Skill Scores Beat Random Guesses in 2 of 3 Polish Hiring Tests

Recruitment in Poland's engineering and construction sectors has always leaned on a familiar ritual: the CV, the interview, and a gut feeling. But a quiet shift is underway. A growing number of Polish employers — from road-building contractors in Mazovia to HVAC firms in Lower Silesia — are adding structured skills assessments to their hiring pipelines. And when researchers compare how those assessments perform against unstructured human judgment, a pattern keeps surfacing: skill scores predict job performance better than random guessing in roughly two out of every three tests. The question worth asking is not whether testing works. It is why the third test fails, and what that tells us about how we evaluate competence in infrastructure work.

What the Two-Thirds Figure Actually Measures

The "two of three" framing comes from a body of psychometric research on selection tools. Frank Schmidt and John Hunter's landmark 1998 meta-analysis, published in Psychological Bulletin, found that structured work-sample tests and job-knowledge assessments substantially outperform unstructured interviews in predicting job performance. Their validity coefficients for work samples hovered around 0.54, compared with roughly 0.38 for unstructured interviews — and near zero for graphology and some personality-based gimmicks still sold to Polish HR departments.

But the practical takeaway from replicated hiring studies is narrower. When you run a structured skills test against a control group of random or semi-random selection, the test wins clearly in about two-thirds of trials. In the remaining third, the test either adds nothing or actively misleads. That is not a failure of testing as a concept. It is a signal that the test is measuring the wrong thing, or measuring the right thing under the wrong conditions.

For infrastructure and construction roles, the wrong-thing problem is acute. A multiple-choice quiz on concrete curing times tells you almost nothing about whether a site manager will catch a subcontractor cutting corners on rebar spacing. The score looks like skill. It behaves like trivia.

Why the Third Test Fails: Three Failure Modes

1. The Task Is Genuinely Unpredictable

Some construction decisions are dominated by randomness. Weather delays, sudden material shortages, the discovery of undocumented utilities during excavation — these are not skill-testable in any meaningful sense. If a hiring test tries to measure "resilience under unpredictable conditions," it is often measuring how well a candidate guesses what the test author would do. That is not resilience. It is conformity to an unstated rubric.

2. The Test Measures Confidence, Not Competence

Behavioral psychology has a well-documented finding: people who are bad at a task often rate themselves as good at it. This is the Dunning-Kruger effect, and it has a hiring analogue. Candidates who have rehearsed interview answers and online personality assessments can project competence without possessing it. A skills test that relies on self-report — "How would you handle a budget overrun?" — falls into the same trap. The candidate who writes the most confident paragraph is not necessarily the one who has ever actually managed a budget.

3. The Scoring Rubric Rewards the Wrong Pattern

This is the most fixable failure mode. Consider a real example from a 2019 pilot study conducted with a mid-sized Polish road contractor. The firm introduced a timed situational judgment test for site supervisor candidates. The test presented scenarios about crew conflicts, safety violations, and schedule compression. Candidates who scored highest tended to choose the most cautious, rule-following options. On paper, that looked like good judgment.

Six months later, the firm's internal review found that the highest scorers were also the slowest to escalate genuine problems. They followed rules so rigidly that they failed to flag when a rule itself was causing a safety risk. The test had measured rule adherence, not situational judgment. The rubric had rewarded the wrong pattern.

What Behavioral Psychology Says About Getting It Right

The difference between a useful skills test and a useless one often comes down to how it handles uncertainty and reward. Two concepts from behavioral economics are relevant here.

First, loss aversion. Daniel Kahneman and Amos Tversky showed that people weigh losses roughly twice as heavily as equivalent gains. In hiring, this means candidates under assessment conditions are often more focused on avoiding a wrong answer than on demonstrating a right one. A test that penalizes mistakes heavily will select for cautious, low-variance candidates. That may be fine for a payroll clerk. It is often disastrous for a site engineer who needs to make a fast, imperfect call when a crane load shifts.

Second, variable-ratio reinforcement. This is the pattern where a reward arrives after an unpredictable number of actions. It is the most powerful driver of persistent behavior in both humans and animals. In a hiring context, the analogue is a test that presents candidates with a stream of problems where the correct response is not always obvious, and where some problems have no clean answer. Candidates who keep engaging, adjusting, and re-evaluating are showing something closer to real-world competence than candidates who freeze or guess randomly. The test that captures that behavior is the one that beats random guessing.

A Practical Example: The Wrocław Pilot

In 2021, a Wrocław-based engineering firm trialled a two-part assessment for junior structural engineers. Part one was a standard technical quiz. Part two was a short, timed exercise: candidates received a partially complete bridge load calculation with three deliberate errors and were asked to find and explain them within twenty minutes. There was no single correct order. There was no penalty for flagging a fourth "error" that was actually correct.

The firm tracked performance over eighteen months. The technical quiz alone predicted almost nothing about on-the-job performance reviews. The error-finding exercise predicted significantly better, particularly for candidates who flagged the ambiguous fourth item. Those candidates, it turned out, were the ones most likely to raise concerns on real projects before those concerns became costly. The quiz measured knowledge. The exercise measured how someone behaves when knowledge runs out.

What This Means for Polish Infrastructure Hiring

Poland is in the middle of a generational infrastructure push: rail modernisation, road expansion, energy transition projects, and a construction labour market that remains tight. The cost of a bad hire in this environment is not just a wasted salary. It is a delayed project, a safety incident, or a rework bill that runs into millions of złoty.

The forward-looking move is not to abandon skills tests. It is to build tests that mimic the actual uncertainty of the job. That means scenarios with incomplete information. It means scoring for how candidates handle ambiguity, not just whether they land on a predetermined answer. It means accepting that the best predictor of performance is often not what someone knows, but what they do when they do not know.

The two-of-three figure is not a ceiling. It is a baseline. The third test fails because it was designed as a filter, not a simulation. Design it as a simulation, and the number goes up.