I built 30 financial decision engines in a week. Google indexed one.

#webdev #nextjs #showdev #india

Last month, I spent two hours trying to decide whether to prepay ₹5L on my home
loan or leave it in a Nifty 50 index fund. I have a math background. I write
software for a living. It should not take two hours.

Every Indian personal-finance calculator I opened was either asking for my
email before showing a result, quietly funnelling me toward a loan partner, or
running on US tax assumptions. I ended up asking Claude. Claude gave me a
plausible answer. I asked it again the next day to check. It gave me a
different plausible answer. Not wildly different — but different enough that
I didn't trust either.

So I opened a new Next.js repo that night.

arthaengine.com is what fell out of that week.
30+ financial tools — EMI, SIP, rent vs buy, the full Indian tax regime
comparison, a job offer comparator, retirement corpus, HRA, gratuity, EPF,
home-ownership cost, a debt-payoff planner. No signup. No ads. No lead
capture. Every number runs in your browser and dies there.

This post is about three things I didn't expect to learn while building it:
Why Indian fintech calculators are structurally broken, why "calculator" is
the wrong word, and why Google has indexed exactly one of my forty-nine pages.

The incumbents aren't trying to help you

Open any Indian finance calculator from a bank, an AMC, or a tax-filing
platform. The math is usually right. That isn't where they fail you.

They fail at the defaults and the next step.

A home-loan EMI calculator from a bank will happily show you your EMI for a
₹50L loan at 8.5% over 20 years. It will not compare that against a 15-year
tenure. It will not run a prepayment scenario. It will not tell you that at
your income, renting for five more years and investing the down payment
likely beats buying.

Because the bank's calculator is a lead-capture form with arithmetic attached.
Of course, it isn't going to tell you not to take a home loan.

The same is true of every "free" calculator from an AMC (routes you to their website)
SIP), an insurance aggregator (sells you term), or a tax-filing platform
(upsells Premium). The math serves the funnel.

This isn't a conspiracy; it's incentives. But once you see it, you can't
unsee it, and you realise the entire category is structurally broken for the
thing a user actually wants: an honest, comparative answer.

Why I stopped calling them calculators

A calculator takes inputs and returns a number.

Loan: ₹50,00,000
Rate: 8.5%
Tenure: 20 years
——————————————
EMI: ₹43,391

Useful. Now what?

The real question was never "what's my EMI." It was "can I afford this house,
and if I can, should I buy it, and if I should, is this the right tenure,
and is prepaying better than investing the difference?" Five decisions
nested inside a number.

A decision engine takes inputs and helps you choose between paths. Every
page on the site defaults to two-sided comparison:

Not "your EMI" — buying vs renting + investing the difference over 10 years, with the breakeven year highlighted
Not "your take-home" — Offer A and Offer B side-by-side, both under the new and old tax regimes, NPV'd over 5 years
Not "your interest saved by prepaying" — prepayment at your loan's effective rate vs the same EMI invested at historical equity CAGR, broken out year by year

This reframing changed the entire UI. Every page has two columns, not one.
The primary CTA is "compare", not "calculate." There's no "submit" button —
Inputs update results live, because the whole point is to play with
scenarios, not to request an answer.

Small design shift, large product implication: the tool stops pretending it
has an answer. It shows you the math and gets out of the way.

The technical problem nobody talks about: Indian tax

I'll spare you the Next.js / Tailwind / Vercel paragraph. You've read a
hundred of them. The interesting engineering problem wasn't the stack.

The interesting problem was the Indian income tax.

Most calculators online handle Indian tax as if it were three if-statements:
Apply slab, subtract deductions, done. That is not how the system works, and
the difference is ₹20,000-80,000 a year for a typical salaried professional.

