How to Pick a Career With Real Data

A four-dimension framework using salary, job growth, automation risk, and education ROI — not vibes.

The problem with most career-quiz recommendations

Most "what should I do with my life" quizzes match you to a personality archetype and then hand you a list of generic job titles. They almost never ground that list in data: what the job pays in your state, whether the field is growing or shrinking, how vulnerable the role is to automation, or whether the education path will earn back its cost. According to the Bureau of Labor Statistics' Occupational Outlook Handbook, the gap between the highest-paying occupation in the United States (anesthesiologists, median above $300,000) and the lowest (fast-food cooks, around $28,000) is more than ten-fold. The career you choose changes your lifetime earnings by hundreds of thousands of dollars. It deserves more than a vibe check.

This guide walks through a data-first framework you can apply to any career. Every claim cites the primary source — Bureau of Labor Statistics, O*NET, or the U.S. Department of Education's College Scorecard — so you can verify the numbers yourself.

Dimension 1 — Median pay

Start with the median annual wage from the BLS Occupational Employment and Wage Statistics (OEWS) report, updated annually. The median is more useful than the mean because it isn't skewed by a small number of high earners. BLS also publishes 10th, 25th, 75th, and 90th percentiles, so you can see the realistic range.

Source: U.S. Bureau of Labor Statistics, OEWS, May 2024.

An elevator installer with a high-school diploma earns more than the median elementary teacher with a bachelor's degree. Pay does not always track years of school. The framework forces you to look at the actual number rather than the prestige of the credential. State variance is also significant: registered nurses in California earn a median of $137,690 versus $74,870 in Alabama. If you're locked into a particular state, use that state's number.

Dimension 2 — Projected job growth (10 years)

The BLS Employment Projections program publishes 10-year growth forecasts for every occupation in the Standard Occupational Classification (SOC) system. The most recent release projects 2024–2034. "Average" growth is around 4%; anything over 10% is "much faster than average."

Standout high-growth careers from BLS Employment Projections 2024–2034:

• Nurse practitioners — projected 45% growth.
• Wind turbine service technicians — projected 45% growth.
• Information security analysts — projected 32% growth.
• Data scientists — projected 35% growth.
• Physician assistants — projected 28% growth.
• Solar photovoltaic installers — projected 22% growth.

Slow- or negative-growth roles to be cautious about: data entry keyers, word processors and typists, telephone operators, and parking attendants. A career with 3% projected growth still hires — it just means competition for entry-level spots is tougher.

Dimension 3 — Automation risk

Different occupations face different exposure to automation by 2035. There isn't a single official "automation risk" number from the U.S. government, but O*NET publishes task profiles for every occupation, and researchers (Frey & Osborne 2013; Felten et al. 2023) have used those profiles to estimate the fraction of tasks that are technically automatable.

Rules of thumb:

• Lower risk — in-person care, fine motor skills, novel problem-solving, unstructured environments. Nurse practitioners, electricians, mental health counselors, dentists, plumbers.
• Higher risk — highly repetitive work, scripted interactions, purely digital tasks with structured I/O. Data entry, basic bookkeeping, telemarketing.
• Mixed risk — software development and analytical knowledge work. BLS still projects 25% growth for software developers through 2034 despite increased AI tooling, because automation augments rather than replaces.

TruePath shows an Automation Risk % for each occupation derived from the O*NET task profile.

Dimension 4 — Education ROI

The U.S. Department of Education's College Scorecard publishes median earnings 4 years after completion for every program at every institution that participates in federal student aid. It also publishes median student debt at graduation for those same programs. The combination lets you compute a real ROI per program.

A useful "break-even year" formula:

break-even year = (total tuition + opportunity cost of not working) ÷ (post-graduation salary − salary without the degree)

For a $40,000 Bachelor of Computer Science at a public university leading to a $90,000 first-year salary (vs. $40,000 without the degree), break-even is roughly 3.9 years post-graduation. For a $250,000 private dental school degree leading to a $170,000 first-year salary, break-even can stretch to 8–12 years depending on debt and interest. The point isn't shortest break-even — it's knowing the number before you commit.

TruePath's Reality Check feature runs this calculation for any career, pulling tuition from College Scorecard and salary from BLS, and lets you adjust loan interest rate, current salary, and work-during-school status.

