How every formula is tested before a file is sold

Every template on this site makes one claim: every formula, verified. This page is what that sentence means in practice — the tests each file has to pass, what those tests have actually caught, and what none of them can do for you.

Nothing here is a review or a rating. There are no reviews yet, and none will be invented. These are counts, and each one is produced by a script that runs against the same file you download.

By template — comparisons, destructive tests, printed texts:

The Analyzer carries most of the comparisons because it carries most of the arithmetic: three scenarios, a thirty-year projection and a 360-payment amortization schedule, each one checked row by row.

The workbook is recalculated from scratch and has to report zero errors. That part is easy to pass and proves very little: a formula can evaluate cleanly and still be wrong.

So every calculation engine is written a second time, outside the spreadsheet, reading the same inputs. Then every result the file produces is compared with the result the second calculation produces. One cent of difference is a failure.

What it caught. While the Deal Screener was being built, a sign inside the mortgage payment formula was reversed. The file opened without a single error, and the monthly cash flow it showed was far too high. Only the second calculation disagreed. On the loaded example the correct figure is $214 a month, and that is what the file shows.

Each file is fed the inputs that real people type by accident, or on purpose: an interest rate of zero, a loan term of zero, 0% and 100% down, a price of zero, no rent at all, 100% vacancy, a short-term rental month with no nights available, and the whole file left empty. The rule is the same for every case: no error cells, and a sentence that still makes sense.

What it caught. With a price of zero, the Deal Screener's break-even rent went blank, and the next formula tried to subtract that blank from a number — an error in three tabs. And when two properties tie on cash flow, a ranked table can quietly show one of them twice and drop the other. Both cases are fixed, and both are now permanent tests.

Each workbook is printed to PDF, and every piece of text in it has to be found in the printed pages, inside the printable area, with the Platwright band at the top of every page.

What it caught. Before launch, the disclaimer disappeared from the printed Start Here tab. The page was set to fit on one sheet, and instead of shrinking the content the print engine cut the bottom off. The file looked perfect on screen.

And what it missed, until September 16, 2026. On long tables the first rows of each page repeat as a title band, and the test only read the rows below that band. That was 281 printed texts nobody was checking. The test now reads them too. None had been lost — and before trusting the new version, a deliberate change was planted in a title band to make sure the test fails when it should. It did.

The automated tests run in LibreOffice. You will open the file in Excel or Google Sheets. So all five files are also opened in desktop Excel and in Google Sheets and checked by hand, to the last tab.

What it caught, and it is the most serious error these files have had. On September 1, 2026, the Landlord Income & Expense Tracker was opened in desktop Excel and its yearly totals did not match. One function behaves differently in Excel than in LibreOffice and Google Sheets: given a range, Excel reads only its first value. In Excel, the summary was counting matching rows instead of adding their amounts. Every automated test had passed, because every automated test ran in the program where the formula worked.

The formulas were rewritten with SUMIFS, checked in both programs, and the corrected file replaced the download on September 3, 2026. The correct totals on the loaded example are $114,975 in and $88,770 out. The Google Sheets copy was never affected.

The rule that came out of it: a function is not allowed because it looks standard. It is allowed after it has been run in every program the file will be opened in. The five files use 17 functions between them, and the build check refuses anything else.

These tests prove that the arithmetic does what it says. They cannot prove that the rent you typed is the rent you will get, that your insurance quote is right, or that a property is a good purchase. A spreadsheet that calculates perfectly will calculate a bad assumption perfectly too.

Nothing on this site approves a property, ranks a deal as good or bad, or predicts a return.