Quantify a vague resume bullet without fabricating numbers

You are a senior engineer who reviews resumes for Indian tech freshers. I will
paste exactly one weak resume bullet point. Help me quantify it honestly.

Step 1 — Ask me 3 clarifying questions, in this exact order:
  Q1. SCOPE: How big was the thing you built? (Number of features, screens,
      endpoints, files, models, datasets — whatever applies. If you do not
      know an exact number, give a rough range.)
  Q2. USERS or USAGE: Who actually used it and how many of them? (Yourself
      only, a college club, a hackathon team, a small group of testers, a
      production deployment — be specific. If nobody used it yet, say so.)
  Q3. TIMELINE or FREQUENCY: How long did it take to build, or how often
      does it run? (One weekend, one semester, runs hourly, runs once a
      day — pick what applies.)

Stop after asking the 3 questions. Wait for me to answer all 3 before
writing any bullets.

If you are running in a single-turn tool that cannot wait for replies, assume
"I don't know" for all 3 questions and produce 3 honest scope-word versions
(use words like "small team", "first version", "demo build", "weekly run")
instead of numbers. Never invent counts, percentages, or revenue figures.

Step 2 — Once I have answered, write exactly 3 quantified versions of the
bullet. Each version should pick a DIFFERENT honest angle:
  Version A: Lead with scope number (size of what you built).
  Version B: Lead with usage or scale (who used it, how often).
  Version C: Lead with timeline or velocity (how fast you built or
             how frequently it runs).

Hard rules — do not break these even if I beg you:
- Only use numbers I gave you in my answers. Do not invent users, percentages,
  performance improvements, latency reductions, revenue, time saved, or any
  other metric.
- If I said "I don't know" or "nobody used it," do NOT add a number to that
  dimension. Use honest scope words instead: "small team", "weekly", "first
  version", "internal use", "prototype". These are legitimate.
- Never write "improved performance by X%" or "reduced latency by X ms" unless
  I explicitly measured it and told you the number.
- No buzzwords ("seamless", "robust", "scalable", "production-grade") unless
  the underlying claim is backed by something I said.
- Each bullet starts with a strong action verb (Built, Shipped, Implemented,
  Engineered, Designed, Deployed).
- Plain text only. No markdown, no emojis.

After the 3 versions, add one short note (1-2 sentences) telling me which
version is strongest given what I described, and why.

Here is my weak bullet:
[PASTE ONE BULLET HERE]

Who this is for

This prompt is for B.Tech and B.Sc freshers in India whose resume bullets technically describe what they built but never tell a recruiter how much. If your bullets read like "Built a web app using React and Node.js" or "Worked on a machine learning model for image classification," you have a quantification problem. Recruiters scanning resumes from a stack of 200 will spend more time on the bullet that says "Built a job-tracking dashboard with 12 filterable views, used by 40 students in my college placement cell" than on the generic version.

It is also for students who tried to quantify their bullets the wrong way — by making up numbers. The classic mistake is writing "improved performance by 30%" when you never measured the original performance, or "served 1000+ users" when only your project partner and one TA ever logged in. Interviewers at Razorpay, Flipkart, Swiggy, or any product company love to ask "how did you measure that 30%?" because they know most freshers cannot answer.

This prompt enforces a different discipline. It asks you three questions about what is actually true, then writes bullets using only those answers. If you never measured latency, the bullets will not mention latency. If only five people used your app, the bullets will say "used by a small group" or "used by 5 testers" — both are honest, and both are stronger than no number at all.

This prompt is rated Intermediate because it requires you to think carefully about your project. The other prompts on this site do more work for you. This one makes you do the work and rewards the honest answer.

How to use it

Step 1: Pick one weak bullet. Open your current resume. Find a bullet that describes a real thing you built but has no numbers. Copy just that one bullet. Do not paste your whole resume — this prompt works on one bullet at a time.

Step 2: Paste and answer honestly. Drop the prompt into ChatGPT, Claude, or Gemini, replace [PASTE ONE BULLET HERE] with your bullet, and send. The AI will ask three questions: scope, users, and timeline. Answer each one truthfully. If you do not know, say "I don't know" or "I never measured this." The AI is designed to handle honest non-answers.

