. That’s when I saw it: Performance and potential scores were coupled so tightly that a full two-thirds of employees received identical ratings on both dimensions. I’m pretty sure let out an audible groan.

Performance evaluation and potential assessment feel like cousins. Both typically involve the manager, the same process, the same nine-box rating scale, the same calibration conversations, and the same implicit theory of who the good people are. But, structurally, they should be doing completely different things. Performance appraisal is retrospective: Did this person meet expectations? Potential is prospective: Will this person meet expectations they haven’t yet had an opportunity to demonstrate? It’s a diametrically different inferential problem, yet we routinely act as if swapping in a different “P” word changes nothing about the nature of the judgment.

The field’s most rigorous thinking about potential assessment looks almost nothing like performance evaluation. The Leadership Potential Blueprint¹ — the most widely cited academic framework on the subject — identifies three distinct dimensions: foundational characteristics like cognitive ability and personality, growth dimensions like learning agility, and career factors like leadership competencies and functional expertise. These toolkits rely on tests of personality and cognitive ability, structured 360-degree feedback, and customized business simulations.

That’s not a performance evaluation — it’s a selection process. And it makes sense. Potential requires us to evaluate a person against a job they haven’t done yet.

But that is not what most organizations do. One benchmark study found that the most common method of potential identification was the opinion of senior leaders (59%), followed by performance appraisals (51%). Only 16% used customized 360-degree feedback or other types of formal assessments.² To be fair, it’s easy to understand why. Assessing potential rigorously is resource-intensive, methodologically demanding, and difficult to explain to managers and employees. Imagine all the difficulty and controversy about measuring past performance; now let’s try it with a hypothetical scenario no one has actually seen. When we make performance evaluation more rigorous, we ground it in observation — competency versus expectations; results versus targets. Potential has none of these.

The annoyingly high correlation I found told me, when presented with an ambiguous question, managers had defaulted to one judgment instead of two. Of course they had! Putting anything other than someone’s current performance would have been a risk. Betting an under-performer could turn it around or proclaiming a high performer had maxed out would require the manager to defend their position, and we hadn’t given them the rational framework to do that.

We conflate performance and potential and then, for good measure, we use them in reverse. It’s why we find the Peter Principle borne out. We’re more comfortable promoting someone who’s a high performer in their current role than we are using assessments that predict their success in the next role. The contamination runs the other way, too. When potential seeps into performance evaluations, ratings easily become a bet on what someone could become instead of what they actually did.

But treating potential like a selection problem will only get us so far. The science here is as extensive as it is humbling. The highest operational validities we find among selection instruments are around .4 — and that’s assuming perfectly reliable instruments, standardized administration, and a clearly defined and reliable criterion. That’s about as likely as the DMV having a 5-star Yelp rating, and it still leaves over 80% of the variance in job performance unexplained.³

The problem is, even if we had magically predictive assessments, we’d still have almost no way of knowing it. That’s because the label of potential may be better at shaping the future than it is at predicting it.

How do you hold a moonbeam in your hand?

Measuring potential changes it

In 1966, psychologists Robert Rosenthal and Lenore Jacobson⁴ walked into an elementary school and told teachers a handful of their students had “unusual potential for intellectual growth.” At the end of the year, those students showed gains in IQ scores that were 50% greater than the rest of the class. Here’s the twist: Those students had been selected randomly. They had no more or less potential than any others. The label might not have changed anything inherent about the students themselves, but it changed how teachers treated them — who they called on, how they responded to a mistake, the type of encouragement they gave.

The same dynamic operates in organizations. Employees designated as high potential don’t just have some special code in their file. They get high-visibility projects that then produce the track record assessed in performance reviews later on. They get exposure to senior leaders who then advocate for them. They get the benefit of the doubt when something goes wrong on a stretch assignment. Research by Eden and colleagues showed that manager expectations shape employee performance through a surprisingly direct mechanism: The manager’s own behavior changes first, and the employee’s performance follows.⁵ Over time, high potentials improve and advance, not just because of differences in underlying capability but also because of the compounding advantage of having the label.

I’ve seen this dynamic in my own career. Some of my most pivotal professional roles were ones I almost certainly wouldn’t have been hired for externally because I didn’t have the demonstrated experience on my résumé. It’s hard to say for sure whether I already had the raw, underlying capability or not. What I do know is that I learned by doing the thing. The track record came later.

