Every November, about 1.2 million Sri Lankans need to file an individual income tax return. Most of them have straightforward situations — a salary, maybe a fixed deposit, possibly some rental income. But the IRD portal is dense, the terminology is opaque, and the average person either overpays an accountant or makes mistakes that cost them refunds. We wanted to see if a product could close that gap.
Not in six months. In three days.
Start With the Knowledge, Not the Code
The first thing we built wasn’t a UI. It was a 2,500-word tax computation spec. Every slab rate, every exemption rule, every WHT rate, every edge case — from listed vs. unlisted dividend treatment to the flat 15% on foreign services income — written down before a single pixel was placed.
This isn’t unusual for us. At specshop, the spec is the product. But for LankanTax it was especially critical because tax calculation has zero tolerance for vagueness. If the personal relief is Rs. 1,800,000 and you code 1,500,000, nobody catches that with a visual QA pass. The spec caught it before it existed.
We structured the spec as a seven-tab Excel workbook blueprint: client info, income sources, capital gains, reliefs, tax computation, document checklist, and a locked reference tab. Each cell had its formula. Each formula had its edge case. Three worked test cases — a simple employee, a USD freelancer, a doctor with multiple income streams — verified every formula chain. This spec became the single source of truth for everything that followed.
The spec isn’t documentation. It’s the executable truth that every surface — landing page, wizard, extension — derives from. Change the slab rate in the spec, everything downstream updates.
The principle behind the three-day buildFour Surfaces, One Spec
From that single tax spec, we built four distinct surfaces — each targeting a different moment in the user journey.
Building Trust Through Transparency
Tax software has a trust problem. People are entering their real income — their salary, their side hustle, their rental property — into a form they don’t fully understand, and the output is a number they’re supposed to pay the government. The instinct is suspicion, not confidence.
We solved this with three design decisions:
1. Show the math, always. Every section has an inline calculation explainer that updates as you type. Employment income? You see “Gross salary + Allowances + Benefits = Total employment income” with live numbers. Foreign income? You see the LKR conversion, the 15% flat rate, the foreign tax credit, the net result. No black boxes.
2. Tell them where to find the evidence. Every income section has an expandable “Where to find this” card that names the exact document (payslip, APIT certificate, bank interest certificate, CDS statement) and explains what to look for. This isn’t help text. It’s confidence.
3. Show the tax updating in real-time. A persistent sidebar (desktop) or bottom bar (mobile) shows the running tax estimate — assessable income, reliefs, slab tax, credits, net payable or refund — all updating with every keystroke. By the time users reach the results page, they’ve already seen the number evolve from zero to its final form. There are no surprises.
The real conversion moment isn’t showing the final number. It’s the three minutes before — watching Rs. 0 climb to Rs. 250,800, then seeing APIT credits pull it back down to a refund. That’s when they trust the tool.
Why the live tax meter existsGive Away the Calculation, Charge for the Filing
The business model is deliberate: the tax wizard is free. Every calculation, every slab breakdown, every “where to find this” card — all free. The value is proven before anyone is asked to pay.
The paywall appears after the results page. You’ve seen your tax. You know whether you owe or are owed a refund. You understand how every number was derived. Now, would you like us to auto-fill the IRD portal, generate the evidence documents, and run a pre-submit review? That’s Rs. 4,999. One-time. This tax year.
It’s a conversion path that earns the right to ask. Every free step builds trust equity that the paid step cashes in on.
Testing With CA Firms
The four surfaces are now ready to put in front of Chartered Accountant firms. Not to sell them software — to test three hypotheses:
Hypothesis 1: CA firms will use the free wizard as a client intake tool. Instead of asking clients to bring a folder of papers, they send a LankanTax link, the client fills in what they know, and the CA gets a structured summary with gaps flagged.
Hypothesis 2: The Chrome extension concept resonates as the free tier. CAs recommend it to clients who want to self-file, and those clients upgrade to “Do It For Me” when they get stuck.
Hypothesis 3: The spec itself — the structured knowledge base of Sri Lankan tax rules — has value as an API or white-label engine for accounting software vendors.
Three days. Four surfaces. Three hypotheses. The spec made it possible. Now the market decides what’s worth building next.
The fastest way to validate a product isn’t to build the product. It’s to build the spec, then let the spec build everything else.