Getting Honest Answers: How to Avoid Bias in Self-Assessments

The great strength of a self-assessment is that the people answering know the organisation better than any outsider. Its great weakness is that they are not neutral. They have reputations to protect, assumptions they have never tested, and a natural tendency to see their own work in a good light.

None of this makes self-assessment unreliable. It means the design and the way you run it matter. This article covers the biases that most often distort maturity assessments, and practical ways to reduce each.

Optimism and the policy trap

The most common bias is simple optimism: people rate their organisation more highly than an independent observer would. A particular version of this is answering based on what the policy says rather than what actually happens.

If a documented process exists, it is tempting to choose the answer that describes having it, even if half the teams ignore it. Leaders are especially prone to this, because they are more likely to know the policy and less likely to see day-to-day reality.

How to reduce it:

• Ask people to answer based on what happens in practice, and say so explicitly at the start.
• Use answer descriptions that focus on observable behaviour ("reviews happen every month and actions are tracked") rather than the existence of documents.
• Include people who do the work, not just people who manage it.

Social desirability

People want to give the answer that makes them, their team or their manager look good. This is strongest when they believe the results will be used to judge them.

How to reduce it:

• Make clear that the assessment is for improvement, not performance management, and mean it.
• Keep individual answers private. Share aggregated results and themes, not who said what.
• Never use maturity scores as personal or team targets. As soon as a score becomes a target, people learn to manage the score rather than the capability.

Anchoring

When people hear an opinion before forming their own, they tend to move towards it. In a group setting, the first person to speak, often the most senior, sets the anchor, and everyone else adjusts around it.

How to reduce it:

• Ask everyone to complete the assessment independently before any discussion.
• In workshops, ask people who gave different answers to explain their reasoning before the most senior person speaks.

Our guide to running a maturity assessment workshop sets out a format built around independent answers and structured discussion.

Central tendency

When unsure, people gravitate to the middle answer. It feels safe and moderate. On a five-point scale this produces a sea of threes; on any scale it hides both strengths and weaknesses.

How to reduce it:

• Use fewer, clearly different levels, so that each answer is a genuine choice. We explain the case for this in why three maturity levels beat five.
• Write each answer as a concrete description, so people can recognise their situation rather than choose a number.
• Look at the spread of answers, not just the average. If everyone has chosen the middle option on every question, treat the result with suspicion.

The halo effect

If people feel generally positive or negative about an area, that feeling colours every answer within it. A team that has just delivered a successful project may rate all of its practices highly, including ones that had nothing to do with the success.

How to reduce it:

• Ask about specific, recent examples. "Think about the last three changes your team introduced" is better than "How good are we at change?"
• Mix question topics so that people consider each one on its own merits.

Recency

Recent events loom large. A serious incident last month can drag down every answer in a risk or security assessment, while a calm period can make everything look fine.

How to reduce it:

• Specify a time frame, such as "over the last 12 months".
• Ask for evidence: "What have you seen that makes you choose this answer?"

Knowing what you don't know

People with less experience in an area sometimes rate it highly simply because they do not know what good looks like. Those with more experience may be harsher judges, because they can see the gaps.

How to reduce it:

• Use answer descriptions that show what higher maturity looks like, so everyone has the same reference point.
• Include at least one person with broader experience, from another team, another organisation or an external adviser.

Design choices that help

Many biases can be reduced by how the assessment itself is designed:

• Concrete answers. Each option describes what actually happens at that level.
• A short hint per question. A sentence clarifying what the question is getting at reduces misinterpretation.
• Reasonable length. Long assessments cause fatigue, and tired respondents default to the middle. Around 25 questions is usually enough for one capability.
• Privacy. Answers that never leave the respondent's own browser, or are only shared when they choose, encourage candour.

Using bias to your advantage

Biases are not only a problem. Differences in how people answer are often the most useful output of an assessment. When leaders consistently score an area higher than frontline staff, that gap is itself a finding: the people setting direction do not see the same reality as the people doing the work.

Rather than averaging such differences away, bring them into the open. Ask, "Why do we see this so differently?" The answer is usually more valuable than the score.

Frequently asked questions

Is an external assessment always more accurate?

Not necessarily. External assessors avoid some biases but introduce others, such as relying on documents and interviews that show the organisation at its best. A well-run self-assessment with a range of participants is often just as accurate, and much cheaper.

How many people should complete a self-assessment?

For a team or department, five to ten people with a mix of roles and seniority is usually enough to see patterns and disagreements. For an organisation-wide view, include people from each major area.

Should we tell people the results will be shared with leadership?

Be honest about how results will be used. If aggregated results will go to leadership, say so, and reassure people that individual answers will stay private.