Here's what a correct calculation actually has to handle:

Two regimes. The new regime (default from 2023) has different slabs, no HRA, no 80C, but a higher standard deduction and a higher rebate threshold.
87A rebate. Full tax rebate up to ₹7L taxable income under new regime, ₹5L under old. This creates a cliff — at ₹7,00,001, tax doesn't jump from zero to a few hundred, it jumps to ₹25,000+. Miss this , and you're wrong for anyone within ₹25K of the threshold.
Marginal relief — a specific provision that exists because of the 87A cliff, softening the jump. Most calculators don't implement it.
Surcharge tiers at ₹50L / ₹1Cr / ₹2Cr / ₹5Cr. Each adds a percentage on top of tax. The tiers differ between regimes.
Surcharge marginal relief — the surcharge jumps also have marginal relief. I've yet to find a competitor that handles this right.
4% health & education cess on top of everything.
HRA exemption — least of three: HRA received, 50%/40% of basic (metro/non-metro), rent paid minus 10% of the basic.
Standard deduction — ₹50K (old) or ₹75K (new, post-Budget 2024).
80C ceiling at ₹1.5L, but 80CCD(1B) NPS (₹50K) is separate and 80CCD(2) employer NPS is uncapped.

Skip any of these, and the calculator is quietly wrong. Most competitors I
tested got 4-5 right. A few got 7. None of the ones I found handled surcharge marginal
relief correctly.

Writing this was the single largest chunk of the week. I ended up building
a pure-function tax module with 40+ test cases lifted from actual Income
Tax Department examples, because the only way to trust the output is to
Prove it against the authoritative source.

Not glamorous work. But it's the work that makes the difference between a
calculator that looks right and one that is right.

Google has indexed one page out of forty-nine

Here's the part that's been eating me this week.

I shipped the site with what I thought was a clean SEO foundation — proper
sitemap, per-page canonicals, Open Graph tags, JSON-LD SoftwareApplication
schema on every calculator, strict robots.txt allowing everything except
auth routes. Google Search Console accepted the sitemap on day one and
discovered all 49 URLs.

A week later, Search Console shows:

Indexed: 1 (the homepage)
Not indexed: 48 — reason: Discovered – currently not indexed

I spent an afternoon convinced I'd broken something. Canonicals wrong? No.
Pages blocked? No. JavaScript-rendered content Google can't see? No — I
checked with URL Inspection, full HTML renders server-side via React Server
Components.

The issue isn't technical. Google doesn't crawl new domains aggressively.
You can do everything right and still sit in the discovery queue because
the domain has no authority signals yet — no backlinks, no engagement, no
reason for Google to spend crawl budget on your sitemap.

This is the most important lesson of the week, and it has nothing to do
with code: technical SEO is necessary but not sufficient. What unblocks
indexing is external signals — other sites linking to yours, real users
visiting, real time-on-page — not a cleaner canonical tag.

Which is partly why I'm writing this. If you read this far and the site is
worth five minutes of your time, that's what breaks the logjam. Not an ego
thing — a literal "Google has to see this domain exists in the wider web"
thing.

What I won't pretend

A few honest things I won't dress up:

The UI is rough in places. Large-number inputs on mobile are painful. I know. It's on the list.
There's no business model. I don't want one. The moment I add an affiliate link, I become the thing that annoyed me into building this.
Some of the "advanced" engines simplify. The FIRE planner and retirement corpus calculator assume constant real returns — fine for directional thinking, not sufficient for financial planning. There's a methodology note on each page. Read it before making a real decision.

Three to try

If you've got a few minutes:

Job Offer Comparator — Paste two offers, get a full CTC-to-in-hand breakdown under both tax regimes, side-by-side
Rent vs Buy — 10-year simulation with breakeven year
Salary Tax — both regimes, with every deduction plus the surcharge/cess/rebate/ marginal-relief logic that most others skip

If you hit a bug or disagree with the math, tell me. I'd rather fix a
Wrong answer, then ship a new feature.

Rajkumar Anguluri. Software developer. Built this solo over a week.
Not monetising, not taking leads, not building a funnel.