From Profile to Gap: Comparing Possible Destinations
This chapter examines how to compare a source occupation with a possible destination once both are represented as skill profiles. The useful result is not a single “match” percentage. It is a visible account of what carries over, what changes, how large the descriptive gaps are, and which facts outside the profile could still overturn the conclusion. The comparison supports a transition hypothesis; it does not certify a worker or choose a career for them.
Compare coordinates, not job names
A profile comparison begins by putting the source and target occupations on the same set of coordinates. In the BLS data, that means reading the same 17 EP skill dimensions for both roles. The comparison can ask, for example, whether the target places more emphasis on mathematics, analytical thinking, computers and information technology, writing and reading, interpersonal work or physical strength than the source does. It can also reveal continuity that a title-based reading would miss: a target may preserve adaptability, detail orientation or problem solving even while moving into a different occupational family.
Consider Data entry keyers and Statistical assistants. The source occupation has an EP score of 4.0 for detail orientation, 4.1 for adaptability, 3.6 for computers and information technology, 2.9 for critical and analytical thinking, and 2.3 for mathematics. Statistical assistants have a somewhat different pattern: 3.7 for detail orientation, 3.7 for adaptability, 4.1 for computers, 3.5 for critical and analytical thinking, and 4.2 for mathematics. The most visible change is not the title. It is the shift toward quantitative and analytical work, accompanied by a reduction in fine-motor emphasis.
That observation can be turned into a concrete transition question. Has the person done work that involves interpreting numerical information, checking a result, identifying an anomaly or explaining a finding? If not, what practice would create that evidence? A profile does not answer those questions, but it makes them harder to avoid. “Administrative experience transfers” is too vague; “the target asks for more mathematics and analytical reasoning than the source profile” is a hypothesis that can be tested.
The same comparison also corrects a common assumption about distance. Using the 17 coordinates, the unweighted Euclidean distance between Data entry keyers and Statistical assistants is 2.66. That calculation squares the difference on each dimension, adds the results and takes the square root. It gives a compact measure of how different the two occupation-level patterns are in this particular representation. It does not say that the move requires 2.66 units of training, or that it is impossible. It says that the broad profile changes substantially enough to deserve a deliberate bridge.
Now compare Data entry keyers with Information security analysts. The distance in the same descriptive calculation is 1.69—smaller than the distance to Statistical assistants—even though the job titles appear to point to very different careers. Both profiles place high importance on adaptability and computer work. The target has higher critical and analytical thinking and writing and reading scores, while the source has greater fine-motor and physical-strength emphasis. The result does not make cybersecurity an obvious next job for a data-entry worker. It reveals a less visible fact: the title difference is not a reliable estimate of the difference in these broad capability requirements.
The comparison is useful because it gives the gap a shape. Some dimensions are nearly continuous; some move upward; some move downward; and some may be irrelevant to the decision if they do not correspond to a critical target task. A person who has worked with records, detected errors and followed strict procedures may already have evidence related to detail orientation and adaptability. That evidence is not the same as cybersecurity knowledge. The profile helps locate the continuity, while the target occupation's actual tasks determine what must be demonstrated next.
The numbers also need a direction. In the BLS workbook, Data entry keyers are projected to fall from 141.6 thousand jobs in 2024 to 104.9 thousand in 2034, a 25.9% decline. Statistical assistants are projected to fall by 2.5%, while Information security analysts are projected to grow by 28.5%. The profile comparison and the employment projection therefore tell different parts of the story. The first describes broad requirement patterns; the second describes the expected change in the occupation's size. A destination can be relatively close in profile and still offer weak prospects. A destination can be attractive on the outlook axis and still demand a bridge that the profile alone does not describe.
The mechanics of a comparison can be pictured as a sequence:
The last box is deliberately conditional. It is the point at which a comparison becomes useful for action without pretending to be a prediction.
A distance is a question, not a verdict
A single distance is easy to misunderstand because it hides the contributions of individual dimensions. Two pairs of occupations can have the same total distance for different reasons. One may differ mainly on mathematics and analytical reasoning; another may differ on physical demands and interpersonal work. The practical meaning of the gap depends on the target's tasks, the person's evidence and the constraints of the work setting.
For that reason, distance is best treated as a way to find the next question. After calculating or visually inspecting a difference, ask which coordinates account for it and whether those coordinates are central to performance. In the Data entry keyers to Statistical assistants comparison, mathematics rises by 1.9 points, while fine motor falls by 1.3. If the proposed move truly involves statistical work, the mathematics gap is likely to be more consequential than the reduction in fine-motor emphasis. But the BLS score does not tell us which specific mathematics, software or statistical concepts are needed. That requires looking at target tasks and employer evidence.
