EN
Four-step wooden staircase with a black metal frame and one blue crossbar.

Two levels no one can distinguish are not two levels

If reasonable assessors cannot explain which evidence separates two adjacent levels, the architecture contains a nominal distinction—not a usable progression.

PRYSMAP4 min read

The symptom often appears during calibration. One manager places someone at the upper end of intermediate; another places them at the lower end of advanced. Both cite the same evidence, and neither can identify which condition changes between the categories. The conversation ends in a negotiation over the number.

1. Define the variable that progresses

A level must change a recognizable demand. The progression may increase autonomy, problem complexity, the reach of consequences, depth of mastery, or the required consistency.

The SFIA framework separates professional skills from levels of responsibility and describes the latter through autonomy, influence, and complexity, along with knowledge and behavioral factors. As a reference, its value lies in making the dimensions of progression explicit.

Not every dimension needs to advance at the same rate. In a technical skill, depth may change before scope. In coordination, the number and diversity of dependencies may be more consequential. The architecture should follow the work rather than impose perfect symmetry.

2. Preserve the same object across levels

If the initial level describes conducting analysis and the next describes managing projects, the skill has probably changed. Progression must preserve the object while raising the demands placed on it.

For example:

  • applies a defined method to familiar problems and recognizes when support is needed;
  • selects among methods when unforeseen variations arise and justifies the choice;
  • establishes criteria for ambiguous problems, assesses limitations, and guides other people’s decisions.

3. Replace adjectives with conditions

Solid, broad, advanced, and excellent do not define a boundary. They may accompany a definition, but they cannot replace one.

Use observable conditions: the type of problem, degree of direction, consequences of the decision, range of contexts, stability of performance, and responsibility for other people’s work. The OPM structured interview guide recommends behavioral examples at each level to support concrete distinctions among degrees of proficiency.

A good anchor allows the reader to imagine both evidence that meets the standard and similar evidence that does not yet meet it.

4. Confirm that a plausible step exists

There should be a recognizable development path between two levels. If the jump requires a simultaneous change in technical mastery, team size, executive influence, and financial responsibility, the architecture may have skipped a stage or combined several dimensions without establishing a hierarchy.

The opposite can also happen: two levels describe virtually the same work and differ only in the frequency of words such as generally and consistently. If that distinction does not change any decisions, the levels may need to be merged.

5. Design evidence for the boundary, not just the center

Prototypical examples are easy: almost anyone recognizes very early-stage or exceptional performance. The quality of a level emerges in borderline cases.

Prepare samples close to the boundary and ask:

  • which evidence shifts the interpretation toward one level?
  • which information is missing?
  • which condition would need to change the conclusion?

6. Test agreement and explanation separately

Two assessors may agree for different reasons. Others may disagree because they observed legitimately different contexts. Calculating agreement alone is therefore not enough.

Ask each person to record the evidence, the expectation applied, and the rule used. Then compare both the level and the reasoning. Reliability concerns consistency; defensibility requires the conclusion to be reconstructable. Neither can substitute for the other.

Three concrete platforms at different heights, separated by a blue line.

7. Remove levels that do not change a decision

Every level has a cost: it increases maintenance, calibration effort, and the sensitivity of conversations. It must provide a distinction that changes expectations, development, mobility, or the assignment of work.

If no one can observe the difference, or both levels lead to the same actions, retaining them adds only an appearance of precision. A shorter scale may be an improvement, not a loss of sophistication.

The final criterion is demanding and simple: every level must state what changes in the work, which evidence would reveal that change, and which decision justifies preserving the distinction.

  4. The cover of Sorting Things Out on a wooden table beside a small paper box.

    Sorting Things Out: every taxonomy describes and decides

    The book Sorting Things Out examines how categories and standards organize the world while distributing visibility, work, and consequences. Reading it through the lens of a skills architecture extends that inquiry to the decisions a taxonomy makes possible.