It even holds true in my personal life, for the most important job I’ve ever had. When my first child was born, I had never held a newborn. I don’t think I had ever changed a diaper. By any reasonable hiring standard, I was wildly under-qualified. Five years and three kids later, I can confidently say nothing could have prepared me for the responsibility of being a parent aside from becoming a parent. The capability came from the experience; it could never have preceded it.

The endogenous, self-fulfilling nature of potential makes it very difficult to establish whether designation predicts success or produces it. When high potentials go on to succeed, it looks like validation, but it may well be accumulated opportunity.

It’s surprising to me this problem hasn’t attracted more attention in the field, but I suppose there are several reasons. Studying it properly would require randomly assigning potential status regardless of assessment, and organizations would never go for that. Even if we could get randomized data, there’s little incentive to seek it out. If your high potential program appears to be working — designated people get good performance ratings and get promoted more quickly — the self-validating loop is comfortable. Interrogating the causality might lead you to confront the role of the company in creating different outcomes for different people.

But if potential is at least partly constituted by doing the thing, and capability develops through the experience rather than preceding it, then a system that requires demonstrating potential before granting access to opportunity has the sequence somewhat backward. In addition to identifying who has potential, we also need to figure out who has had the opportunity to show it.

The phenomenon discovered by researchers in elementary schools was named the Pygmalion Effect. I’ll admit I always assumed Pygmalion was the subject of the myth, the one being shaped and transformed. But I had it backwards, the same way we use “Frankenstein” to refer to the monster when, in Mary Shelley’s original work, he’s the scientist. It wasn’t until I reread the story that I realized Pygmalion was the sculptor, and, critically, not just an observer — the creator.

How do you keep a wave upon the sand?

Performance is changing before our eyes

Here’s where things get really inconvenient. The science of personnel selection works best when the outcome is stable. We do the job analysis; identify the knowledge, skills, abilities and other characteristics it requires; build or borrow a tool; and hire based on the expectation that the score will predict future performance. The entire notion of criterion-related validity rests on a key assumption — that what success looks like today is a reasonable proxy for what success will look like when it comes time to test whether your prediction was right.

That assumption is under mounting pressure. Lightcast’s Speed of Skill Change Report⁶ found the average occupation in the U.S. saw a third of its required skills change between 2021 and 2024. For the top quartile of occupations, that figure was 75%. And the pace is only increasing. The magnitude of change in those three years was comparable to the previous five. Research by Microsoft and LinkedIn⁷ project the skills required for a given job will change by 50% by 2030, and generative AI could raise that figure to as high as 68%. Changes are happening in quarters and months, not years. The job configurations and competencies organizations are hiring for today may look materially different by the time a new hire is fully onboarded, let alone evaluated in a couple of performance-evaluation cycles or promoted a level or two. When the criterion drifts faster than we can validate the predictor, the scaffolding of the whole system begins to buckle.

This creates a problem that touches all of our talent programs, including hiring and potential assessment, and neither has offered a clean answer. Rigor in hiring comes from its specificity, which turns out to be brittle in the face of change. Potential is more flexible, but that very conceptual fuzziness along with its endogeneity make it under-validated. It’s like trying to capture a photograph of a hummingbird with a camera meant for a still life; it all turns out a blur.

How do you find a word that means ‘potential’?

We sculpt the image

There should be a law that every Substack article must contain at least one Arrested Development gif.

This is usually the point where I get the urge to just abandon ship. These feel like intractable problems. But, as usual, the way out is less about having a tidy solution and more about thinking more deeply about our intent. What are we actually trying to do when we measure potential?

Most high potential programs are, at their core, about leadership pipeline. Who will run things? Who will sit in the highest seats? And that means every program encodes a prototype — an implicit theory of what leadership looks like. Are those future leaders extroverted and decisive or brainy and deliberate? Are they visionary disruptors or meticulous operators? Your answer probably depends on the profile of your current leadership team. Organizations think about potential as some sort of universal, naturally occurring resource. It isn’t. It is a concept authored by the people who succeeded at a particular time, under a particular set of conditions.

The people defining potential today are, in many cases, people who climbed a ladder that was relatively stable compared to what we may see in the next few years. They’re now being asked to identify who will succeed in an environment that may look quite different — more dynamic, less linear, more complex, and more dependent on the capacity to operate under chasmic uncertainty rather than managed complexity. Under those circumstances, past performance may be a weaker predictor of future performance than it has ever been before.