Putting the framework into practice

1. List 5–8 careers that pass an interest filter — fields you'd happily read about for 20 minutes.
2. Look up median pay in your state via BLS OEWS or CareerOneStop.
3. Check growth projection via BLS Employment Projections.
4. Estimate automation exposure using O*NET tasks (or TruePath's pre-computed estimate).
5. Run education ROI using College Scorecard tuition for the cheapest in-state program.
6. Rank by break-even year and pick the top 2 that you'd enjoy doing day-to-day.

Done manually, this takes about three hours per career. TruePath automates the lookup and computation for all 968 occupations in the BLS SOC system.

A worked example: Texas, 22 years old, high school diploma

Pay: BLS OEWS Texas, May 2024. Growth: BLS Employment Projections 2024–2034. Tuition: College Scorecard medians, in-state public.

Common mistakes

1. Comparing average pay instead of median

Means are skewed by high earners. Use median for realistic outcomes.

2. Ignoring state and metro variance

RN pay in the Bay Area is roughly double rural Alabama's. Cost of living differs but not as much as the pay.

3. Confusing growth rate with absolute openings

45% on a 5,000-person occupation adds 2,250 jobs over 10 years. 4% on a 3-million-person occupation adds 120,000. Both matter.

4. Ignoring debt

$50K of debt with $55K starting salary ≠ $5K of debt with same salary. Always include the loan payment in break-even.

How TruePath helps with this

The framework above is exactly how TruePath structures its career database. Every career profile in TruePath shows median salary (national + your state, sourced from BLS OEWS), 10-year growth projection (BLS Employment Projections), automation risk percentage (O*NET-derived), required education, training time, and key skills. The Reality Check feature runs the break-even calculation automatically using College Scorecard tuition figures, your current salary, and a loan interest rate you can adjust between 0–30%. The AI Roadmap feature takes a chosen career and generates a 100+ step personalized plan including specific course codes, institutions, and monthly cost estimates for your state.

Free users can browse all 968 career profiles, take the 15-question assessment, save favorites, and run two comparisons. The Full Access tier is $75 one-time and includes three AI roadmaps, three Reality Checks, 100 AI questions, and unlimited comparisons — no subscription.

Frequently asked questions

What is the single best way to pick a career?

No single dimension captures career fit. Combine four data points — median pay in your state (BLS OEWS), 10-year growth (BLS Employment Projections), automation risk (O*NET task profile), and education ROI (College Scorecard tuition divided into post-graduation salary differential). Rank options by break-even year and pick the top option you would actually enjoy doing day-to-day.

How accurate are BLS salary medians?

BLS OEWS is the most rigorous occupational wage dataset in the U.S. It surveys roughly 1.1 million establishments and 80 million workers over a three-year rolling cycle. The estimates are accurate to within a few percent for any occupation with 1,000+ workers nationally. For occupations with fewer workers, BLS suppresses the figure rather than publish unreliable estimates.

Are interest-based career quizzes useless?

Not useless, but incomplete. The Holland Codes (RIASEC) framework that powers most career quizzes correlates with job satisfaction in roughly 30% of variance across studies. The other 70% comes from pay, work environment, schedule, growth trajectory, and management quality. Use an interest assessment as input 1 of 5, not as the answer.

Should I factor in remote work?

Yes. BLS OEWS does not directly publish remote-work percentages, but O*NET work activity data and recent Bureau of Labor Statistics tele-work surveys allow estimation. TruePath publishes an estimated remote work % for each career; rates range from 0% (electrician, dental hygienist) to 70%+ (software developer, data scientist).

How do I weight job growth vs current pay?

A career with high current pay but flat growth (e.g. petroleum engineer) is a fine choice if you can enter the field now and pay back your debt within 5–7 years. A career with lower current pay but high growth (e.g. wind turbine technician) is a fine choice if you are willing to ride the wave for a decade. Avoid careers with low pay AND low projected growth.

Does TruePath show data for my specific state?

Yes. Every career profile shows the BLS OEWS state-level median wage for your state (selected at signup). Comparisons are personalized to your state — for example, a Software Engineer vs Registered Nurse comparison in Texas shows $129,260 vs $82,210 instead of national medians.

How often is the data updated?

BLS OEWS releases new wage estimates annually (typically late March / early April for the prior year). BLS Employment Projections release every two years. TruePath ingests the latest published BLS dataset within 60 days of each release.