calendar
Edited on October 29, 2026

Competency Mapping: A Step-by-Step Guide With a Worked Example

Build a competency map for one role, write observable levels, pilot the criteria and assess relevant work. Use the worked example and editable worksheet to choose a response to gaps.

Carolina Martin, Learning Experience Manager at Coursebox AI
Carolina Martin
Customer Success Manager
Key takeaways

Competency mapping identifies what a role requires and describes the knowledge, skills and behaviours at its proficiency levels. Assessment against that map can reveal development needs, process problems or gaps that require other support.

  • A competency is usable only when each level is written as behaviour a rater could see or check in a record.
  • Use the required-level indicators to design suitable learning objectives and assessment tasks. Investigate why a behaviour is missing before deciding whether training is the response.
  • Pilot the map on representative work and compare independent raters. Refine unclear criteria; do not rewrite the map merely to reproduce a manager's existing ranking.
  • Published frameworks such as O*NET, SFIA and the APS Integrated Leadership System make a faster first draft than a blank page.
Instructional Designers

Go from storyboard to an interactive course in minutes

I've been handed enough competency frameworks to recognise the list of nouns: "Communication", "Customer focus", "Problem solving". The first thing I try is to design a work sample for each line. If I cannot say what a person would do or produce, the line needs more definition.

Competency mapping identifies what a role requires and describes the knowledge, skills and behaviours at its proficiency levels. Assessment against that map can reveal development needs, process problems or gaps that require other support.

Observable indicators give the map a practical use: a rater can check work against them, and a designer can build suitable practice and assessment. They provide the basis for the task, not necessarily a single quiz question. This guide follows nine steps with one fictional contact-centre team leader example, from evidence gathering to a justified development response.

I'm an instructional designer at Coursebox, where we build AI tools for course and assessment design, including an AI rubric generator for marking work against set criteria. In Step 7, I give the finished map to our course agent and show what it got right, what it got wrong and what I fixed. The rubric was the part that went wrong.

info:What this guide covers

This guide is about the role side: what good performance looks like for a job, written as behaviour at each level. It is not a people inventory and it is not a grid template.

What is competency mapping?

For a working map, connect each competency to observable indicators and relevant evidence. That lets an assessor explain a rating and a manager investigate a gap. The map does not, by itself, decide whether the cause is missing skill, unclear policy or an obstacle in the work.

A competency is the mix of knowledge, skill and behaviour that together produce performance in a job. Take a contact-centre team leader handling an escalated complaint. They need to know the complaints and refunds policy (knowledge), restate a complaint and write a clear confirmation (skill), and stay with an upset customer instead of passing them up the line (behaviour). Any one of the three on its own does not resolve the complaint.

A finished competency map has four parts:

  • Competencies: the capabilities the role needs, each explained through relevant actions or work products. The six in this worked example are a manageable starting set, not a universal limit.
  • Proficiency levels: a scale, such as four levels from Developing to Expert.
  • Behavioural indicators: what a person at each level does, written so a rater can see it.
  • Required level: the level this particular role needs.

Types of competencies

Many frameworks sort competencies into four types. The labels vary, and some frameworks merge two of them, so treat the types as a check that your map is balanced rather than as a fixed taxonomy. Here is one of each for the team leader:

  • Core: applies to everyone in the organisation. Example: handles customer data under the privacy policy.
  • Functional or technical: specific to the work of the role. Example: resolves escalated complaints.
  • Behavioural: how the person works with others. Example: hands over open cases to the next shift with the customer's status and the next action.
  • Leadership: how the person develops and directs a team. Example: coaches agents after a call.

Competency mapping vs skill mapping vs a skills matrix

These terms overlap in practice. This guide uses the following convention to separate the role requirements, the inventory of people's skills and the grid used to record evidence.

TermWhat it describesThe question it answersWhat you end up with
Competency mappingThe roleWhat does good performance look like in this job, level by level?A map of competencies with behavioural indicators and a required level
Skill mappingThe peopleWhat skills does each person on the team have today?An inventory of current skills across a team
Skills matrixThe recordWho is at which level on each skill or competency?A grid of people against skills, filled with ratings

These tools can connect role expectations with evidence about people. An existing competency map can inform a skills inventory, and a matrix can record the ratings. Organisations may combine them rather than build three separate tools in a compulsory sequence.

