Competency framework

Goal

A competency framework makes a decision about a person defensible. It writes down in advance what the role requires, competency by competency and level by level, then places the individual on that scale from observed behaviour. It answers two questions: does this person reach the level the role asks for? On which competency is a level missing?

The decisions it supports all have a bar to clear, and in each of them the gap between the expected level and the observed level is the useful result, because it names the action.

The Johari window answers a neighbouring and distinct question, that of what a person knows about themselves and what those around them observe. Its product is a development conversation. The framework produces a level per competency, comparable from one person to another and from one year to the next. A blind spot named in a feedback session will feed a later assessment.

The deliverable is twofold: the framework itself, a grid of competencies and levels with one indicator per cell; then, for each person assessed, a dated observed profile carrying the gaps and the actions they trigger.

Usage

When to use it

• Recruiting for a poorly defined role: write the expected profile before the advertisement goes out, so that every interview turns on the same criteria.
• A promotion decision: place two internal candidates on the same scale, each backed by observed facts.
• Staffing a project team: spot the missing competencies before the roles are allocated.
• Process reorganisation: the target roles change, and their competency requirements are rewritten with them.
• The annual development plan: aggregating the observed gaps gives the real demand for training.
• Preparing for a Federal Professional Examination or a certification: compare the candidate's profile with the qualification profile required.

When not to use it

• A development conversation with no expected level: nothing to compare against, open a Johari window to work on perceptions.
• Assessing the results of a period: the framework places behaviour on a scale, take the role's performance indicators to judge whether quantified objectives were met.

Description

The elements

A competency is an observable characteristic that can be tied to performance in a role. Spencer and Spencer separate the visible part, knowledge and skills, from the submerged part, motives, character traits and self-image, hence their iceberg image. The visible part develops through training in a few months; the submerged part develops slowly and is settled mostly at selection.

Competencies are grouped into domains, also called clusters: analysis, communication, business knowledge, tools. The domain serves to navigate the framework and to compose a target profile without leaving out a whole family.

The proficiency scale orders three to five levels, from beginner to expert. Below three, the scale stops distinguishing anything. Above five, assessors agree poorly on the boundary between two neighbouring levels. SFIA, a framework aimed at IT occupations, has seven and compensates with five generic descriptors of autonomy, influence, complexity, knowledge and business skills, to be read profession by profession.

The behavioural indicator is the sentence that places somebody on that scale: it describes what the person does, under what conditions and with what reach. "Prepares and runs the interviews of a domain on her own" can be verified; "a good communicator" cannot. One is needed per level and per competency, which takes up most of the drafting work and makes two assessments comparable.

The target profile sets the level expected of each competency selected for a role; the job description, for its part, sets out the tasks of the post. One framework therefore serves several roles, each drawing its competencies and its target levels from the shared grid.

The grid of levels

The framework reads as a matrix: one competency per row, one level per column, one indicator in each cell.

Building the framework

1. Set the scope
   One role, a family of roles or the whole organisation. A company-wide framework built in one go takes a year and comes out obsolete; start with the roles on which a decision is expected.
2. Identify the competencies
   Spencer and Spencer's method starts from the Behavioural Event Interview, McClelland's adaptation of Flanagan's Critical Incident Technique: it is run with incumbents known to perform well and with a comparison group, and the behaviours that separate the two groups are then kept. Failing that, adopt a published framework and cut it to the context, which saves the drafting but calls for the same validation.
3. Fix the scale
   The number of levels, their names and one generic descriptor per level, identical for every competency.
4. Draft the indicators
   One per cell, in the present tense, in the active voice, with a verifiable object.
5. Validate the grid
   A review in a workshop with role incumbents and their line management, then a test on three people whose expected level is already known.
6. Set the target profiles
   The level expected per competency, decided by the owner of the role. A profile with every competency at the top level describes a post that never gets filled.
7. Publish and revise
   The framework is open to the people it assesses, with an annual review.

The qualification profile, the Swiss version

Swiss vocational education and training keeps a competency framework per occupation at national scale, under other names. For each initial programme, a VET ordinance issued by SERI sets the legal frame of the occupation. The training plan, drawn up by the professional organisation and approved by SERI, describes how it is carried out. It contains the qualification profile, which lists the operational competencies grouped into domains, with the level of requirement expected for each. An IT apprentice on a Federal VET Diploma programme is assessed against that profile, the same one in every host company in the country. Above it, the Federal Professional Examination and the Advanced Federal Professional Examination each carry their own qualification profile, and a candidate measures herself against it before entering.

Assessing a person and reading the gap

The assessment sets a person against the profile of their role on the strength of facts: a deliverable produced, a situation handled, feedback from a stakeholder. Three arrangements combine according to what is at stake: self-assessment, assessment by the direct line manager and, for heavy decisions, an assessment drawing on several sources or one run externally.

Calibration decides the worth of the result. Two or three assessors go over the borderline placements, each backed by a named piece of evidence.

The gap is read per competency: expected level minus observed level. A one-level gap on a competency is handled by coaching or by a supervised assignment. A two-level gap repeated across several competencies says that the role does not suit this person at this date, and the decision then turns on recruitment. Overshoot is read too: a competency above target signals a person under-employed or a target profile badly set.

