What makes a generated chart editable, and what makes it a picture

There is one question worth asking about a generated chart. Click it, and ask to edit the data. Either a spreadsheet opens with your categories and values in it, or nothing happens — because what you were handed is an image that has been pasted onto a slide and cropped to look like a chart.

Both arrive as a .pptx. Both look identical in a screenshot. The difference only shows up when the quarter changes and somebody has to update the number. This page is about what has to be inside the file for the first outcome, why the second is so much easier to build, and the test that separates them in about thirty seconds.

Published 23 August 2026 BATech Studio Written by hand · no charting library
What one chart in our decks is made of
Rows in the workbook

One per point

Column A the categories, column B the values, B1 the series name. The same shape the chart's own formulas point at.

Charting libraries used

None

The chart XML and the workbook are both written by our renderer, which is why this page can describe the bytes rather than a feature list.

Jobs delivered

69 of 79

Every kind, all time, on the one renderer described here.

A chart is three parts, not one

An image is one part: a PNG in ppt/media/ and a rectangle on the slide that points at it. A real chart is three, and every one of them has to be written correctly or the chart is dead on arrival.

What a chart adds to the package 
ppt/
├── charts/chart1.xml           the chart itself — type, axes, series
├── charts/_rels/chart1.xml.rels points at the workbook below
└── embeddings/data1.xlsx       the spreadsheet the numbers live in

The slide points at the chart, the chart points at the workbook, and the workbook holds the numbers. Break any link in that chain and PowerPoint shows an empty frame where the chart should be — a failure that unzips perfectly and passes every validator, because nothing about the XML is malformed. It is simply pointing at nothing.

The chart does not contain its own numbers

This is the part that surprises people. The values you see plotted are not stored in the chart as data; they are stored as a formula pointing into the spreadsheet, with a cached copy alongside so the chart can draw before the workbook is opened.

What our chart part actually says 
<c:tx>  <c:f>Sheet1!$B$1</c:f>          the series name
<c:cat> <c:f>Sheet1!$A$2:$A$7</c:f>     the category labels
<c:val> <c:f>Sheet1!$B$2:$B$7</c:f>     the values

Those references have to resolve against a workbook that really exists, really is a spreadsheet, and really has those cells in those places. "Edit data" is PowerPoint following that pointer. If the workbook is absent, or is a file that only claims to be one, the menu item is there and does nothing.

So the workbook has to be a genuinely valid workbook

An .xlsx is a ZIP of XML, the same construction as the presentation around it. Ours is written by the same ZIP writer that builds the deck: a workbook part, a worksheet, the relationships between them, and the cells.

Two decisions in it are worth naming, because both are the kind of thing that is invisible when right and hard to diagnose when wrong. Text cells are marked t="str" so the file needs no shared-string table — the one part of SpreadsheetML genuinely worth avoiding for a file this small. And the workbook is stored rather than compressed inside the deck, because it is already a ZIP; deflating a deflated file costs processing on every export and gives nothing back.

The lesson, stated plainly

A chart that cannot be edited is not a worse chart. It is a different object wearing the same picture. The question to ask a generator is not whether its charts look good in the preview — it is what happens when you right-click one.

Three checks, on any deck, from anyone

01 Right-click the chart and choose Edit Data.

A spreadsheet should open with your categories down one column and your values down the next. That is the embedded workbook being followed. If the menu item is missing, the object is an image. If it is present and nothing opens, the chart part exists and its workbook does not.

02 Rename the file to .zip and look for one folder.

Unpack it and open ppt/embeddings/. There should be one .xlsx per chart in the deck. An empty folder, or no folder at all, beside slides that visibly contain charts means every one of them is a picture.

03 Change one number and watch the slide.

Edit a value in the workbook and close it. The chart on the slide should redraw. This is the check the other two only imply, and it is the one that matters in the month the figures move — which for most decks is the month after they were made.

Why a picture is the easy answer

Because rendering is a solved problem and writing chart XML is not. Any charting library will produce a good-looking PNG in a line or two; putting that PNG on a slide is another line. Writing a chart part means encoding the chart type, the axes and their scales, the series, the number formats, the colour references into the theme, the cached values, and the formulas that point into a spreadsheet you must also construct — and then getting all of it past several readers that disagree about what is optional.

The result is that "makes charts" is true of both kinds of product, and only one of them hands you something you can still use in three months. It is not usually stated either way, which is why the checks above are worth more than any description, including this one.

Where the numbers come from

Yours. The model that writes a deck is instructed at length never to invent a figure, and the renderer will not manufacture one to fill a chart that has no data. A chart appears because you supplied numbers for it.

That constraint is the whole reason the workbook can be trusted: it holds exactly what was given, in the order it was given, one row per point. A generated figure that looked plausible would leave the building under your name and in your branding, and no amount of editability makes that acceptable.

What it costs

During the alpha a credit is one US dollar. A slide deck is 25 credits, a branded document 10, a data report 15, a single image 1, and ten credits are granted free on signup — enough for a whole branded document before you decide anything. Charts are part of a deck and are not charged for separately. Every price here is an indicative estimate rather than a quote; the published catalogue is on the product page.

If you want the wider engineering account, the long version is next door. If you want the answer for your own numbers, generate a deck on the free credits and run the three checks above on it. That is a better test than anything written here.