Why map competencies?

A map gives everyone who makes a people decision the same written description of good work. That shows up in four places:

  • Hiring: interview questions come straight from the indicators. For the team leader: "Tell me about a customer who wanted something outside policy. What did you offer, and what did you do after the call?"
  • Development: a missing behaviour is specific enough to investigate. If a team leader does not confirm the customer's choice in writing, check their skills, time, tools and understanding of the policy before choosing a training or workplace response.
  • Succession: evidence of the work required at a more senior level can identify development candidates. It does not establish readiness for the whole operations role, which may have additional requirements.
  • Performance reviews: managers and staff argue about evidence ("here is the call") instead of impressions ("I feel you're doing well").

A map does not remove bias from any of these. Two managers can watch the same call and see different things, which is why Step 5 includes a calibration check.

Approaches to competency mapping

There are four common ways to build the first draft. They differ in speed and in how close they get to the real job.

  • Job analysis from scratch: you observe the work, interview job holders and managers, list the tasks and group them into competencies. It fits new or unusual roles. It is the slowest approach.
  • Critical incident interviews: you ask people who know the job to describe specific occasions when someone handled a situation very well or very badly: what happened, what the person did and what the result was. John Flanagan set out this critical incident technique in 1954. It is the best source of behavioural indicators I know, because the stories are already about behaviour.
  • Borrowing a published framework: you start from an existing model, such as O*NET for occupations or SFIA for digital roles, and rewrite it for your organisation. It is the fastest approach, but published wording is generic and needs rewriting into your context.
  • A strengths-based lens: tools such as Gallup's CliftonStrengths describe a person's natural talents. They help with development conversations alongside a map, but they describe the person, not the role.

I usually combine the second and third: borrow a framework for structure, then use critical incidents to rewrite each competency in the language of the actual job.

How to do competency mapping, step by step

The worked example runs through all nine steps, and it is fictional. A mid-sized insurer runs a contact centre where five team leaders each manage a team of agents. When a complaint escalates, the agent passes the call to their team leader. The insurer has a written complaints and refunds policy; "within policy" in this example means within those rules, and the example needs no amounts or limits. The operations manager has noticed the same complaints coming back and asks L&D for training on escalations.

Step 1: Set the purpose and pick one role family

Decide what the map is for before you write a word of it. A map for hiring needs indicators you can test in an interview. A map for development needs indicators fine enough to show the next step. A map for succession needs the levels above the current role.

Then pick one role family. Mapping the whole organisation at once produces a long list of generic competencies that fit nobody.

  • Purpose: development, so the indicators must show each step up.
  • Role family: contact-centre team leaders, five people.
  • Sign-off: the operations manager approves the map before anyone is assessed against it.

Step 2: Collect data on what good performance looks like

Use more than one source, because each one misses something:

  • The current job description: fine as a list of duties, but it rarely says what doing them well looks like.
  • Published occupational data: the US O*NET database is a free first draft. Its entries for Customer Service Representatives and for First-Line Supervisors of Office and Administrative Support Workers give a task list to start from.
  • Interviews with subject matter experts: the operations manager and the strongest team leaders.
  • Critical incidents: ask each interviewee for one escalation that went well and one that went badly, and what the team leader did.
  • Recorded or observed work: listen to a sample of real escalation calls.
1,016
occupational titles in the O*NET taxonomy, of which 923 are data-level occupations with detailed descriptions you can use as a first draft
Source: O*NET Resource Center overview (sponsored by the US Department of Labor); these are US occupations, so the wording needs adapting for roles elsewhere

In the example, the critical incidents did the most work. The operations manager described a team leader who offered a customer two ways forward, emailed the choice the same afternoon and noted why the claim had stalled. That complaint did not come back. Another team leader apologised well, fixed the problem on the call and wrote nothing down. The customer rang again a week later, and nobody could see what had been agreed.

Step 3: Define the competencies

Turn the data into a set of competencies that covers the actual role. I often start with six to eight so the first draft is workable, then add or regroup when the evidence requires it. Test whether raters can apply the indicators rather than using the count as a quality rule.