Step 3: Read the three versions. You will get three quantified bullets, each leading with a different number. Pick the one that best represents your actual work, or combine the strongest phrase from each into your own version.

Step 4: Repeat for every weak bullet. This prompt is meant to be run multiple times, once per bullet. Going through your top 6 to 8 project bullets one by one takes about 20 to 30 minutes and is the single most impactful resume edit most freshers can make in one sitting.

Example before and after

Before (the weak bullet, as a student wrote it): "Built a movie recommendation system using Python and machine learning for college project."

The AI's 3 clarifying questions: Q1 SCOPE: How big was the recommendation system? (Number of movies, users, features, models?) Q2 USERS: Who actually used it? (Just you, classmates, anyone else?) Q3 TIMELINE: How long did it take, or how often does it run?

The student's honest answers: Q1: Around 5000 movies in the dataset, used 2 algorithms (collaborative filtering and content-based), 1 hybrid model on top. Q2: Demoed to my class and 2 professors during the project showcase. No deployed users. Q3: Built it across one semester, around 3 months of part-time work.

After (the 3 versions the prompt produces):

Version A (lead with scope): Built a hybrid movie recommendation system combining collaborative filtering and content-based algorithms across a 5,000-movie dataset.

Version B (lead with usage): Built a movie recommendation system demoed to a class of students and 2 faculty reviewers during the end-of-semester project showcase.

Version C (lead with timeline): Built a hybrid movie recommendation system over one semester, combining collaborative filtering and content-based models on a 5,000-movie dataset.

Note: Version A is strongest here — the 5,000-movie dataset and the hybrid approach are the most concrete signals for an ML-leaning recruiter.

Notice what is NOT in any of these bullets. There is no "improved recommendation accuracy by 22%" because the student never measured it. There is no "served 1000+ users" because nobody used it outside the demo. There is no "production-grade" or "scalable" because it was a semester project. Every claim is true, and every claim is verifiable in an interview.

Common mistakes to avoid

Answering the questions vaguely. If you answer Q1 with "I don't really know how big it was," the bullets will be weaker than they could be. Take 2 minutes to actually count. Open your project's GitHub repo, look at the number of files, components, endpoints, or whatever applies. The clarifying questions are the most important part of the prompt.

Letting the AI write a "performance" number you never measured. Sometimes the model slips and adds "improved accuracy by 15%" even though you never said that. Read every word. If you see a number you did not give, delete it. The prompt is explicit, but you are the final filter.

Using the same prompt run for a different bullet. Each bullet has different scope, users, and timeline. Running it once and trying to reuse the same answers for another project will give you bullets that do not match the second project. Run the prompt fresh for each bullet.

Picking the most impressive-sounding version when a different version is more honest. If Version A leads with "5,000-movie dataset" but only 200 movies actually had complete data, Version A is misleading. Pick the version where every word is defensible in an interview.

Skipping the prompt entirely and just adding fake numbers. This is the failure mode this prompt exists to prevent. Recruiters interview thousands of freshers a year. They know what a fabricated metric sounds like ("processed 1M+ records," "99.9% accuracy," "10x faster"). Honest, smaller numbers always beat impressive-sounding fake ones.

When not to use this prompt

Do not use this prompt if your bullet is wrong at the structural level — for example, if it does not name a technology, does not describe what you specifically built, or is written in passive voice. In those cases, use the bullet point rewriter prompt first to fix the structure, then come back here to quantify.

Also skip this if you genuinely have nothing to quantify. Some experiences — like attending a 2-day workshop or reading a book on system design — should not be on your resume at all. Quantification cannot save a bullet that should not exist. If your bullet is "Attended GeeksforGeeks DSA bootcamp," delete it; do not quantify it.

Finally, do not use this prompt for soft-skill or leadership bullets like "Coordinated team of 4 for college fest." Those bullets need a different framing, often built around outcomes and responsibilities rather than scope/users/timeline. This prompt is built for technical project bullets where size, usage, and velocity are the right dimensions to measure.