Keep the rental or sell it. That is the whole question, and answering it properly takes about fifteen inputs and a discount rate you can defend.
The tool runs entirely in the browser. No account, no backend, no server that could leak what you own or what you owe on it.
Where the AI isn’t
Every calculator like this now has a chatbot bolted to the side that tells you what to do. That makes it worse. If a model produces the verdict, you cannot audit how it got there, and the entire value of the thing is that you can check its work.
So the math is deterministic and every assumption is visible: cap rate, cash-on-cash, current equity, monthly cash flow, a CAPM-derived hurdle rate, NPV and IRR at a ten-year exit.
There is one place a model belongs — pulling a messy listing into form fields. Unstructured text into structured data is the thing models are genuinely good at. It fills the inputs; it does not touch the calculation. Knowing where to stop is most of the design.
The export outlives the app
The workbook downloads with five sheets, and the projections are written as real Excel formulas pointing at the inputs sheet. Change a cell and everything recalculates.
That matters because you can hand it to your accountant or your spouse and argue with the assumptions offline. A tool that hands you a number asks you to trust it. A tool that hands you a working model does not need you to.
Proving the privacy claim
“Your data never leaves your device” is the kind of claim users have to take on faith, and plenty of apps say it while firing analytics on every keystroke.
So there is a panel in the corner that reads the browser’s own PerformanceResourceTiming entries and draws every request the page makes. Open it and watch what happens after load: nothing. No API calls, no telemetry, no beacons. It reads the browser’s accounting rather than the app’s own code, so it cannot miss a request — including one the app tried to hide.
Building the instrument turned out to be more interesting than the claim, and it became its own project.
Next
The financial model now has 38 tests with values worked out by hand from the formulas rather than captured from the code — the point being to catch a wrong formula, not just a changed one. After that, Monte Carlo across rent growth, vacancy and appreciation, so the output is a distribution and a probability of clearing the hurdle rate instead of a single confident number.