Analytical Thinking

These are different abilities. A person excellent at reconciliation may be poor at causal reasoning, and someone strong at building an estimate may be careless about verifying an input. Writing analytical thinking in a job description without saying which of these you mean produces a requirement nobody can screen for and nobody can be rejected against defensibly.

The interview question about a time the candidate solved a complex problem is answered from a prepared story. Give them a task instead.

• Use real material, anonymised. A payroll register with three planted inconsistencies, a pipeline report where the drop-off is at one stage, an attendance extract that does not tie to the muster.
• Keep it to twenty or thirty minutes, and let them ask questions. What they ask before starting is often more informative than the answer.
• Score what you decided in advance: did they find the discrepancy, did they check the total, did they say what they could not conclude, did they state an assumption rather than hide it.
• Look for the candidate who says the data is insufficient where it genuinely is. Confident conclusions from inadequate evidence are the failure mode this assessment is meant to catch.
• Give the same task to every candidate for the role and mark against a key, which is what makes the comparison meaningful.
• Pay attention to accessibility and language. A timed reading-heavy exercise assesses reading speed alongside reasoning, and where the job is not performed under time pressure that is a barrier rather than a signal.

The function that asks for analytical thinking in candidates produces some of the most common analytical errors in organisations.

• Attrition computed on a denominator nobody defined, so the figure is not comparable with last year or with anybody else's.
• Percentages on small populations. Three exits from a team of nine is thirty three per cent, and it is also three people.
• Comparing time to hire across roles with different notice periods and treating the difference as recruiter performance.
• Survey results reported as a score with no view of who did not respond, when non-response is usually the finding.
• Correlation offered as cause, most often that a training programme improved retention among the people who chose to attend it.
• A change reported without the range. A median that moved two points in a distribution that spans forty is noise presented as a result.

The practical discipline is to state the denominator, the population size and the period on every number, and to say what the figure would need to be for anyone to act differently. A metric that would change no decision is not analysis.

• Ask for the reasoning, not the answer. A manager who requires the working to be shown teaches more than any training module.
• Require assumptions to be written down alongside estimates, which is the single habit that most improves analytical work.
• Review decisions afterwards against what was predicted. Without that loop, confident reasoning is never corrected.
• Teach the specific tools the role needs, since analytical ability in the abstract is not trainable in a two-day workshop while spreadsheet competence and reconciliation technique are.
• Give people access to the underlying data rather than to a dashboard alone. Reasoning cannot develop where the only available object is a summary somebody else produced.