How to Build an In-House Assessment Centre: A Step-by-Step Guide

Why this research matters

Assessment centres are one of the most valid selection and development tools available. Here is how to design, run, and evaluate one without hiring an external firm.

The Case for Building In-House

Assessment centres have a predictive validity coefficient of approximately 0.65 for job performance - one of the highest of any selection method, significantly outperforming unstructured interviews (validity ~0.38), reference checks (~0.26), and years of experience (~0.18) (CIPD, citing Schmidt & Hunter, 2024). They are also one of the most expensive selection tools to run - which is why many organisations outsource them entirely to specialist firms, paying significant fees per candidate.

The case for building in-house capability is compelling for any organisation that hires more than 20 professional or managerial roles per year. Once the design and assessor training investment is made, the per-candidate cost drops dramatically, the process becomes faster and more flexible, and internal assessors develop a deeper understanding of what "excellent performance" looks like in the organisation's own context.

Step 1: Define the Purpose and the Competencies

Before designing a single exercise, be precise about what the assessment centre is for. External selection and internal development require different designs. Selection centres need to discriminate between candidates reliably - they need to be standardised, structured, and objectively scored. Development centres are more exploratory - they aim to reveal an individual's strengths and development areas, and can be more flexible in design. Conflating these purposes produces a centre that does neither well.

The competency framework drives everything else. For each competency you want to assess, you need at least two exercises that create observable opportunities to demonstrate it. If leadership is a competency, a group exercise and a role-play will both give assessors opportunities to observe it - providing triangulation. Never assess a competency with only one exercise.

Step 2: Design the Exercise Matrix

A typical one-day assessment centre includes 4–6 exercises. Each exercise should be role-relevant and should create naturalistic opportunities for candidates to demonstrate target competencies. The most evidence-supported exercise types include:

• In-Tray / Inbox Exercise: Candidates receive a simulated inbox of emails, memos, and reports and must prioritise and respond within a time limit. Assesses planning, decision-making, and written communication.

• Case Study Presentation: Candidates analyse a business problem and present recommendations. Assesses analytical thinking, strategic orientation, and oral communication.

• Group Discussion/Exercise: Candidates work together on a task or debate. Assesses teamwork, influence, and leadership in group settings.

• Structured Competency Interview: Behavioural questions with STAR-format probing. Provides candidate self-report evidence to triangulate with observed behaviour.

• Role Play: Candidates manage a simulated conversation - with a manager, client, or direct report. Assesses interpersonal skills, coaching, or client management.

Step 3: Train Your Assessors

Assessor quality is the single greatest risk to assessment centre validity. Untrained assessors collapse into halo effects (rating attractive candidates highly across all competencies) and make premature judgements based on first impressions. Assessor training should cover: the Observe-Record-Classify-Evaluate (ORCE) model; dimensional thinking (rating competencies, not candidates); and calibration practice with video case studies. A minimum of two days of assessor training is the industry standard before deploying assessors in live centres.

Step 4: Run the Centre and Wash-Up

The wash-up session - where assessors pool and reconcile their ratings - is where most of the quality control happens. Use a structured matrix: competencies on one axis, exercises on the other, with each cell containing each assessor's rating. Work through competency by competency, not candidate by candidate, to reduce halo effect. Where assessors disagree significantly, return to the evidence: what specific behaviours were observed? The final output should be a competency profile per candidate, supported by specific behavioural evidence, not a ranked list.