On this page
What the Dreyfus model is and where it comes from
Stuart and Hubert Dreyfus published their model of skill acquisition in 1980, studying how airline pilots, chess players, and language learners progress from rule-following to intuitive expertise. The model defines five stages: Novice, Advanced Beginner, Competent, Proficient, and Expert. Each stage describes a qualitative shift in how a person processes information and makes decisions, not a quantitative increase in knowledge.
In competency mapping, the Dreyfus model provides the vertical axis. Competencies are the rows (what you measure). Proficiency levels are the columns (how developed each competency is). Without a shared scale, two raters looking at the same person will produce different scores because they are calibrating against different mental models of "good."
The five levels, defined by behaviour
Each level is defined by what the person does, not what they know. This is the distinction that makes Dreyfus useful in a 360 assessment: raters observe behaviour, not internal knowledge. The shift from one level to the next is not about accumulating more facts. It is about how the person uses context.
- 1Level 1: Novice Follows explicit rules and checklists. Needs step-by-step guidance. Cannot adapt when the situation deviates from the script. Asks "what do I do next?" rather than "what does this situation require?"
- 2Level 2: Advanced Beginner Recognises recurring patterns from experience. Starts to see that some situations are similar to ones they have handled before. Still relies on rules but begins to notice when a rule does not quite fit. Needs supervision on exceptions.
- 3Level 3: Competent Plans work based on priorities, not just rules. Can distinguish what matters from what does not in a given context. Makes deliberate choices and accepts responsibility for the outcome. This is the level most organisations expect for a fully ramped employee in role.
- 4Level 4: Proficient Sees the situation as a whole rather than a set of individual factors. Responds intuitively to familiar patterns and only deliberates on the novel parts. Teaches and mentors others at Levels 1 through 3. Recognises when a playbook needs updating.
- 5Level 5: Expert Acts from deep intuition built on extensive experience. Does not rely on rules or guidelines for routine decisions. Others reference this person when trying to understand what excellent performance looks like. Performance at this level is the standard for the behavioural anchor.
Competency only
First-cycle framework, small teams, leadership overviews.
Pros
- + Fast to build
- + Easy to discuss
- + Low maintenance
Cons
- - Calibration noise
- - Hard to debate a rating
- - Thin IDP input
Competency + behaviours
Most mid-size organisations. Sweet spot for annual reviews.
Pros
- + Observable evidence
- + Defensible ratings
- + Clear IDP targets
Cons
- - More writing upfront
- - Behaviours need maintenance
Competency + behaviours + skills
Technical functions, regulated industries, skills-based hiring.
Pros
- + Training maps to skill
- + Certifications integrate
- + Hiring bar sharper
Cons
- - Heavy to build
- - Ages faster
- - Needs AI to maintain
How to write behavioural anchors that hold up
A proficiency level without a behavioural anchor is a label. Labels produce inconsistent ratings because each rater fills in their own definition of "competent" or "proficient." The anchor is a one-sentence description of what someone at that level actually does in the context of a specific competency.
A good anchor passes the observation test: could a colleague who works with this person confirm or deny the statement based on what they have seen?
- 1Start with a verb "Identifies stakeholder concerns before they surface in meetings" is observable. "Has good stakeholder awareness" is not. The verb forces you to describe behaviour rather than traits.
- 2One behaviour per anchor "Delegates tasks with clear outcomes and follows up weekly" contains two behaviours. Split it. A rater might see the delegation but not the follow-up, and the combined anchor forces them to average two observations into one score.
- 3Make the gap between levels visible If your Level 3 and Level 4 anchors sound similar, raters will cluster their scores. The language at each level should be distinct enough that a rater can point to specific behaviours that place someone at one level and not the adjacent one.
- 4Test with three people Read the anchor to three managers who know the role. If they cannot independently agree on what level a specific employee sits at, the anchor is too vague. Rewrite it until they converge.
Calibrating raters across the scale
Even well-written anchors produce inconsistent ratings if raters have not been calibrated. Calibration means aligning raters on what each level looks like before the assessment window opens. Without it, a generous manager rates Level 3 performance as Level 4, and a strict manager rates the same performance as Level 2. The assessment data tells you more about the rater than the assessee.