For the team leaders, the list came to six:

  1. Resolves escalated complaints
  2. Coaches agents after a call
  3. Plans the team's roster against forecast call volumes
  4. Handles customer data under the privacy policy
  5. Hands over open cases to the next shift
  6. Runs the team's daily briefing

Writing four levels for all six takes time, so I work one competency fully in this guide, "Resolves escalated complaints", and show a second as a single row so you can see the map has more than one line:

CompetencyTypeLevel 3 indicatorRequired levelHow assessed
Coaches agents after a callLeadershipNames one specific behaviour from the call for the agent to change, plays back the moment it happened, and checks it on the agent's next three callsLevel 3Coaching session observed by the operations manager, plus the agent's next three calls

Step 4: Write behavioural indicators for each level

An indicator describes what a person at one level does, in words a rater could check against a call, a record or a task. If a rater cannot check it, nobody can assess it or train for it. My rules for writing them:

  • Start with a verb you could watch or check: restates, offers, confirms, logs. Not "understands" or "is aware of".
  • Make every behaviour separately checkable: if a level contains three behaviours, the rater marks three things, not one impression.
  • Leave out adjectives: "excellent", "strong" and "effective" mean whatever the rater wants them to mean.
  • Make each level add something: a rater should be able to say what the person did that the level below does not.

Here is the scale for "Resolves escalated complaints". The four levels and their names are ours, not a standard. SFIA, a framework for digital roles, uses seven levels of responsibility. The count is less important than every level being observable.

  • Level 1 Developing: Handles a routine complaint with support and follows the policy for escalating an issue outside their authority.
  • Level 2 Capable: Independently restates and resolves routine complaints within policy, escalating when approval or specialist help is required.
  • Level 3 Proficient: Offers two options within policy, confirms the customer's choice in writing the same day, and logs the root cause in the CRM.
  • Level 4 Expert: Meets the relevant Level 3 requirements and identifies repeat causes across the week's escalations, proposing a process change to the operations manager.
An illustrative competency level matrix showing Developing, Capable, Proficient and Expert; Level 3 is the example target, with required behaviours needing evidence.

View the full-size diagram

The Level 3 line now does two more jobs. As a learning objective: given an escalated complaint, offer two options within policy, confirm the customer's choice in writing the same day and log the root cause in the CRM. As an assessment item: a recorded or simulated escalation, marked on those three behaviours one at a time.

warning:The line I rewrite most often

"Communicates effectively with customers" appears in almost every framework I'm handed. Ask two managers what Level 3 of it looks like and you get two answers. "Restates the complaint in the customer's words before offering anything" gets one answer, and you can hear it on a recording.

Competency map worksheet

An editable Word worksheet for mapping one role. It has the filled "Resolves escalated complaints" row with all four levels, a blank map with one row per competency type, an indicator check that marks each Level 3 behaviour Met, Not yet met, Not observed or Not applicable, and a team record for ratings, gaps and training responses. It is a planning aid with no score, and no email is required.

Step 5: Validate the map before you roll it out

Expert review helps establish that a map covers the work. Before rollout, pilot its clarity and consistency on representative tasks and people. The following checks can expose problems, but a small pilot is not full validation for hiring, promotion or predictions about performance.

  • Try it on representative work. Compare ratings with independent evidence of how the task was performed. Include people with different experience and look for missing requirements, unclear indicators and ratings unsupported by the work. Two people who both meet the standard may correctly receive the same level; do not rewrite the map merely to reproduce a manager's ranking.
  • Calibrate two raters. Have them assess the same work independently, then discuss disagreements against the criteria. In this example, "same day" means by 11:59 pm on the local calendar day of the call. Write that deadline before rating; do not decide it after seeing a person's result. More recordings may be needed to check whether the clarified rule is applied consistently.

Research also supports rating competencies separately rather than as one overall score. Dave Bartram, then at the test publisher SHL, tested selection measures against eight broad competency factors, the Great Eight, one at a time.

29 validation studies
covering 4,861 people were pooled in a meta-analysis that used the Great Eight competencies as the yardstick for testing whether personality and ability measures predicted line managers' ratings
Source: Bartram (2005), Journal of Applied Psychology 90(6); studies from the test publisher SHL, rated by line managers rather than measured on output, with validities of 0.20 to 0.44 for individual competencies

Bartram's meta-analysis found that personality-based measures correlated with managers' ratings on all eight competencies, while ability tests correlated with four. The work supports examining separate criteria, but does not validate this example map. A local pilot checks usability; stronger claims about prediction need suitable evidence beyond that pilot.

