Auditing Your Own Process

Most organisations have never looked at their own assessment data. The audit below requires no expertise and produces findings almost every time.

The principles in “Auditing Your Own Process” become easier to maintain when ownership and time spent are visible. Teams evaluating project time tracking can use it to coordinate the administrative work around assessment, identify stages that repeatedly expand and plan capacity, while keeping the hiring decision tied to job-relevant evidence instead of raw activity totals.

For an independent benchmark, compare the process with U.S. OPM assessment resources; the useful test is whether the local method remains job-related, proportionate and explainable.

The six figures

Pass rate by stage. Where does your funnel narrow, and is that where you intended?

Pass rate by group, by stage. The adverse impact note covers the calculation.

Dropout by stage, with reasons where you have them.

Time from application to decision.

Assessor agreement on the same candidates.

And assessment scores against how hires actually performed six to twelve months later.

What each usually reveals

Pass rates show a stage that removes almost nobody, which is pure cost, or one that removes almost everybody, which is doing the whole selection unexamined.

Group differences concentrate in one stage more often than they spread evenly.

Dropout clusters where the process is slow or the demand is high.

Timelines are longer than anybody in the organisation believes.

Assessor agreement is worse than assessors expect.

And score-performance relationships are frequently absent, which is uncomfortable and is the most useful finding available.

Doing it

Pull the last twelve months of hiring data.

Aggregate across similar roles to get useful numbers.

A spreadsheet is sufficient; this does not require a system.

A day, once a year.

The hardest figure

Scores against performance.

It requires that somebody recorded scores, that performance is measured somehow, and that a person is willing to join them up.

Small samples prove nothing and will still show you whether your top scorers are conspicuously failing.

If there is no relationship at all, that is worth knowing before you spend another year running the process.

Who should see it

Whoever can change the process.

Not only the recruitment team, who frequently cannot remove a stage a senior stakeholder insisted on.

The audit is most useful as an argument, and it is a strong one because it is your own data rather than a vendor's claim.

Acting on it

Remove stages that have never changed an outcome.

Investigate the stage producing group differences before removing it.

Fix scheduling and speed, which are usually the cheapest improvements available.

And calibrate assessors, which the calibration note covers and which is the single highest-return intervention for most organisations.

What not to conclude

That a stage with a low pass rate is doing good work. It may be filtering on something irrelevant.

That a null score-performance relationship means assessment is useless. It means yours is not working, which is fixable.

And that a group difference is caused by whoever assessed most of the affected candidates, which is usually wrong and always unhelpful.

What to check

Which of the six figures could you produce today?

Has anybody looked at them in the last year?

Does the person who could change the process see them?

And when something was found last time, did anything change?

The point

Six figures, a day, once a year: pass rates by stage and group, dropout, timeline, assessor agreement, and scores against performance..

Underlying all of this

Almost everything in this collection reduces to one discipline: write down what the job requires, assess that thing directly, record the evidence, and look at your own outcomes afterwards. None of it requires buying anything, and organisations that do those four things consistently outperform ones running longer processes built from instruments chosen before the requirements were known.