Funnel Maths and Your Master Target List

Most students decide how many places to apply to by feel. Twenty sounds like a lot. Forty sounds like a lot of work. Then twenty applications produce two assessment links, one interview and a rejection, and it feels like failure, when it was simply arithmetic.

This article does the arithmetic first, so your effort matches the size of the task, and then builds the list of companies that the arithmetic says you need.

• How a job-search funnel works, and why most applications produce nothing.
• How to estimate the number of applications you need, without pretending to know rates you do not know.
• Why improving an early stage is worth more than applying harder.
• How to build and prioritise a master list of 60 to 80 companies, using AI and checking its output.

Every application passes through stages, and at each stage most people stop:

1. Application → screen. A recruiter or a system looks at you, and you get either an assessment link, a recruiter call or silence.
2. Screen → interview. You pass the assessment or call and get invited to interviews.
3. Interview → offer. You get through the interview rounds.

If a quarter of interview loops end in an offer, you need about four interview loops per offer. If two in five screens lead to an interview, you need about ten screens to get those four interviews. If one in eight applications leads to a screen, you need about eighty applications to get those ten screens. Each stage multiplies the one before it.

You will see confident figures online: "one offer per hundred applications", "referrals get five times the callbacks". Treat them with suspicion. Rates depend on your profile, your targets, the role, the month, the channel and the company, and almost nobody publishes reliable data for freshers in India. This course does not give you "the real rates". It gives you a way to plan with assumptions, then measure your own rates within a few weeks and plan with those instead.

This short program works backwards from the number of offers you want, using rates you choose. The rates in it are assumptions for illustration.

# File: funnel.py# How many applications does a target number of offers need?# The conversion rates below are ASSUMPTIONS for illustration, not statistics.# Replace them with rates measured from your own tracker.import mathSTAGES = [    ("application -> assessment or screen", 0.12),    ("assessment or screen -> interview loop", 0.40),    ("interview loop -> offer", 0.25),]def applications_needed(offers_wanted, stages):    needed = offers_wanted    for _, rate in reversed(stages):        needed = needed / rate    return math.ceil(needed)def expected_counts(applications, stages):    counts = [applications]    for _, rate in stages:        counts.append(counts[-1] * rate)    return countsdef main():    apps = applications_needed(1, STAGES)    print(f"Applications needed for 1 offer: {apps}")    labels = ["Applications", "Assessments or screens", "Interview loops", "Offers"]    for label, count in zip(labels, expected_counts(apps, STAGES)):        print(f"  {label:<24}{count:>6.1f}")    print("\nWhat if only the first rate changes?")    for first_rate in (0.06, 0.12, 0.20):        changed = [("first", first_rate)] + STAGES[1:]        print(f"  screen rate {first_rate:.0%}: {applications_needed(1, changed)} applications")main()

Applications needed for 1 offer: 84  Applications              84.0  Assessments or screens    10.1  Interview loops            4.0  Offers                     1.0What if only the first rate changes?  screen rate 6%: 167 applications  screen rate 12%: 84 applications  screen rate 20%: 50 applications

Two lessons are in that output.

Lesson 1: volume matters. With these assumptions, 20 applications would be expected to produce about two screens and less than one interview loop. That is not a bad search; it is a search that has not yet had enough tries. Students who stop at 20 often conclude something is wrong with them when the sample was simply too small.

Lesson 2: the earliest weak rate matters most. Halving the screen rate from 12 to 6 per cent doubles the applications needed, from 84 to 167. Raising it to 20 per cent cuts them to 50. Applying "harder" with the same resume keeps the rate where it is. Fixing the resume, adding a referral or targeting better moves the rate itself. That is why Modules 2 and 3 come before the application sprint, and why article 42 teaches you to find your weakest stage.

After about 30 applications, calculate from your tracker:

• Screen rate = assessments or recruiter calls ÷ applications.
• Interview rate = interview invitations ÷ screens.
• Offer rate = offers ÷ interview loops completed.