A health worker wraps a colour-coded mid-upper arm circumference (MUAC) tape around a young child's arm to screen for malnutrition
Lagos Food Bank Initiative / Pexels

Case study: A competency framework for nutritionists in emergencies

The humanitarian nutrition sector had a recognised need to build capacity and to be clearer about what emergency nutritionists do and the skills they need. A team from UCL's Institute for Global Health, with co-authors from the Emergency Nutrition Network, World Vision, Concern Worldwide, Save the Children UK and others, built a technical competency framework for them.

  • What they did: According to the paper in Public Health Nutrition, they reviewed existing competency frameworks and interviewed people about the methods other sectors had used. They then identified competencies from interviews with field experts, feedback from course trainees, academic course content and job specifications. They wrote behavioural indicators for each competency, and members of the Global Nutrition Cluster reviewed the draft over several rounds.
  • Result: A framework of 20 competency areas with 161 behavioural indicators, sorted into three levels that match progressively more senior roles.
  • Limitation: The authors say it still needed research to validate it, to work out how to assess against it and to measure its effect on performance. It is a well-documented build, not evidence that the framework improved anyone's work.
  • Lesson: Multiple data sources and practitioner review help build a credible draft. For one role, select the competencies its work requires, then pilot whether the indicators are clear and applied consistently. That is a starting check, not proof of predictive validity.

Step 6: Assess people against the map

Choose the method that produces the evidence the indicator describes. For the team leaders, that means the call itself and what was written afterwards:

  • Observation: watching the task live, as in a workplace verification of competency.
  • Recorded work: calls, tickets, emails or CRM entries the person produced on the job.
  • Scenario questions: a realistic case with the information needed to act, answered in writing or out loud.
  • Manager and self ratings: useful for a conversation, but a rating is a proxy. A confident self-rating is not evidence that the behaviour happens.

Mark each requirement separately: met, not yet met or not observed. Use not applicable only for a permitted exclusion, with its reason. A required behaviour that did not occur needs further evidence, not an automatic pass. For this example, Level 3 requires all three stated behaviours and the relevant lower-level capability; an average must not hide a missing requirement.

Here is the example, with the result:

  • Competency: Resolves escalated complaints.
  • Level 3 indicator: Offers two options within policy, confirms the customer's choice in writing the same day, and logs the root cause in the CRM.
  • How it is assessed: One recorded escalation call per team leader, plus the CRM record and any written confirmation sent to that customer, scored by two raters who mark each of the three behaviours separately.
  • Gap: Three of the five team leaders are at Level 2. All three restate the complaint and resolve it within policy, and all three offer two options. None of them confirms the choice in writing the same day or logs the root cause. The other two are at Level 3.

Record the ratings in a skills matrix so you can see the whole team at once. If you run the same exercise across several roles, a skills gap analysis template helps you rank which gaps to train first.

Step 7: Turn gaps into training and development plans

First investigate why the three team leaders miss those behaviours. For this fictional example, the manager confirms that the policy is clear, staff can access the CRM field and have time to use it, but need practice writing the confirmation and recording the cause. Those assumptions justify the focused training response; a different cause would call for a different fix.

  • Training response: One short module for the three Level 2 team leaders on the written confirmation and the CRM root-cause field, ending in a scenario assessment scored on the Level 3 indicator. The two Level 3 team leaders skip it.

Training is not the only response. If the CRM root-cause field is hard to find, fix the system first. A training needs analysis questionnaire helps separate skill gaps from process problems. For more examples of assessment built around observable standards, see our competency-based training examples.

Notes from testing: giving the map to our course agent. Our course agent plans and builds a course from a typed request. I wanted to know whether it would build assessment from the indicators or drift into generic customer-service content. If you only need a bank of knowledge questions, our AI quiz generator is the quicker route, but a Level 3 behaviour needs a scenario and a rubric, so I used the course agent in a test workspace.

What I asked for. "Build a short course for contact-centre team leaders on resolving escalated complaints. Learners are at Level 2 and need to reach Level 3. Make it one module, with a scenario-based assessment and a rubric scored on the Level 3 indicator. Use only the source document below." I pasted the four level indicators and "Role requirement: Level 3" into the prompt, with no file upload.

