Four calculators, a spreadsheet, and no idea if the numbers agreed. So I built one that does.
A free tool that models a Canadian's whole financial picture as one connected system: a budget that feeds the investment plan, TFSA, RRSP and FHSA contributions checked against the room you actually have, retirement drawdown, and net worth measured against published Canadian data.
A free planning tool for Canadians that replaces the four browser tabs and a spreadsheet many of us use to answer one question: am I going to be alright?
- Role
- Sole designer, director and product owner
- Scope
- Concept, UX, UI, information design, QA
- Built with
- Claude wrote the code. I made every other call.
- Status
- Live since July 2026. Updated October 2026, three months in.
- Product design
- UX research
- Information design
- Front-end
- 52 — Public comments from six online communities
- 2 — Education listings, including the Alberta Teachers' Association library
- ~80 — Changes shipped in the first three months
It started as my own spreadsheet
Standard financial tools give you isolated numbers, not clear answers.
A raise, a tax strategy & retirement date are all deeply connected, yet every online tool evaluates them in a vacuum. You end up re-entering the exact same data across half a dozen tabs, only to stitch the math together yourself in a spreadsheet prone to silent errors.
Why stitching doesn't work
Every one of these is survivable alone. Together they make the picture untrustworthy.
The inputs never match
Each wants your salary, province and age again, its own way, with its own assumptions baked in.
Nothing carries forward
The take-home one tool computes isn't what the next starts from, so errors compound quietly down the chain.
The interesting questions span tools
TFSA or RRSP depends on your marginal rate, remaining room and drawdown plan at once. No single-purpose tool sees all three.
No sense of scale
A number alone doesn't tell you if you're doing well. Few tools show you what is actually typical in Canada.
Four tools, the same inputs typed four times, and a reconciliation step that exists only in your head.
One engine, flexible entry points.
Financial tools tend to force a bad trade-off: they either wall you off behind a 20-minute profile before showing any value, or they trap you in a vacuum with a fast answer that ignores the rest of your financial life.
Two steps, answer first.
Step one asks for salary and province. Step two already shows your take-home, then asks where to go.
Ask the one question on your mind; the answer comes from the same model of your whole plan.
The budget is the engine, not a side panel.
Budgeting and investing usually live in separate tools. They're really one system.
Your budget is the live cash flow behind every long-term projection. Trim $50 a month and the savings it frees up is one tap away from your investment plan, so the retirement date you see is built on what you actually spend.
The real 2026 limits. Contribute past them and the overflow lands in a taxable account.
Closing the tax refund loop.
Contributing to an RRSP or FHSA earns a tax refund, and a refund calculator tells you what comes back, then stops. What happens after it lands in your account is where the money is.
So the model treats the refund as a decision: spend that cheque, or reinvest it? Side by side, the two compounding curves make the gap hard to ignore.
The edge a single-purpose calculator cannot model: it needs your rate and your plan together.
Answers first, evidence second.
The plain-English answer leads, and the charts and sliders sit underneath as evidence, not a puzzle to decode.
Clarity is accessibility.
Financial tools rarely fail at the math. They fail right after it, when someone has a correct number and no idea what to do next.
So plain language is an accessibility requirement here. Every screen answers in a sentence before it shows a chart, and asks for one input at a time. Underneath, every field is labelled, results are announced as they change, focus is always visible, text meets AA contrast in all three themes, and the layout holds at 320 px wide.
A 100 PageSpeed accessibility score is the floor, not the finish.
The 500 KB charting library now loads after first paint. Mobile PageSpeed went 78 to 100.
Rigor behind the UI.
A polished interface means nothing if the math underneath is wrong. Flat tax rates fall apart in retirement, where brackets, credits and clawbacks shape every withdrawal.
So I built a real tax engine and benchmarked it against an independent calculator. The first retirement run caught a bug in my own engine: it charged the OAS clawback, then taxed those same dollars again, overstating a well-pensioned retiree's bill by about $1,600 a year. It now matches to the dollar.
Every number is checkable. All 43 constants the engine uses are published with their official source and a confidence rating, on a methodology page generated from the engine's own data. Since October, nothing ships without passing the engine's tests, and they fail if the engine, that page and the calculator pages disagree.
Replacing shortcuts with real Canadian tax logic.
A flat-rate shortcut breaks in retirement, so the engine models six layers, including the clawback it once taxed twice.
- 74 / 76 — Sampled tax points within $2 of the reference in July, up from 30
- 12 — Jurisdictions benchmarked point by point in July
- 43 — Tax and benefit constants published with a source and confidence rating
Quebec is not Ontario with a different pension name.
Quebec pay runs on QPP instead of CPP, adds QPIP, takes a reduced EI rate, gets a 16.5% federal abatement and has its own worker deduction. The same $85,000 salary keeps about $3,600 less than in Ontario. Treat that as a rounding difference and a Quebec paycheque comes out wrong.
