Where these numbers come from
A benchmark is only worth something if you know how it was made. Here is exactly what is calculated, what is estimated, and what an AI model touches — including the parts we consider weak today.
The short version
Your percentiles are computed by ordinary, deterministic code — the same input always produces the same output, and no AI model invents a figure. What varies is the reference data those calculations compare you against: some of it is measured by an official statistical agency, and some is still curated by us. Every salary figure on your report is labelled with which one it is.
Which figures are sourced today
Salary percentiles for the roles below come from the U.S. Bureau of Labor Statistics Occupational Employment and Wage Statistics survey — United States (national, cross-industry), survey year 2025, retrieved 2026-09-01. These are measurements of actual US wages, not estimates.
| Role | Matched occupation (SOC) | Match quality |
|---|---|---|
| Financial Analyst | Financial and Investment Analysts (13-2051) | direct match |
| Marketing Manager | Marketing Managers (11-2021) | direct match |
| Account Executive | Sales Representatives, Wholesale and Manufacturing (except technical products) (41-4012) | closest available |
| Operations Manager | General and Operations Managers (11-1021) | closest available |
| HR / People Manager | Human Resources Managers (11-3121) | closest available |
| Data Scientist / Engineer | Data Scientists (15-2051) | direct match |
| Product Designer | Web and Digital Interface Designers (15-1255) | closest available |
| Product Manager | Computer and Information Systems Managers (11-3021) | closest available |
| Backend Engineer | Software Developers (15-1252) | direct match |
| Frontend Engineer | Software Developers (15-1252) | direct match |
| Fullstack Engineer | Software Developers (15-1252) | direct match |
One caveat we apply to all of them: BLS publishes the spread of pay for an occupation but does not label seniority. Reading your level off that curve (entry → 10th percentile, mid → median, senior → 75th, lead → 90th) is our modelling assumption. The dollar figures are measured; the level mapping is ours.
What AI does — and doesn't do
The engine is deliberately split so that a language model never produces a number.
| Step | Who does it | Can it be wrong? |
|---|---|---|
| Reading your resume | AI model (or a keyword parser if no key is set) | Yes — it can misread a job title, miss a skill, or misjudge seniority. Check the role and years shown on your report. |
| Percentiles & salary bands | Deterministic code | Reproducible — no AI involved. Same input, same output, every time. |
| Skill gaps & course matches | Deterministic code | Reproducible — derived from the reference dataset, not generated. |
| The written summary | AI model, given only the computed figures | Wording only — instructed to use the supplied numbers and invent none. Phrasing may still miss nuance. |
How a percentile is calculated
We place you inside a comparison group — role × level × country — and score four axes independently:
| Axis | How it's scored |
|---|---|
| Skills | Share of in-demand skills for your role that appear in your resume, weighted by how often each is expected at your level. |
| Experience | Your years mapped onto the typical range for your level. |
| Salary | Your pay placed in the estimated band (p10 → median → p90). Without a salary we use a neutral midpoint rather than guessing. |
| Resume strength | Presence of concrete achievements and quantified results — a readability signal, not a market measurement. |
The overall score is a weighted blend: skills 30%, salary 25%, resume 25%, experience 20%.
The honest limitations
We would rather you know these than discover them later.
| Limitation | What it means for you |
|---|---|
| Skill-demand data is estimated weakest point | The “% of postings” figures behind skill gaps and course matches are approximate anchors we curated, not measured from live job postings. They are plausible, not authoritative — unlike the salary bands above, which are measured. Sourcing them is our top priority. |
| Skill coverage is much narrower than pay coverage | 11 role families, United States only. If your job doesn't match one closely, you are compared to the nearest fit — and the result is weaker for it. United States only either way. |
| No employer or industry adjustment | Company size, industry, city and equity are not modelled. Pay in New York finance and small-town nonprofits is not the same market. |
| A resume isn't a person | We score what your document says. Skills you never wrote down are invisible to us. |
| Projections are illustrative | The 1–3 year career track shows a plausible path at typical pay levels. It is not a forecast and not a promise. |
Where we're taking this
Replacing curated anchors with sourced data is the roadmap: public salary statistics, aggregated job-posting data for skill demand, and — as the product grows — our own anonymized, consented user data. As each lands, this page will say which figures are sourced and which are still estimates.
Not advice
CareerWise is an orientation tool, not career, financial, legal or employment advice. Verify anything that matters — especially a salary figure you plan to negotiate with — against sources for your own city, industry and employer before acting on it.