What it produced. One section with three pages and an estimated 35 minutes, which is longer than I'd call short. The pages were a 10-minute lesson on the move from Level 2 to Level 3, a 10-minute practice page with a three-decision branching scenario, and a 15-minute assessment. The assessment was a written assignment, not a quiz, with three parts headed "Offer of two within-policy options", "Same-day written confirmation" and "CRM root-cause entry". It tested the Level 3 behaviours and nothing drifted into generic empathy questions. It also invented no policy details: no refund limits or amounts appeared anywhere.

What was wrong. The rubric did not match the map. The agent's hand-over note, its summary of what it had built, promised "a single 100% rubric criterion", but the page had two rows, both named "Level 3 Proficient indicator". One was worth 99% ("Full marks require all three observable elements...") and the other 1%. Neither named any of our four levels. The scenario had no actual complaint in it, only an instruction to label two resolutions Option A and Option B, and it contained an unfilled placeholder: "Write the CRM entry using the exact root-cause phrase stated root cause."

The fix, and how I checked it. I sent one message to the AI editor, which changes a single page from a typed instruction. I asked it to give the scenario one concrete complaint (an order delivered to the wrong address twice, with redelivery on a booked date or cancellation with a refund to the original payment method, the customer choosing the refund, and a stated root cause, with no amounts or limits). I picked a delivery complaint because I could describe it in one line; for the insurer's own course I'd use a stalled claim, and the three rubric rows would stay the same. I also asked it to replace the rubric with three rows, one per Level 3 behaviour, with a line under it mapping the result to Level 3 or Level 2. The editor replied: "Replaced the rubric with the three requested Level 3 criteria, weighted equally. Added the Level 3/Level 2 result mapping beneath the rubric." A reply is not a check, so I reloaded the page and read it. The rubric now had three rows: "Offers two options within policy" at 33.33%, "Confirms the customer's choice in writing the same day" at 33.33% and "Logs the root cause in the CRM" at 33.34%. Each row maps to one behaviour in the Level 3 indicator, and none sits outside the map. The placeholder was gone, the wrong-address complaint was in the scenario, and the 99% and 1% rows had been replaced rather than kept.

The fix has limits. The promised Level 3/Level 2 mapping line was absent after reload. The rubric still uses percentages, so one missing behaviour can be hidden by the total. A human assessor would need to apply the all-three rule separately; I did not test that grading or its reporting. The scenario also supplies the root-cause wording. That checks accurate recording of a stated cause, not diagnosis. What I verified was three more specific rubric rows and a concrete scenario, not an operational level-based grading system.

Step 8: Build the map into hiring, reviews and succession

A map kept only in the L&D folder gets used only for training. Put the same wording into the processes where other people decisions get made:

  • Hiring: write interview questions and a work sample from the required-level indicators. For a team leader vacancy, play candidates a recorded escalation and ask what they would do next.
  • Performance reviews: rate against the same indicators and ask for evidence, such as two recorded calls, instead of a general rating.
  • Succession: look for people ready for the level above. In the example, one of the two Level 3 team leaders is a candidate for Level 4 development; the next step is reviewing a week of escalations for repeat causes, which also prepares them for the operations role.
  • Job descriptions and promotion criteria: use the indicators word for word, so the role someone applies for is the role they are assessed against.

Ask HR to own the wording in these documents, so a change to the map flows into all of them at once.

Step 9: Review the map on a schedule

Set a review date, but also review when something changes the job. Triggers I'd watch for:

  • A new system: the insurer replaces its CRM and the root-cause field moves or disappears.
  • A change to the role: team leaders take on quality checks that used to sit with a separate team.
  • A new rule: a regulator changes how complaints must be handled or recorded.
  • Results that look wrong: almost everyone rates at Level 4, or nobody moves after training.

If you train within the Australian vocational system, the standard is a unit of competency, a nationally published description of one workplace task and how it is assessed. You map your training and assessment to it rather than rewrite it, and you review when a new release of the unit comes out.

