Chart Advisor: Best Chart for Your Data

Upload

Start with a file or a sample

Any number of columns is fine. You confirm the few things a program cannot know for sure.

Drop an .xlsx or .csv file here, or click to chooseUp to a few hundred thousand cells works comfortably. Larger files should be summarised first.

No file handy? Try a fictional sample:

Privacy and limits

Everything runs on your computer in this page. Nothing is sent anywhere or saved. Formula cells are read as the last value Excel saved, so a file that was never recalculated can show empty cells. Old .xls and .xlsb files, password-protected files and pivot-table reports are not supported; save the data as .xlsx or .csv first.

Prepare

Check that the table was read correctly

The most common reason for a wrong chart is a wrong table. Confirm the header row, then look at what was excluded.

Sheet Header rowsHow many rows hold the column names Last data rowLeave blank to read to the end Numbers written as textFor example 1.234,5 or 1,234.5 Dates written as textFor 05/03/2026 Exclude total rows

Click a row in the preview to mark it as the header row (highlighted yellow). Red rows were excluded as totals or repeated headers.

Columns

Confirm what each column is

The types below are guesses. Fix any that are wrong; the recommendations follow your choices.

Date = a point in time · Period = an ordered label such as Jan, Feb or a year · Number = something you add up or average · Yes/No flag = an event you count as a rate · Category = a group · ID / text = not used for grouping · Ignore.

Focus

What do you want the chart to say?

Pick the measure, then optionally a grouping, a time column and a target. The grey/accent story is built from these choices.

MeasureWhat you want to show How to summarise it Group byOne bar or line per value Over timeDate or period column Group time byOnly for date columns Compare with a second numberDraws a scatterplot (optional) TargetIn the same unit as the result, e.g. 85 for 85% Which direction is good? Show top N groupsThe rest are combined as "Other"

Charts

Your best option, plus alternatives

The first card is the best practice for your choices. Alternatives carry the same message in a different shape. Switch between a typical default chart and the recommended design, edit the headline, then download the picture.

Press "Recommend charts" in step 4.

Chart types considered

★

Transparency

How it decides, and where it can be wrong

Every recommendation comes from a short list of rules. You can see them all, and none of them look at your business context.

If you chose…	You get	Rule behind it
One group-by column	Sorted horizontal bars (and a dot plot if values are close together)	Compare categories on one zero baseline. Bars must start at zero; a dot plot may use a labelled tighter axis.
A time column	A line chart (or columns for a few periods)	Lines show direction and pace. An incomplete last period is drawn dashed and left out of the headline.
Time + group by	Lines with one highlighted, or a heatmap	Many coloured lines become spaghetti. Grey for context, one accent for the point.
Two numeric columns	Scatterplot with the 3 biggest outliers labelled	Shows relationship and exceptions; correlation is not causation.
Nothing else	One big number with its target	One value is better stated than charted.
A target	Target line; groups that miss it are accented	A number without a benchmark cannot be judged.
Alternative designs	Up to two alternatives under the best practice; more behind "Show more options"	The best practice is always first and never replaced. Alternatives carry the same message in another shape, each with "Choose this when…" and "Trade-off…". Triggers: lollipop (6+ groups); Pareto (total or count, 5+ groups); strip plot (average or median per group, 8+ rows per group); small multiples and slopegraph (time + 3 or more groups); waterfall and stacked 100% (time + group, total or count).

What it cannot do

It does not know your audience or decision, so edit the headline before using it.
It cannot tell whether a column was entered correctly or what a code means.
Headlines state facts about the numbers; they do not explain causes.
It does not fix pivot-table reports, several tables on one sheet, or a mix of units in one column.

Good habits

Check the preview and the column types every time; a confident chart on a misread table is still wrong.
Keep the notes under each chart with the chart when you share it.
When groups have few rows, treat their values with care.
The design principles follow ideas in Storytelling with Data (Knaflic, 2015) and The Big Book of Dashboards (Wexler, Shaffer, Cotgreave, 2017). The rules and thresholds here are our own choices and may be wrong. This page is an independent tool, not affiliated with or endorsed by the authors or publishers.