Tougather vs Excel: a spreadsheet comparison for real work

Deze pagina in het Nederlands

Where Excel is better

A comparison with no losses in it is an advertisement. Every line below was read out of the code.

Nine functions, refused by name

The library is 167 functions merged from two files. Nine names are deliberately absent, and typing one gets the reason rather than a bare#NAME? — in the column formula editor before you run it, and in the error itself when you type one into a cell. RAND and RANDBETWEEN: a cell that changes on every render cannot be compared with what it said a second ago. UNIQUE, SORT, FILTER and TRANSPOSE: each spills into cells you did not select, and there is no spill model here — the grid has no notion of one formula owning a rectangle. INDIRECT and OFFSET: both build a reference out of text while the sheet runs, blinding the cycle guard to dependencies before it evaluates them.

There is no volatile-function model either, which is a larger absence than a list of nine names looks. NOW is implemented, but read once when the sheet computes rather than ticking. Nothing here recalculates on a clock.

Macros and VBA

There are no macros. The reason is one line in the code: what a macro usually automates — repeat this over every row — is what a formula column already is. That is sound for the common case and covers nothing else. A button that rebuilds a report, an add-in that calls an internal system, a userform somebody in finance wrote years ago: real work Excel does, and nothing here runs it. If the file ends in .xlsm, this is not where it goes.

Fifty thousand rows, and what happens past that

The ceiling was measured with a script, not estimated. Fifty thousand rows is 3.4 MB stored, opens in 0.56 seconds, and puts a keystroke on screen in 118 milliseconds. At a hundred thousand the workspace is 6.8 MB and the browser refuses the write outright: the storage it gives a page is full.

Which thing broke matters. Rendering and scrolling are comfortable at fifty thousand, and the address parser takes three column letters and seven row digits, so addressing was never the limit. Storage is — the price of work that lives in your own browser. A refused write raises a bar across the app saying the work on screen is no longer being saved. Excel does just over a million rows a sheet.

Formulas do not survive the file

This one costs people most and is easiest to miss. The .xlsx import reads cell values, and the date formats out of styles.xml so date serials come back as dates. It does not read the formula element at all. Export runs the same way: a formula column is written out as the numbers it computed. So =B2*C2 goes in as 4200 and comes back out as 4200, and the workbook you hand a colleague is values, not a live model.

Financial models

Plainly: somebody building a financial model keeps Excel. A three-statement model is scenario switches, a recalculation loop, dynamic arrays and add-ins, and it gets checked by people who open it in Excel and expect the formulas to still be there. The eleven finance functions here — PMT, FV, PV, NPER, RATE, NPV, IRR, SLN, SYD, EFFECT and NOMINAL — will do a loan schedule. They will not do a model.

Where this one wins

A formula is a rule about a column

In Excel a column of formulas is a thousand copies of one idea, and the interesting bugs live in the four rows where somebody edited a copy. Here the column carries the formula and every row computes it against its own values. The per-cell form works too — type =B2*C2 into a cell and it evaluates — and where both exist the column wins, because a rule any single cell can override is not a rule. Fill still behaves: =B$2*$C2 dragged anywhere keeps the halves you pinned, rewritten over your own source text so spacing and capitalisation survive.

Every table is a sheet

There is no separate workbook file: every table in a project is already addressable as a sheet by its title. Costs!B2:B40 and [Costs]![Amount] both work, and two tables sharing a title get numbered rather than resolved by a coin flip. A formula column may reference another; a cell asked for while it is already computing returns #CYCLE! rather than hanging the tab.

Pivots that cannot go stale

A pivot here is a table block carrying a spec, re-derived from its source every render. Seven aggregates, five date buckets — day, week, month, quarter, year — and optional totals. Because it is a table, sorting, charts and .docx and .xlsx export already work on it. The trade: no slicers, no external sources, no snapshot you can freeze.

Column validation goes the other way from Excel's. Required, number, date, email, url, bounds, length, uniqueness and an arbitrary formula over the row all exist, but a failing cell is kept and marked rather than refused — a rule that throws away typing loses work on the day the rule is wrong.

Side by side

Eight of these twelve rows go to Excel, which is the honest shape of a spreadsheet comparison against the spreadsheet. The rest of the suite is in the full Microsoft 365 comparison, and how Google Sheets compares is the closer fight.

What crosses when you open a file

CSV is the better road in: the parser sniffs the delimiter across comma, semicolon, tab and pipe, and handles quoted fields, doubled quotes, CRLF and a byte-order mark. Going out, sheet names are sanitised the way Excel demands — five forbidden characters replaced, 31 characters maximum, no two differing only in case. What comes back is a real .xlsx, but not your original workbook. Function names are identical on both sides, so the 20 formulas you'll actually use apply either way.

Who should keep Excel

Keep Excel if a workbook runs macros, if you work past fifty thousand rows, if you build models, if you depend on dynamic arrays or INDIRECT, or if the file goes to somebody who will open it in Excel and expect the formulas to be there. None of that is a gap being quietly closed; it follows from a spreadsheet that lives in a document, in your browser.

This is for the other spreadsheet — the budget, the results table, the survey answers: numbers belonging to a document rather than a workbook, where the real cost is three copies that stop agreeing in March. A chart holds the id of the table it reads from, which is the whole reason the sheet sits in the document instead of beside it. All our tool comparisons has the rest of the set.

Questions people ask first