Documentation / Data / Minitab worksheets

I — Both directions

Your worksheets are not trapped in either tool

Open an .MTW worksheet in SVEND, run the study here, and take a worksheet back out. Most quality engineers have years of work in Minitab files and no intention of abandoning it, which is a reasonable position and not one this page is trying to talk anyone out of.

Status: Available · Both directions

.MTW  →  a SVEND dataset  →  any analysis on the bench
.MTW  ←  a dataset, a gage study, or a DOE design

The relevant property is the file format as a boundary that can be crossed in both directions without retyping. A study created here opens there, and a study created there opens here.

II — Bringing a worksheet in

Drop the file on the Analysis Workbench’s dataset rail, the same place a CSV goes. Columns keep their names, numeric and text columns are distinguished, and the dataset is then an ordinary dataset: every analysis on the bench can read it.

Case	What happens
The common generation	Read directly. The layout is self-describing and uncompressed, and the reader was derived from the bytes of a sixty-file survey corpus rather than from a specification.
The later OLE generation	Refused by name rather than as “malformed”. The refusal states which generation the file is, because that distinction is only worth drawing if it reaches the person holding it. Save as the earlier format, or export a CSV.
A missing cell	Arrives as absent rather than as a number. See below. This is the one that matters most.
Something that is not a worksheet	Refused before anything is allocated.
A very large file	Refused against the same size ceiling every upload on the platform uses.

A * cell is not a measurement of 1.23456×10³⁰. Minitab writes a missing value as that sentinel, and it decodes perfectly cleanly, it is simply enormous. Handed to a control chart or an R&R as a reading it destroys the mean, the limits and the verdict while reporting no problem at all. SVEND decodes it as absent. Across the survey corpus, a quarter of the readable worksheets carried at least one.

III — Taking one back out

From	You get
A dataset	The table as a worksheet, column names intact.
A Gage R&R study	A worksheet with `RunOrder`, `StdOrder`, `Part`, `Operator` and the readings — the column shape a Gage R&R expects, in worked order. The measurement column takes the name of the declared characteristic, so a reader sees a bore rather than an unqualified “measurement”.
A DOE design	The design as a worksheet, ready to carry to whichever tool the rest of the analysis lives in.
An analysis result	CSV, Excel or JSON — see import and export.

The gage export is the interesting one. Issue the randomized run sheet here, capture the readings off the instrument, and the worksheet that comes out is already in the shape the analysis wants, with the standard order preserved beside the run order, so a re-randomization cannot detach a value from its part. The resulting study opens for a Minitab user with no further modification.

IV — What we checked

Reading and writing a format is easy to claim and easy to get subtly wrong, so the number worth publishing is the round trip: every readable worksheet in the survey corpus was read, written back out, and read again.

47 of 47 worksheets round-tripped identically 50,646 cells compared, of which 20,944 absent 0 values altered

What that number is and is not. It is a statement about the reader and the writer agreeing with each other across real files, including on where the gaps are. It is not a claim of coverage over every worksheet Minitab can produce: the twelve OLE-generation files in that corpus are refused rather than read, and they are counted as refused rather than quietly left out of the total.

V — Where this sits

Minitab is a good tool with a long history and an enormous installed base, and a quality engineer who knows it well is not doing anything wrong. Plenty of people will use both: a bench that is always open in a browser for the everyday work, and a desktop package for the things it has done well for decades. That is a perfectly sensible way to work, and the file format is what makes it possible.

The direct comparison (what each is stronger at, and where each is the better choice) is on the comparison page, which concedes the cases where Minitab is the right answer rather than pretending there are none.

VI — Boundaries

The .MTW layout is not published as a specification. This reader was derived from the bytes of a public survey corpus, so it is honest to say it covers the files we could obtain rather than the format in the abstract.
The later OLE-compound generation is refused by name. Reading it is a different mechanism and is not built.
Minitab project files (.MPJ), graphs, session output and stored macros are not read. A worksheet is a table; a project is not.
Minitab is a trademark of its owner. Nothing here implies review, certification, sponsorship or endorsement, and no compatibility claim is made beyond the worksheet reading and writing described above.

CSV, Excel and JSON → The Gage R&R study bench → The Analysis Workbench →