Competency frameworks you can borrow

Starting from a published framework is faster than a blank page, as long as you rewrite its indicators for your own roles. These are the ones I see most often, with when I'd look at each:

  • APS Integrated Leadership System (ILS): the Australian Public Service framework. The ILS guide for APS 1 to 6 employees sets out five capability clusters (supports strategic direction, achieves results, supports productive working relationships, displays personal drive and integrity, and communicates with influence) and describes behaviours at each APS level. It is the clearest public example I know of behaviour written level by level.
  • SFIA: the Skills Framework for the Information Age, for digital and IT roles. It defines seven levels of responsibility, and its professional skills are aligned with those levels.
  • European e-Competence Framework (e-CF): a European framework that, as the EU's ESCO description of the e-CF puts it, classifies around 40 competences for ICT professionals in five ICT business areas, with proficiency levels. It is maintained by the European standards committee CEN/TC 428 and suits ICT roles in Europe.
  • Lominger: originally 67 competencies, developed by Michael Lombardo and Robert Eichinger. The library is now owned by Korn Ferry. Look at it if your organisation already uses Korn Ferry assessments, so the wording matches what managers have seen before.
  • SHL Great Eight: eight broad competency factors from the test publisher SHL, the model behind Bartram's meta-analysis in Step 5. It suits selection, where you need competencies you can predict.
  • DDI: Development Dimensions International, a leadership consultancy with its own competency model. I have not checked its structure at source, so if you are considering it, ask to see the behavioural indicators at each level before you commit.
  • ATD Talent Development Capability Model: the Association for Talent Development's model for people who work in talent development. Look at it if the role you are mapping is in L&D.
  • CliftonStrengths: Gallup's assessment of 34 CliftonStrengths themes. It assesses a person's talents rather than a role's requirements, so use it beside a map, not as one.

Common mistakes

  • A list of nouns with no indicators. "Teamwork" with no levels cannot be assessed or trained.
  • An unmanageable list. If raters cannot work through the indicators reliably, regroup them or review the assessment process. A long framework may still be appropriate for a broad role or sector.
  • Rating attitude. "Positive attitude" cannot be seen. Describe what a person with that attitude does.
  • One manager's view as the only data. One person's picture of good work, with no critical incidents or recorded work behind it, bakes their preferences into the map.
  • A map nobody reviews. The job changes, the indicators don't, and ratings drift away from the work.

Where competency mapping came from

The modern idea of competency goes back to the psychologist David McClelland. In 1973 he published "Testing for competence rather than for 'intelligence'" in American Psychologist. As the title says, he argued for testing competence rather than intelligence. Flanagan's critical incident technique, published nearly 20 years earlier, became one of the main ways to find out what competence in a particular job looks like.

Putting your first map to work

A competency map pays off when it changes a decision: who gets trained, on what, and who is ready for the next role. That depends on the indicator lines. A map of nouns produces a document. A map of observable behaviour produces assessments and training you can check.

Start with one role and a competency relevant to its work. Collect examples of performance, write the indicators and design an assessment task for the required level. Pilot it on representative work with two raters. Refine unclear wording and investigate the causes of gaps before adding training or extending the map.

Frequently Asked Questions

Competency mapping identifies what a role requires and describes the knowledge, skills and behaviours at its proficiency levels. Assessment against that map can reveal development needs, process problems or gaps that require other support.

Set the purpose and role scope, gather evidence of the work, define competencies and observable levels, then pilot the map and compare raters. Assess people against the criteria, investigate the causes of gaps and choose suitable support. Review the map when the work changes.

The terms overlap. This guide uses competency mapping for describing role expectations and behaviours by level, and skills mapping for connecting needed skills with what people can demonstrate. An existing role framework can help, but a separate competency map is not a prerequisite for every skills map.

Core competencies apply to everyone in the organisation, functional or technical competencies are specific to a role's work, behavioural competencies cover how a person works with others, and leadership competencies cover developing and directing a team. Labels vary between frameworks.

Set a regular review date, such as once a year, and also review whenever a system, the role or a regulation changes, or when ratings stop matching what managers see in the work.

Carolina Martin, Learning Experience Manager at Coursebox AI

Carolina Martin

Customer Success Manager

Carolina is a highly experienced learning designer with tech start-up experience. She is based in Sydney and previously designed courses alongside Coursebox founder Travis in a previous job. She is passionate about delivering engaging learning experiences and brings the same perspective to customer experience too. She manages our customer support documentation and product testing processes.