This brings us to a core misconception of what the field of talent management actually does. We’ve told ourselves we’re in the business of discovery. We like to think we create rigorous, neutral processes that observe, evaluate, and identify talent “objectively” (a word I can’t use without breaking out in hives). We measure performance and uncover hidden gems while the L&D side of the house does the developing. But in practice, this self-image doesn’t hold up, and here’s why:

• We don’t discover potential — we conflate it with the performance record in front of us.

• We don’t observe it — we constitute it through the very act of designation, shaping the trajectory we later hold up as validation.

• And we don’t evaluate it against a stable, well-defined target — we select toward a prototype that is itself a historical artifact, created by people whose own path to success may be a poor map for the terrain ahead.

If that’s true, then the field isn’t just doing measurement badly. It’s doing a fundamentally different thing than we think it’s doing. It’s making choices about who gets developed, toward what model of leadership, under what theory of the future. These questions will not be resolved with better measurement alone.

(For what it’s worth, in the absence of the most rigorous measures, there are a few ways I’ve found to make potential more useful in practice. First, shorten the timeframe. Don’t ask where they’ll be in five years. Instead, ask, is this person ready to take on a more complex, cross-functional project next quarter? Is this IC ready for a management role the next time one opens up? Second, drop the word “potential” entirely. Instead of debating their hypothetical ceiling, focus on what you can reasonably extrapolate from observed behavior. Has this person demonstrated the agility required to manage a broader scope, or to perform in a different function?)

I don’t think all of this means we should abandon the pursuit. Potential does exist. Some people do grow faster, adapt more readily, lead more effectively under conditions of greater complexity, and the work of identifying and developing those people is important. But it means holding the conclusions more loosely and the assumptions more explicitly. Who is in the room when the prototype is defined? Whose theory of leadership is being encoded into the nine-box? When was it last revisited, and by whom?

Remember, Ovid’s Pygmalion didn’t discover the woman in the ivory. No, he imagined her with such conviction, such specificity, such devotion to a particular vision of what she could become that she became real. It seems to me he fell in love with possibility itself. Perhaps that’s the most accurate thing we can say about potential, too: It’s a possibility, not a property. Humans aren’t meant to be captured, static, in stone. The most interesting part of potential happens when it becomes a living, breathing thing — in other words, when it isn’t potential anymore.

Organizations are doing the same thing. Every calibration session is an act of imagination, and we are the creator like Pygmalion whether we realize it or not. Will we have the curiosity and humility to hold the image lightly, and perhaps be surprised by what emerges — who, under different circumstances, in a different role, another moment, might exceed our expectations and become something extraordinary? Asking that question will tell you more about your organization than any nine-box ever could.

Edelweiss by Tobias Oetiker

1. Church, A. H., & Silzer, R. (2014). Going behind the corporate curtain with a blueprint for leadership potential. People & Strategy, 36(4), 50-58.

2. Church, A. H., Rotolo, C. T., Ginther, N. M., & Levine, R. (2013). How are top companies assessing their high-potentials and senior executives? A talent management benchmark study. Consulting Psychology Journal: Practice and Research, 65(3), 199–223.

3. There is at least one critical advantage potential assessments have over external hiring. An internal employee has already demonstrated a trove of potential-relevant information: how they’ve responded to setbacks, how quickly they learn from experience, how receptive they are to feedback, how they perform across varied contexts. These are much harder to measure accurately in a 30-minute interview with an external candidate. Most organizations already have this rich, relevant data, they’re just not using it.

4. Rosenthal, R., & Jacobson, L. (1966). Teachers’ expectancies: Determinants of pupils’ IQ gains. Psychological Reports, 19(1), 115–118. https://doi.org/10.2466/pr0.1966.19.1.115

5. Eden, D., & Shani, A. B. (1982). Pygmalion goes to boot camp: Expectancy, leadership, and trainee performance. Journal of Applied Psychology, 67(2), 194–199. https://doi.org/10.1037/0021-9010.67.2.194

6. Lightcast. (2025, January). The speed of skill change. https://lightcast.io/resources/research/speed-of-skill-change

7. Microsoft & LinkedIn. (2024, May). AI at work is here. Now comes the hard part. 2024 Work Trend Index Annual Report. Microsoft. https://www.microsoft.com/en-us/worklab/work-trend-index/ai-at-work-is-here