The same logic applies when the profile gap is smaller. Customer service representatives and Medical secretaries and administrative assistants have a descriptive distance of 0.89. The target has a 0.4-point higher interpersonal score and a 0.4-point lower mathematics score, while adaptability remains similar. Their projected employment directions differ: customer service representatives are projected to decline by 5.5%, while medical secretaries and administrative assistants are projected to grow by 4.2%. This is a plausible example of why adjacent work can be worth examining. It is not proof that every customer-service worker can enter medical administration, because the target may require familiarity with healthcare workflows, privacy practices, scheduling systems or employer-specific standards that the broad vector does not record.
Nursing assistants and Home health and personal care aides offer another contrast. Their distance is 0.65, and the target has a 17.0% projected increase compared with 2.3% for nursing assistants. Their broad interpersonal and adaptability profiles are close, but the work settings, responsibilities and employment arrangements can still differ materially. A small gap may reduce the amount of new capability that needs to be built; it does not eliminate the need to inspect the work itself. A person may prefer one setting, lack access to another, or find that the physical and scheduling demands change the value of the move.
The vector can also show why a title is not a demand forecast. In the retained BLS data, Statisticians and Data scientists have the same 17-coordinate skill vector, yet their projected employment changes are 8.5% and 33.5%. Customer service representatives and service sales representatives likewise share a vector while their projections differ, at -5.5% and 3.1%. Within this representation, the occupation profiles are indistinguishable on the selected skill coordinates, but the employment outlook is not. That does not mean the occupations have the same tasks, credentials or pay. It means the vector and the projection are measuring different properties.
This is why a profile-only recommender would be overconfident. It might find a mathematically close target while overlooking a licence, degree, security clearance, portfolio, supervised-practice requirement or physical condition. It might also treat a broad occupation average as if every employer used the same tools and standards. Conversely, a title-only recommender could discard a potentially useful path because it fails to recognise shared capabilities. The correct response to either error is not to abandon structure. It is to add the missing checks.
The first check is task criticality. Which target tasks depend on the changed dimensions? A 0.5-point difference in writing and reading may matter greatly if the target role produces regulated records or decisions that must be explained. A similar difference in a less central dimension may matter less. The BLS profile does not supply that task-level weighting, so the reader must not imply that all coordinates are equally important simply because a distance formula treats them that way.
The second check is individual evidence. A target requirement is a demand placed by the role; a person's evidence is what they can demonstrate. Evidence may come from a work sample, a completed project, an assessment, a reference, supervised practice or an outcome that another person can inspect. A person who has used spreadsheets is not automatically a data analyst. A person who has resolved complex customer cases is not automatically a medical secretary. But those experiences may be useful starting evidence if they can be connected to target tasks and supplemented where necessary.
The third check is entry feasibility. The supplied BLS skill profile does not tell us the education, experience, licensing, location or employer-specific technology conditions for a particular opening. The OECD evidence shows why education and field of study matter to the skills people bring into work, while also warning that credentials do not fully determine occupational mobility. A comparison should therefore ask what gate applies, whether it is legally required or merely customary, how it can be met, and whether an alternative form of evidence is accepted.
The fourth check is opportunity direction. Employment projections are not guarantees, but they prevent a profile comparison from being read in isolation. If the target is shrinking, a close match may still be a poor long-term destination. If the target is growing, that growth may come with rapid skill change, strong competition or demanding entry conditions. The projection tells us which direction deserves investigation; it does not settle the decision.
The fifth check is work context. A transition can change autonomy, schedule, physical demands, pay, location and access to further learning even when the vector looks favourable. The same broad interpersonal requirement can be expressed in a clinic, a call centre, a care home or a sales environment. The work setting determines what “good performance” looks like and which capabilities can be practised. It also affects whether the person can sustain the move.
A compact way to use the comparison is to ask four questions in order. What remains continuous between the source and target profiles? Which changed dimensions are central to the target tasks? What does the separate employment outlook suggest? What evidence or entry condition could still change the conclusion? If those questions cannot be answered, the distance score is not yet decision-ready.
There is an additional reason to resist a verdict. The BLS vector is an occupation-level aggregate, not a measurement of the individual. It can reveal that Information security analysts place more emphasis on analytical thinking and writing than Data entry keyers, but it cannot reveal whether a particular data-entry worker has already developed those capabilities elsewhere. It can show that a target role is projected to grow, but it cannot show whether the worker has access to the training or networks needed to compete for it. It can identify a promising bridge; it cannot walk the bridge.
The next stage is therefore not to find the nearest title. It is to combine similarity with direction and feasibility. The series' central distinction now has a practical form: profile closeness tells us where continuity may exist, the gap tells us what to investigate, and the outlook tells us whether the destination deserves attention. Before learning design and labour-market change are added in full, that three-part judgment needs to be tested as a reusable way of reading any proposed move.