Skills Gap Analyzer — Paste the Job Description, Get a Learning Plan

When candidates read a job description and feel underqualified, the feeling is usually accurate for 2-3 items on the list and exaggerated on the other 12. The tool separates the two piles. It extracts the top 25 keywords from the JD, compares them to your actual skills, and shows you precisely which skills are missing and roughly how long each takes to reach interview-convincing fluency. Once you see the diff visually, the gap goes from "terrifying" to "a 4-week project."

The bigger insight: most job descriptions are negotiable. The "required" list is what a hiring manager would love to have — the "must-have" threshold is usually 60-70% of it. A candidate at 65% coverage with strong adjacent skills and a clear learning trajectory beats a candidate at 85% coverage who stopped growing. The gap number on this tool is a compass, not a cliff.

The tool tokenizes the job description, strips 70+ stopwords (the, and, we're looking for, ideal candidate, etc.), builds unigrams and bigrams, weights bigrams 2x because multi-word phrases carry more signal, and returns the top 25 by frequency. This mirrors the way ATS keyword search works, so the missing-keyword list you get here is close to the missing-keyword list a recruiter's filter would produce.

It's not a semantic matcher. It's a literal string matcher, which is what ATSs run. "Data analysis" and "analytics" are different keywords to the tool because they're different keywords to the ATS. That means if the JD uses "analytics" three times and your resume uses "analysis", you lose the keyword — and fixing it is as simple as swapping one word.

The tool classifies each missing skill into one of three bands:

• Vocab / concept keyword (≈6 hours). A phrase you can understand by reading. Examples: "OKRs", "net revenue retention", "CAC", "agile ceremony names". The time is for reading, note-taking, and being able to discuss it in an interview without sounding like you just Googled it.
• Tool / framework keyword (≈30 hours). Something you need hands-on reps with. Examples: SQL, Python read-level, Amplitude, Mixpanel, Stripe, Tableau, Figma. The time is for tutorials, a small project, and enough troubleshooting to have a real anecdote about it.
• Pattern / domain keyword (≈70 hours). Real expertise. Examples: executive communication, roadmapping, pricing strategy, churn analysis. The time is for deliberate practice plus real applied work, often inside your current job.

These numbers are calibrated to "interview-convincing" fluency, not mastery. Interview-convincing means you can talk about the skill for 20 minutes with a specific example, and a senior person in the room thinks "they could hit the ground running with 2-3 weeks of ramp." Mastery takes years. For a job application, you need interview-convincing.

The worst way to close a 12-item skills gap is to do a little of each. A candidate who spent 100 hours dabbling across 12 skills has zero convincing answers for any of them. A candidate who spent 100 hours deep on two skills has two killer portfolio stories and strong adjacency for the rest.

So the plan the tool outputs sequences skills one at a time. Pick the single biggest missing item. Work it hard for 2-3 weeks. Get one real project out of it. Add the keyword to your resume verbatim, where it's true. Then move to the next item.

This also solves a common problem: when candidates try to close a skills gap before applying, they tend to spend months preparing and then apply to nothing because "I'm not ready yet." Sequencing forces you to ship — you apply to roles after closing 1-2 items, not after closing all 12. Most offers come from candidates who applied at 65% coverage with an upward trajectory, not candidates who waited for 95%.

Once you've learned enough of a skill to have one real artifact, update the resume. The formula: verb + object + specific context.

• After learning SQL: "wrote 40 SQL queries analyzing product usage for the enterprise cohort; identified the activation bottleneck that drove the Q3 roadmap."
• After learning Amplitude: "built 6 Amplitude dashboards used daily by the PM and design team; found the onboarding drop-off that cut signup-to-first-value from 14 days to 5.3."
• After learning pricing: "designed the pricing trade sheet that became the template for 3 at-risk renewals; NRR on the cohort moved from 104% to 128% in 18 months."

Notice the sequence: learn → apply to a real (even if small) project → write the bullet. Without a real artifact, the bullet will not hold up to "tell me about a time you did X" in the interview.

The hours-to-close estimates assume decent-quality self-study. Some channels reliably hit those numbers; others don't.

• What works. Structured courses with projects (Coursera, Udemy paid, specific YouTube series by recognized experts), books with exercises, hands-on sandboxes (Mode Analytics for SQL, Observable for JavaScript, Kaggle for data science), paying for one-on-one tutoring 2-4 hours, pairing on work with a colleague who's stronger.
• What doesn't work. Reading blog posts without exercises, watching videos without coding along, collecting certificates from tutorials you didn't apply, "audit mode" on courses where you never submit assignments.
• Wildcard. ChatGPT/Claude as a tutor works surprisingly well if you treat it like a pair — you write code, you ask "what's wrong with this approach," you argue with the answer. It fails hard if you let it write the code for you.

Every hour you spend reading JDs is an hour you're not spending on applications or prep, so speed matters. A skilled reading pass takes 4 minutes for a standard JD:

1. Minute 1: skim the top for role title, level, team, and comp. Decide if it's in your range. If not, close the tab.
2. Minute 2: skim the "must have" list. Mark any item you genuinely can't cover. If you have more than 3 can't-cover items, move on.
3. Minute 3: skim the "nice to have" list. These are negotiation points, not gates.
4. Minute 4: note the 3-5 keywords you'd want to add to a tailored resume. Those go into the cover letter's hook paragraph.

Applications that come out of a 4-minute JD read and a 10-minute tailoring pass beat applications that come out of a 45-minute re-read and a 2-hour soul search. The soul search doesn't add signal to the recruiter.

• Treating every keyword as equal. "Microsoft Office" as a missing skill is not the same as "product strategy" as a missing skill. The tool's hours-to-close reflects that; do not spend 10 hours on Excel to close a gap when 1 hour would cover you.
• Confusing adjacent skills with missing skills. If the JD asks for "Snowflake" and you have strong Postgres, that's an adjacent skill, not a missing one. On the resume, note both: "Postgres at depth; Snowflake by proxy via similar warehouse architecture." In the interview, you can answer Snowflake questions with Postgres analogies.
• Learning in the abstract. Reading 3 books on pricing is not the same as running one pricing experiment. The tool assumes the latter.
• Not re-measuring. After 4 weeks of learning, re-run the tool with an updated skills list. The gap shrinks, the confidence grows, and it becomes much easier to apply.