Small numbers are noisy. Two screens from 30 applications might be bad luck or a real problem. Treat a rate as a signal once it rests on at least 20 or 30 tries at that stage, and look for big differences, for example between channels (referral versus cold application) or between company groups.

If your plan needs around 80 applications in the next six weeks, that is about 13 to 14 a week, or two a day. That is manageable, and it also tells you something: you cannot spend an hour personalising every application. Article 22 shows how to give the first 20 or 30 applications full care and then speed up without losing quality.

The arithmetic says you need many companies, so you need a list long enough to apply from for weeks: aim for 60 to 80, split by your reach, target and safe groups from article 3.

No single source covers the market, and each one shows you different companies.

AI research tools can suggest companies quickly, but they can also invent companies, merge two companies into one, give careers page addresses that do not exist, or describe a company as it was years ago. Ask for suggestions with reasons, and check each one.

A useful request (in a search-connected assistant that shows sources):

I am a fresher in India looking for [backend / data analyst / QA] roles,open to [cities or remote]. Suggest 25 companies in these groups:funded startups in [domains], global capability centres in [city],and Indian product companies in [domains].For each: name, what they build in one line, the group, and the sourceyou used. Only include companies where your source is from the last 12 months.If you are unsure a company exists or is hiring, leave it out.Do not give careers page URLs; I will find those myself.

Asking it not to give careers URLs is deliberate. You will look those up yourself on the company's real website, which is also the moment you check that the company exists and has roles you could apply to.

How to check each suggestion: open the company's own website, find its careers page, and confirm at least one of: an open role matching your profile, a history of hiring freshers (job titles such as "associate", "graduate", "0 to 1 years"), or a team doing work you want. Remove anything you cannot verify in two minutes.

Once you have 80 to 120 verified names, prioritise them. You can do this in a spreadsheet with three columns of your own judgement, or ask an assistant to propose a first ranking from the facts you collected:

Below is a table of companies I have verified, with what they build, their group,the stack in their job posts, and whether they hire freshers.My profile: [stack you can defend in an interview], [target group from myreach/target/safe split], [location preference], graduating [month year].For each company, propose Priority (High / Medium / Low) using only:stack overlap with my profile, fit with my target group, and evidence offresher hiring in the table. Give a one-line reason per company.Do not add facts that are not in the table.[paste table]

Then read the ranking and change anything you disagree with. You know things the table does not, such as which domains excite you, which seniors you can ask and where you would actually move to.

Companies open and close roles constantly. Two habits keep the list useful:

• Weekly: add five new companies from any source, and move dead ones (no fresher roles for months) to an archive tab.
• Alerts: set job alerts and news alerts for your top 15 companies (article 16), so new roles and news reach you without checking by hand.

Divya wants junior data analyst roles in Pune or remotely. Her split is 20 per cent reach (Indian product companies), 60 per cent target (capability centres and startups), 20 per cent safe (services companies' analytics tracks).

• She collects 40 names from job boards, 25 from a list of capability centres in Pune, 15 from her alumni page and 20 from an AI suggestion request.
• Verification removes 18: three companies from the AI list that she cannot find at all, two that were acquired, and thirteen with no analyst roles in the last year.
• She prioritises the remaining 82. Her top 15 get alerts; the rest are spread across six weeks of applications.

Three companies "disappearing" during verification is normal. That is why the verification step exists.

• Planning with someone else's numbers. Use assumptions only until your own tracker gives you rates.
• Stopping at 20 or 30 applications and concluding the market or you are the problem. That is usually too few tries to tell.
• Trusting an AI-generated list without checking it. Invented companies and dead links waste hours later.
• A list that is all reach companies. Keep the split.
• A list that never changes. Add and archive weekly.

• Next: Your Pipeline Tracker and 10-Week Execution Plan turns the list and the maths into a weekly routine.
• For more on finding and judging job descriptions with AI, see Job Descriptions, Skill Gaps and Portfolio Projects with AI.
• To check AI-researched facts properly, see How to Fact-Check AI-Generated Answers.