Same keywords, different careers: how HireSalt reads a career arc

Put Candidate A and Candidate B side by side and the skills lists match. Both list Python, SQL, experimentation, dashboarding, stakeholder reporting and mentoring. Both have six years of work, and both now hold a Lead Data Scientist title. Read those facts alone and the two look interchangeable.

Both candidates and their job histories are invented for this article. The figures are what HireSalt’s scoring arithmetic returns for those histories. No real applicant appears here.

The difference is in the order of the jobs. You only see it when you open each candidate’s page.

The two work histories

Candidate A starts as a Junior Data Analyst. Each of the next three roles is one level up: a Reporting Analyst, a Senior Data Scientist, then the Lead Data Scientist role. Each role lasts about 18 months. Scope grows at every step as well. Suppose the first role lists one scope phrase, the last lists four, and the two in between list two and three.

Candidate B also has four roles of about 18 months. The first three sit at the same level: Data Analyst, Reporting Analyst and Analytics Engineer. The fourth is the Lead Data Scientist title that ends A’s history. B’s scope phrases stay at one for three roles and rise to two at the end.

What you see on the page

Open Candidate A’s page and scroll to Career Story. The panel has four parts: a header, a timeline of the last six roles at most, a Career Arc column and a Skill Arc grid. Gaps and flags sit at the foot of the Career Arc column. They stay folded until you open them, and they appear only when there is something to show.

The header gives you an Advancement score out of 100, a fit label, a seniority label and a career anchor. Both candidates get the same right-pointing arrow beside their score. You might read that as steady progress. It is not. The arrow measures whether the latest role ran about as long as the earlier ones. It does not measure whether the person climbed, so it cannot tell these two apart. Seniority reads Ascending for both, because neither history steps down.

The timeline is where the two diverge. Each role is a node with its level code, and each step up is marked PROMOTED with the levels gained. On A you see three promotions of one level each. On B you see three nodes at IC2, then one jump of two levels at the end.

The Career Arc column holds an arc summary, four measures and the Natural next roles. The Skill Arc shows how the candidate’s skill profile has grown and shifted over time. Each row is a skill and each column is a recent role. A filled dot marks where a skill first appears, a green dot marks one carried over from an earlier role, and an empty ring marks a role without it. Because both lists match, the Skill Arc is where you would see when each skill arrived.

Where the score comes from

The Advancement score is a weighted mix of four measures, and progression carries the most weight. Progression is levels gained per year. A gained three levels in six years and B gained two, so A scores 100 and B scores 66. Scope is next. It checks whether each role lists more scope phrases than the one before. A rises at every move. B rises once. Tenure marks very short stints down, so both candidates score the top mark for 18-month stints. Growth looks at how far the level moved overall, scaled by years worked.

Here is what the cards and header show.

Tenure cannot separate them. Scope and Progression do most of the work.

What the model writes

The model extracts the timeline from the resume. The arithmetic runs on that timeline and takes nothing from the job description. In the same request, the model also writes the fit label, the next roles, the career anchor, the flags and a short arc summary.

The fit label is judged against the job the resume was scored for. So the same history can read differently when you open it against a different posting.

The panel lists up to three Natural next roles, each tagged high, medium or low confidence. Treat them as the model’s view of where the history points. If the role you are hiring for appears in that list, the arc lines up with it. If it does not, the match may still be good, but the case is less direct and worth probing on the call. These roles do not predict how someone will perform or how long they will stay.

The flag

On B’s panel you see one flag: Lateral moves only. A pattern check raises it when three roles in a row sit at the same level, which is what B’s first three jobs do. A has no pattern flag. The model can still add flags of its own, so read the list rather than assume it is empty.

Career Story does not change the score

Career Story runs after scoring. The match score is already set when it starts, and Career Story writes only its own field. If that step fails, the score stays as it was and the panel does not appear. Career Story is narrative context. It does not predict outcomes. For how the criteria score is built, read how HireSalt scores a resume. Every part of the panel is described in the Career Story help page.

Two limits

Keep two limits in mind when you act on the panel. Scope counts phrases and does not weigh them. Two phrases about a large team score the same as two about a small one. The levels rest on the model reading each title, and total years come from the resume too, so a misread title moves the numbers. Very long resumes are also cut short before the model reads them, so roles near the end can drop out.

On the phone screen

Ask each candidate something different. With A, ask what the four scope phrases on the last role cover, since the scope climb ends there. With B, ask about the three roles at the same level. Ask what each one was for, and whether the move to Lead was the first real change in scope.

Both candidates are Lead Data Scientists now. The skills list cannot show which one spent three roles at the same level on the way. The timeline can.