I modelled each piece separately, for all 13 provinces and territories, then designed a Quebec French edition, now in beta, around one rule: the English string is the key. A missing translation falls back to correct English instead of a blank, and English visitors never download the French dictionary.
Grounded in real context.
A future projection tells you where you are heading, but it cannot tell you if that target is realistic or competitive.
I integrated age-bracketed benchmark data from Statistics Canada and CRA taxfiler records directly into the UI. Users can immediately compare their income, net worth, and registered savings against their actual peers.
To maintain strict data integrity, the interface explicitly distinguishes verified government statistics from extrapolated model estimates. Built-in validation rules ensure mathematical logic holds true, so registered savings can never exceed investable assets or overall net worth.
A percentile alone is trivia; the card adds where you sit, your rank and the gap to the next tier.
Benchmarks only help if people trust them, so every figure is tagged as published or modelled.
Zero lines written. Full design ownership.
Claude wrote every line of code. I acted as product lead and creative director, guiding the model with clear design architecture and system logic rather than basic task tickets.
AI excels at generating plausible output. It cannot tell you what shouldn't exist, nor can it spot when correct math fails human comprehension. That taste, curation, and restraint remain the core design job.
What each tool was actually for
The interesting part is the division of labour, not the stack.
Google Sheets
The mathematical foundation and benchmark used to validate every calculation engine feature.
Miro
Information architecture and user flows, mapped out before a single pixel was drawn.
Figma
Design system scales, grids, and UI tokens translated into strict layout rules rather than static mockups.
Claude
Production code execution. Fast at building, but indifferent to whether something is worth building.
Me
Product direction, taste, and copy editing. Deciding what gets cut, what stays, and ensuring the final result is accessible, concise, and defensible.
Your numbers stay yours. The rest, I say out loud.
Your financial inputs stay in your browser. No accounts, no backend database: your salary, balances and goals are never sent to a server. Share a plan and the numbers travel inside the link itself, not through mine. Local-first is the right default for data this sensitive, and it costs something: no sync between devices and no cloud recovery.
What the app does measure is published in plain language on its own page: anonymous visit and feature counts, with advertising features switched off. The tool is free with no paywall, funded by an optional tip jar and one labelled referral partner that never changes a calculated number. One switch, Community mode, removes both for the whole visit.
Being specific about what is measured earned more trust than a vague promise of none.
Live product, real research loops.
A four-tap feedback widget is my most useful research channel, backed by privacy-limited event counts that show where people stop, never what they typed. The top request from that feedback was household mode, and it shipped: couples are now taxed as two people, not one, which had overstated their bill by thousands a year.
Three months live.
Update, October 2026. The app went live in late July, and about 80 changes have shipped since. Some of the best didn't start with me.
The 52 comments came mostly from Canadian personal-finance forums. The Alberta Teachers' Association library and LEARN Quebec list it, and FCAC's Financial Literacy Month 2026 page names it as a participant, which is a listing, not an endorsement.
It also grew three single-question pages: a take-home pay calculator, a Coast FIRE calculator and an RRSP refund calculator.
Who found what
One report usually means the system is wrong for everyone, so the system is what gets fixed.
A reader's Ontario paycheque looked wrong
July. A low-income credit was missing, so I added it and re-benchmarked every province.
A retiree's plan assumed they still worked
September. Already-retired plans now start from today, and rental income counts.
Published standards set the defaults
Return and inflation follow FP Canada's 2026 guidelines. Default CPP is Service Canada's average new pension: $858 a month, not $925.
My own audit: Quebec was missing a deduction
September. Adding Quebec's worker deduction is why the gap above is now about $3,600, not the $3,900 this page used to say.
My own audit: medals were labelling people
October. Medals now start at the median. Below it you see a plain percentile and what would move you up.
An independent calculator, line by line
October. Against TaxTips.ca's 2026 calculators ($30k to $100k, single filer), Ontario and Quebec agree to the dollar; five more provinces within rounding. The one gap, the Canada Workers Benefit at low incomes, is disclosed, not guessed at.
A curator: classrooms needed a way in
October. Talk Money cards for Financial Literacy Month: with the app's made-up sample plan, a class can talk about pay without anyone sharing their own.
This is very impressive. [...] For reference, I'm a CFP and CPA [...]. Nice work!
A public comment on a Canadian personal-finance forum
What I said no to
AI made building cheap, so the hard part is deciding what shouldn't exist. A few I built, or nearly built, and why they're gone.
An advertising pixel
Removed. Measuring ad conversions wasn't worth an ad network's tracker on a finance page.
A newsletter
Built, then pulled. I couldn't be sure it met Canada's anti-spam law, and even switched off it cost every visitor 7.5 KB.
Referral links that swapped per creator
Deleted the day I built it. Anyone could copy the link and put their own code in.
A paid supporter mode
Declined. The themes were already free, so it meant taking something away to sell it back.
A benefits warning for low-income retirees
Declined. The rules behind it aren't modelled, so it would have been a confident-looking guess.
It's still evolving on purpose. The people using it keep finding what I missed, and they shape what comes next.