Chart Builder

Turn a CSV into a clear bar, line, pie or scatter chart and export it as PNG or SVG.

Runs in your browser. Nothing is uploaded — your input is processed on this device and discarded when you close the tab.

Build a chart

Drop a CSV file here or click to browse. The file is read on this device only. Data is capped at 5,000 cells (rows × columns) so charts stay honest and the tab stays responsive.

Or paste CSV data

Pasted data is parsed exactly like an uploaded file, with the header option below.

Load pasted data Load sample data First row is a header row

Chart type Bar Line Pie Doughnut Scatter Histogram

Label / X column

Value / Y columns (tick one or more)

Load data first — the columns will appear here.

Aggregation for duplicate labels Sum Average Count of rows Minimum Maximum

When several rows share a label — three sales rows for “Jan”, say — aggregation decides how they become one bar or point.

Sort Data order (as in the file) Label A → Z Label Z → A Value, low → high Value, high → low

Top N (0 = show all)

Colour palette Colour-blind safe (Okabe–Ito) High contrast Ocean (blue–teal)

The default palette is designed to stay distinguishable under the common forms of colour-vision deficiency.

Chart title

X-axis label

Y-axis label

Load data to draw a chart.

Download PNG Download PNG (2×) Copy image Download SVG Transparent background (exports)

Exports are yours to use: charts you generate here may be used in documents, slides and reports, including commercially, with no attribution required. The 2× PNG renders at double resolution for print and retina slides; the transparent option removes the white background from both PNG and SVG exports.

Aggregated data behind the chart

The exact values plotted — check them before you trust the picture

Data preview — first 50 rows

Cells that are not numbers are highlighted, named in the warning above with their row and column, and excluded from the chart. Empty cells are skipped and counted separately.

Preview — first 50 rows

How to use the Chart Builder

1. Load your data. Drop a CSV file onto the dropzone, paste CSV text and choose Load pasted data, or use Load sample data. If you have just cleaned a file in the CSV Cleaning Studio, a banner offers the handed-off copy directly. Leave “First row is a header row” ticked unless your file has no headers, in which case columns are named Column 1, Column 2 and so on.
2. Pick the columns. Choose the label column — the categories or X values — then tick one or more value columns. Each ticked column becomes its own series of bars or its own line.
3. Choose an aggregation. If your file has several rows per label, decide how they combine: sum for totals, average for typical values, count for “how many rows”, minimum or maximum for extremes.
4. Pick the chart type and tidy up. Bar and line suit most category data, pie and doughnut show shares of a whole, scatter plots one numeric column against another, and histogram shows the distribution of a single numeric column across bins you control.
5. Sort, limit and label. Sort by label or by value, cap the chart at a Top N so a long tail cannot crush the leaders, then add a title and axis labels — a chart without them is a picture, not an explanation.
6. Check, then export. Read the aggregated data table and the warning line: both tell you exactly what was plotted and what was excluded. Then download a PNG for documents and slides, or an SVG when the chart needs to stay sharp at any size or be edited later.

Worked example: sales by month and channel

A small shop exports six rows of sales data: one row per month per channel, with Sales and Returns columns. Charted raw, row by row, the file would draw six bars with the month labels repeated — hard to read and easy to misquote. Aggregated by month, it becomes a three-bar story. Load it with the tool itself:

The sample is already set up the way the example needs — Month as the label column, Sales ticked as the value, aggregation set to Sum — so the chart shows Jan 200, Feb 240 and Mar 280: January is 120 online plus 80 in store, February is 150 plus 90, and March is 170 plus 110. Switch the aggregation to Average and the same bars become 100, 120 and 140, the typical month per channel rather than the total. Now untick Sales, tick Returns instead, and keep Sum: the chart shows returns of 8 in January, 10 in February and 9 in March — the returns did not follow the sales climb in March, which is exactly the kind of pattern a raw six-bar chart hides. Tick both columns together for a two-series view, or sort by value, high to low, with Top N set to 2 to keep only February and March. The aggregated table under the chart always repeats the plotted numbers, so anything you quote from the picture can be checked against the table in seconds.

What the Chart Builder does, and why it matters

Most data that matters to a small team already exists as a CSV: the sales export, the survey download, the server log summary, the spreadsheet someone emailed. The hard part was never storing those numbers — it is seeing them. A column of forty figures communicates almost nothing; the same figures as bars make the largest, the smallest, the trend and the odd one out visible in about a second. That speed is not a cosmetic extra. People decide from the picture — in the meeting, in the report, in the funding deck — so a tool that turns a file into a defensible chart quickly, without an upload or an account, removes the two usual failure points: the chart that never gets made because the software was too heavy, and the chart that gets made carelessly because nobody checked what it actually plotted.

The step this builder treats as first-class is the one most quick chart tools skip: aggregation. Real exports are rarely one row per category. They are one row per transaction, per day, per channel — dozens of rows competing for the same label. Plotting them raw produces either a forest of overlapping bars or, worse, a chart that silently shows only some rows and looks authoritative while doing it. Here, every row for a label is combined by a rule you choose and can see — sum, average, count, minimum or maximum — and the aggregated table beneath the chart prints the resulting values in plain figures. If a number in a presentation is challenged, the defence is not “the tool said so”; it is a table anyone can recompute from the source file.

Honesty about bad data is built the same way. A cell containing “many”, “TBC” or a stray note cannot be plotted, so it is never quietly treated as zero — a substitution that would drag down averages and invent totals. Instead the cell is highlighted in the preview, named by row and column in a warning, counted, and excluded, while empty cells are skipped and counted separately. You keep the decision: fix the file (the CSV Cleaning Studio sits one step earlier in the workflow for exactly this), or accept the exclusion knowing precisely what it was.

Finally, the output respects where charts end up. The PNG suits slides and documents; the SVG is a genuine vector redraw of the chart’s data — bars, slices, lines and points as editable shapes — so it stays sharp in print and can be restyled later. The default palette is the Okabe–Ito set, designed to remain distinguishable for readers with colour-vision deficiency, because a chart that only works for some of its audience has only done some of its job. Everything runs in your browser: the file is never uploaded, which makes the tool safe for sales figures, internal metrics and other data that should not be pasted into a stranger’s server just to draw twelve bars.

Common errors and limitations

• The chart looks like a forest of tiny bars. Your file has many rows per label and aggregation is doing its job, but there are simply too many labels. Sort by value and set a Top N, or aggregate the file further in the CSV Cleaning Studio first.
• Bars show nulls or gaps for some labels. Those labels had no numeric values in the ticked column — every cell was empty or non-numeric. The warning line names the excluded cells; a group with no usable numbers is shown as a gap rather than a misleading zero.
• Pie slices do not match your totals. Pie and doughnut use the first ticked value column only, and only positive values can be slices. Negative or zero groups are omitted from the slices — check the aggregated table before quoting shares.
• Scatter shows fewer points than you have rows. A point needs both coordinates to be numbers. Rows where either the X or the Y cell is empty or non-numeric are skipped and reported in the warning count.
• The file is refused. Data over 5,000 cells (rows × columns) is rejected with an explanation. The cap keeps aggregation and rendering instant on ordinary hardware; trim the file or split it, chart the part that carries the story, and note the cap beside the chart if the subset matters.
• Averages hide spread. Average aggregation says nothing about variation: two groups can share an average with completely different shapes. When the spread is the story, switch to a histogram of the underlying column instead of arguing from a single bar.
• Sorting is by the first value column. With several series ticked, value sorting orders labels by the first ticked column. Untick the others temporarily if you need to rank by a different series.

Frequently asked questions

Related tools

Related articles