Stop fixing spreadsheets by hand.

The errors that quietly make a report wrong, found and marked in the sheet itself: red for errors, amber to check, blue for tidying.

• A typed-in number among formulas, which stopped updating
• A total whose SUM stops short of the rows beside it
• Amounts and dates stored as text, so they never add up
• An ID used twice, naming the row it clashes with
• The same name spelled two ways, splitting every chart
• Values far outside the rest, and dates years out

Jump to any cell, fix a whole kind in one click, or ignore it for good. Outliers and duplicate IDs are reported without a fix, because only you know which row is wrong.

Marks are borrowed, not taken: each cell's own fill is remembered first, so clearing them puts your formatting back. And undo is ours, not Excel's: every fix snapshots the cells, so they come back exactly as they were.

One click repairs the whole export and puts a clean, properly typed Excel table on a new sheet. Every problem is listed in plain words with a count and examples first, and you can switch off any fix you don't want. The original is never touched.

• Title, "Generated by" and note rows
• Headings split over two rows
• Subtotal and grand total rows
• Group headings like Region: Europe
• Headings repeated mid-table
• Numbers typed as text: $1,204.50 1.130,00 € (250) 980- 1,2O0
• Dates in any format, day- or month-first
• The same name spelled three ways
• N/A and - placeholders
• Duplicate and blank rows

MEx reads what each column holds and suggests up to eight charts that together tell the story: the trend over time, a comparison, a share, a two-way breakdown, a top 10, a relationship. Every one comes with the reason it was suggested.

You decide what goes in. Say "build 1, 2 and 5", "only the first four", or ask for your own: "add a doughnut of Units by Channel". Choose which columns it may use, and how many charts you want. Nothing is fixed.

It works straight from a messy export and never adds up prices or percentages. Colours stay the same for the same thing in every chart. The dashboard is saved in the workbook: next month, paste in the new rows and say refresh - charts find their columns by name, so a new export with them in another order still works.

Up to 8 charts, 4 headline numbers, one tidy grid. The numbers behind every chart go on a companion sheet so you can check them.

The expensive part of a monthly report isn't the thinking, it's the repetition: the same clean-up, the same combine, the same dashboard, in the same order, from a file with the same shape and different numbers.

So MEx remembers. Every command that works is recorded as you go — there's no mode to switch on — and /recipe save month end keeps the sequence. Next month, one line runs the lot.

What's stored is the command, not the result. "Clean up this sheet", not "clean A1:F250" — so a recipe still works when the sheet is four hundred rows longer. If a step needs a sheet this workbook hasn't got, it's skipped and named rather than pointed at whatever sheet looks similar.

Recipes live inside the .xlsx, like the dashboards do. Send the file to a colleague and their copy of MEx can run the same routine. No account, no server, nothing to sync.

"Last month's file and this month's disagree by 412 and I can't find where." MEx matches the rows on a key column and reports every cell that moved, with both values and the amount of the change — plus the rows that appeared and the rows that went away, kept separate, because a missing row and an edited row need different explanations.

Numbers are compared with a tolerance, so a figure rounded to the penny in one export and not the other isn't reported as a difference. Rounding noise is silence; real movement is a line in the report.

The differences go on their own sheet, ready to send to whoever asked, and the changed cells are highlighted in the newer sheet so you can go and look.

If no column is unique enough to match rows on, MEx says so and compares whole rows instead — which finds what was added and removed, but not what was edited. It tells you that rather than quietly reporting less.

The monthly reporting deck is the other half of the job. Build it once by hand, then rebuild it every month because the numbers moved. Copy-paste takes an hour and is wrong the moment anything changes; paste-link breaks when the workbook is renamed, moved, or its ranges shift.

MEx Automate for PowerPoint takes the spreadsheet and builds the slides: a title, the headline numbers, a slide per chart with the plain-English finding already written underneath, and a closing slide of what the numbers show. Same engine as the dashboards, so a messy export works directly and the deck says exactly what a dashboard from the same file would say.

Or just paste what you have. Notes from an email, an outline, a list, a table copied out of Excel. MEx reads the structure the way you wrote it: a short line with a list under it is a slide title, an indented line is a sub-point, numbered lists stay numbered, a long list becomes two slides, and a pasted table becomes a table and a chart of its numbers. Every word on the slides is a word you pasted — nothing is invented, so there is nothing to fact-check.

Charts are drawn, not pasted. PowerPoint has no way for an add-in to insert a chart, so instead of a screenshot you get ordinary PowerPoint shapes — rectangles, lines, text — in your own theme. Recolour them, move them, animate them. They stay sharp on a projector and a screen reader can read them.

Next month, press Refresh. Each slide remembers what it shows — "revenue by region, from the sheet named Sales" — not a path to a file. Rename the workbook, move it, add four hundred rows, reorder the columns: pick the new file and the slides redraw. Nothing to relink, because nothing was ever linked.

Only the shapes MEx drew are replaced, so a note or an arrow you added to a slide survives the refresh. Slides this month's data can't produce are left alone rather than deleted — that's worth seeing, not hiding.