- 1Run a 30-minute anchor walkthrough Before the assessment opens, walk raters through 2 or 3 competencies with their anchors. Show an example profile and ask each rater to rate independently. Then compare. Where they diverge, discuss the anchor until they converge. This single session reduces inter-rater variance by roughly half.
- 2Use the same scale across the entire assessment Mixing Dreyfus (5 levels) with a 4-point Likert on some competencies and a 10-point scale on others destroys comparability. Pick one scale for the entire campaign. If you use Dreyfus, use Dreyfus everywhere.
- 3Show the anchor during the rating Do not ask raters to remember the anchor from a training session two weeks ago. Display the anchor next to the rating input. Raters who see the anchor while rating produce scores that cluster closer to the true mean than raters who rate from memory.
In a 360 assessment, the multi-rater design already reduces individual bias by averaging across 5 to 8 raters. Calibration reduces the remaining variance so the average is closer to the true signal.
Build a framework with anchored proficiency levels
313 pre-built competencies with Dreyfus 0 to 5 anchors. Customise, import your own, or let Huna AI generate from job descriptions.
Three mistakes that break the scale
Most scale failures are not about the model. They are about how the model is implemented.
- 1Missing Level 0 (no exposure) The original Dreyfus model starts at Novice, but in a competency framework some employees have no exposure to a competency at all. Without a Level 0, raters are forced to rate someone at Novice when the honest answer is "this competency does not apply." Huneety uses a 0 to 5 scale for this reason: 0 means the competency is outside the person's current scope.
- 2Anchors written as personality traits "Is a natural leader" is a trait, not a behaviour. It cannot be observed, only inferred, and it invites bias (the person who "looks like" a leader gets rated higher). Rewrite every trait-based anchor as an observable action.
- 3Using the scale for performance, not capability Dreyfus measures how developed a competency is. It does not measure how hard someone works or how much output they produce. A Proficient employee having a bad quarter is still Proficient. Conflating capability with performance turns the scale into a backdoor performance review, and raters will sanitise their ratings accordingly.
Dreyfus on Huneety
Huneety uses the Dreyfus 0 to 5 scale as its default proficiency model. Every competency in a framework carries behavioural anchors at each level, and raters see the anchor inline during the assessment. The competency mapping module lets you write custom anchors per competency or start from the 313 pre-built competencies that already have anchors attached. If you use a different scale (4-point, 7-point, or a custom proficiency model), the platform supports that too. Dreyfus is the default because it balances granularity with rater usability.
FAQ
Quick answers
- How many proficiency levels should a competency framework have?
- Five is the most common (Dreyfus). Some organisations use 4 or 7. Fewer than 4 and raters cannot differentiate meaningfully. More than 7 and the adjacent levels become indistinguishable. Dreyfus at 5 (or 6 with Level 0) is the standard because it balances precision with usability.
- Can I use Dreyfus alongside SFIA or Korn Ferry?
- Yes. SFIA 9 uses a 7-level responsibility model. Korn Ferry uses its own proficiency scale. Huneety maps across all three. The underlying competencies are the same; the scale is a lens. You can import a SFIA-based framework and rate it on a Dreyfus scale if that is what your raters are trained on.
- What is the difference between Dreyfus Level 0 and Level 1?
- Level 0 means the person has no exposure to the competency. It is not in their role scope. Level 1 (Novice) means they are actively developing the competency but still follow rules and checklists. The distinction matters because rating a non-applicable competency as Novice inflates the gap analysis and produces misleading development priorities.
- Do I need to write behavioural anchors for every competency?
- For competencies you assess, yes. An anchor-less competency produces unreliable ratings because every rater calibrates against their own mental model. For competencies that are only tracked (not assessed), anchors are optional but recommended for consistency when you do assess them later.
Continue learning
Competency framework examples
Six real frameworks by function (engineering, sales, leadership, customer success, marketing, operations) with behavioural anchors.
Read the guide
Soft skills vs hard skills in a framework
How to structure frameworks that cover both technical and behavioural capabilities without one drowning the other.
Read the guide