What makes a framework fail

Too many competencies

A grid of thirty competencies cannot be held in mind, calls for a day of assessment per person and falls out of use within the year. Eight to twelve competencies per role are enough to decide on. What does not fit in the grid stays in the job description.

Do two neighbouring levels say two different things?

"Masters process modelling" at senior level and "masters process modelling well" at expert level make a grid that measures the assessor's generosity. The drafting test: hand two people of different levels to somebody who does not know them, to be placed on the reading of the indicators alone. Two different cells have to come out.

Self-assessment without evidence

Left to itself, self-assessment drifts towards the top of the scale and the gaps disappear. It becomes usable again when every level claimed calls for a fact: the name of the project, the deliverable, the date, the witness.

Published then shelved

A framework built for one recruitment round and never revisited ages at the speed of the roles it describes, and the assessments it carries stop being comparable from one year to the next. Tying it to the HR calendar, the annual appraisal and the development plan makes it last.

Mistaking it for a certification

The framework describes what each level covers; it certifies nobody. A certification calls for an examination arrangement of its own, with its evidence and its board, which the Federal Professional Examination and the professional business analysis certifications organise around their profile. A company announcing analysts "at level 3" without calibrated assessment is publishing an opinion.

AI considerations

A conversational assistant speeds up the drafting. From a job description and a few transcribed interviews, it proposes a first list of competencies and a set of indicators per level, which the validation workshop then goes over line by line. It is also a second reader of the scale: submitting the four indicators of one competency and asking what separates level 2 from level 3 brings out the cells that do not discriminate. Finally it maps two frameworks onto each other, an acquired entity's against the acquirer's or an in-house grid against a published one: a long and verifiable reconciliation job.

The assessment stays human. Placing a person rests on facts that only her colleagues and her manager have observed, and a model that has observed nothing will produce a plausible and wrong rating. Observed profiles are personal data within the meaning of the Federal Act on Data Protection, filed in a record that sometimes decides a promotion: processing them in an online service is settled before the assessment round opens, with the data protection adviser. A model trained on a company's job descriptions reproduces its biases, and a competency worded in the image of the current incumbent shuts out the candidates who would do the work differently.

Examples

A health insurer based in Lausanne opens a senior business analyst post and assesses an internal member of staff against the target profile of the role. Two assessors run the exercise: the head of the department and the manager of the last programme she worked on. Every level claimed rests on a dated fact, the model of the benefits re-invoicing process for modelling, the scope agreement obtained between the benefits department and customer service on the policyholder portal for negotiation. The two placements that diverge are settled in a calibration session, the day before the annual appraisal.

Four competencies out of five reach the target, one of them above it. The only gap is on negotiation with stakeholders, observed at level 2 against a target of 3: she obtains scope agreements between two departments of the same division; she has not yet led an arbitration between divisions. The gap holds on one competency and on one level, which points the decision towards coaching: the promotion is granted, carrying six months of follow-up by the project portfolio manager, and the next arbitration between divisions is handed to her in a pair. The overshoot in data analysis goes into the record too, since it names the person to hand the domain's reporting requests to.

Visualisations

The grid of levels is a matrix of text values in which the reader hunts for one particular cell: a table carries it without loss. The assessment is judged on a comparison, two values per competency whose gap is the result. A chart puts the two values side by side and makes the gap visible at a glance, where a column of figures forces the reader to work it out row by row.

Cost

Tooling

A spreadsheet is enough and remains the most widespread tool: twelve competencies over four levels fit in one sheet, the target profiles in a second, the assessments in a third. Its limit arrives once the observed profiles run to the hundreds and have to be aggregated to build a development plan.

The talent management modules of HR software, Abacus, SAP SuccessFactors or Workday, host the framework, tie it to the posts and keep the history of assessments. In return they impose their own scale structure, to be checked before the grid is drafted.

SFIA covers IT occupations and lends itself poorly to roles outside IT. The IIBA's Business Analysis Competency Model describes the business analyst's competencies with their indicators and their levels, under member access. For an occupation covered by a VET ordinance, the training plan published by SERI supplies a qualification profile already validated and kept up to date.

Sources

• McClelland, D. C. (1973), Testing for Competence Rather Than for Intelligence, American Psychologist, 28(1), 1-14: the founding article, which shows the limited predictive power of aptitude tests over performance at work and proposes putting the assessment of observable competencies in their place.
• Boyatzis, R. E. (1982), The Competent Manager: A Model for Effective Performance, New York, Wiley: the study of two thousand managers in forty-one jobs that established competency modelling as a method, with a model of nineteen competencies in four clusters.
• Spencer, L. M. and Spencer, S. M. (1993), Competence at Work: Models for Superior Performance, New York, Wiley: the reference for building a framework, holding the definition of competency, the iceberg model, the competency dictionary, the Behavioural Event Interview and the graded behavioural indicator.
• SERI, list of occupations in initial vocational education and training (de, fr, it): access, by occupation, to the VET ordinance and to the training plan that carries the qualification profile, the Swiss version of the technique.
• SFIA Foundation, How SFIA works: the description of the seven levels of responsibility and of the generic attributes of the IT competency framework, cited as a widespread example.