Scatter plot maker from CSV
Two CSV columns as a scatter chart, drawn on your device.
Open a CSV to plot one column against another
Drop it here or choose a file. The first file opens as a scatter chart. Set a number column as X, tick another as Y.
The sample: made-up afternoon readings in Belfast, 365 rows, humidity against temperature.
Tip: drop several files to compare them.
Your file never leaves your device. It is read and charted by your browser. Nothing is uploaded.
Open a CSV to plot one column against another
- Chart type
- Scatter: the points alone, with no line from one row to the next.
- Sample
- Belfast_Humidity against Belfast_Temperature, both chosen for you.
- Your own file
- The first one you open: one number column against the file’s usual X. You set which column is X and which is Y.
The scatter chart in brief
- Points
- One dot per row: the Y column’s value above the X column’s value, with no line joining them.
- X column
- Any number, date-time or duration column, or the row number. It need not be sorted.
- Y columns
- Up to 50 on one chart, each in its own colour, on a left or a right Y axis.
- Large files
- An overview first; zoomed far enough, every point of the range in view.
- Correlation
- Pearson’s r and r² of two columns, as numbers, from the Math dialog.
- Not built
- A trend-line option on the chart, or point colours and sizes taken from a third column.
- Image
- The chart as a PNG file, in the light or the dark colours.
- Processing
- Drawn by your browser. The CSV stays on your device.
- Price
- Free, with no account.
How to make a scatter plot from a CSV file
- Open the CSV. Drop a CSV on the tool at the top of this page, or choose it there. TSV, TXT, Excel, JSON, MATLAB, Parquet, HDF5 and LabVIEW TDMS files open the same way. With no file to hand, “Try a sample weather file” opens 365 made-up afternoons in Belfast.
- Look at what opened. The first file opened on the page starts as a scatter chart (for a later one, press Scatter over the plot), but the tool cannot know which two columns belong together. One number column is plotted, against the X that any chart of the file starts with: its first sorted date-time column, else its first sorted number column, else the row number.
- Set the X column. The top line of the column list names the X, for example “X: time”. Click “change” beside it and pick the number column that belongs on the horizontal axis. If that column was the one plotted, it moves to the X and the chart waits for a Y.
- Tick the Y column. Tick the column to plot against the X, and untick any that should not be there. Each ticked column is one set of dots in its own colour.
- Zoom in and save a PNG. Drag across the chart, scroll or pinch to zoom; a double-click goes back to the whole range. “Save as an image” in the plot’s ⋯ menu saves the chart as a PNG.
A worked example: temperature against humidity
The sample holds one reading at 15:00 for each day of 2025 in Belfast, all of it made up: 365 rows with the air temperature and the relative humidity. On this page it opens with Belfast_Temperature across the bottom and Belfast_Humidity up the side. The dots fall from the top left to the bottom right, so the warm afternoons were the dry ones.
| Axis | Column | Smallest to largest |
|---|---|---|
| X | Belfast_Temperature | 3.5 to 21.5 °C |
| Y | Belfast_Humidity | 58 to 99 % |
How closely do the two follow each other? Click Correlation (r) over the plot, choose the two columns and run it: Pearson’s r is -0.9094 and r² is 0.8270. A value of r near −1 means that as one column rises the other falls, almost along a straight line.
Checking the number in Python
These lines read the downloaded sample and work out the same correlation with NumPy, then a straight-line fit:
import numpy as np
import pandas as pd
d = pd.read_csv("megarows-sample-belfast-temperature-humidity.csv", skiprows=[1])
t, h = d["Belfast_Temperature"], d["Belfast_Humidity"]
r = np.corrcoef(t, h)[0, 1]
slope, intercept = np.polyfit(t, h, 1)
print(len(d), "rows")
print(f"r = {r:.4f}, r squared = {r * r:.4f}")
print(f"humidity = {slope:.4f} * temperature + {intercept:.4f}")The printed result:
365 rows
r = -0.9094, r squared = 0.8270
humidity = -1.8567 * temperature + 103.9989Run on 11 November 2026 with numpy 2.1.3, pandas 2.2.3, Python 3.12.15. A test repeats the correlation with the site’s own code whenever the site changes and compares it with NumPy’s value. Download the Belfast sample (CSV, 9 KB).
Choosing the X and Y columns
A scatter plot answers one question: when this column is large, is that one large too? The column you think of as the cause, or the one you control, usually goes on the X; the one that responds goes on the Y.
- The X can be any number, date-time or duration column, or the row number. It does not have to be in order: the axis runs from the column’s smallest value to its largest, wherever those rows sit in the file.
- The Y is every column ticked in the list. Two ticked columns give two clouds of dots against the same X; a column with a very different range can go on the right-hand Y axis.
- Text columns cannot be an axis of a scatter chart. A row with an empty cell in the Y column has no dot.
- Dated amounts, such as sales or orders, open as monthly totals, as they do in the plotter, and stay a summary here: the chart shows the totals, not the rows, until “Summarise the rows” is unticked in Plot settings.
Hover over a dot to read the values of the nearest point. An X chosen by hand is remembered in this browser for files with the same column names, so the next export from the same logger starts on that X.
Scatter plots of large files
A screen has far fewer pixels than a long file has rows, so a chart of a large file starts as an overview: a selection of real rows, never averages, that keeps the lowest and the highest values, about four points for each pixel of the chart’s width at most. The corner of the plot reads “Overview · zoom in for full detail”.
Zoom in and the range in view is read again from the file. Once it holds no more points than the chart has room for, every one of them is drawn and the corner changes to “Full detail”. Our own tests do exactly this with a scatter chart of 1,000,000 rows whose X column is in no order: the overview reaches from the smallest X to the largest, and a narrow zoom shows each point in range.
So the overview of a dense cloud keeps its highest and lowest values while the inside is thinned. Point density is not shaded; for the shape of one column on its own, the histogram maker counts every value. For files of many millions of rows as lines over time, the CSV plotter is the place to start, and the MegaRows home page has a 1,000,000-row sample to try.
The correlation is checked
Pearson’s r as SciPy and NumPy give it. The Math dialog’s correlation is compared with reference values on test signals, and with NumPy on the Belfast sample above. The full list of checks.
Your CSV stays on your device
The page reads the file with your browser and draws the dots there. No part of the CSV, and not its name, is sent to a server, so a file that may not leave the company laptop can still be plotted.
After the scatter plot
- CSV histogramThe spread of one column on its own, in equal bins.
- CSV statisticsThe mean, median and quartiles of both columns.
- CSV plotterThe same two columns as lines over time, or as bars.
- Large CSV viewerRead the rows behind a point: sort and filter the table.
- Sample filesMore made-up weather files to practise on, free to reuse.
Frequently asked questions
How do I make a scatter plot from a CSV file?
Open the CSV in the tool on this page: the first file opened starts as a scatter chart. Click “change” on the X line of the column list and pick the number column for the horizontal axis, then tick the column to plot against it.
Is the scatter plot maker free?
Yes. Drawing a scatter chart of two CSV columns, zooming into it and saving the PNG are all free.
Do I need an account to make a scatter plot?
No. The chart appears with no registration and no email address. Signing in is optional; it adds workspaces that keep a chart’s set-up.
Is my CSV file uploaded?
No. Your browser reads the CSV and draws the points on your own device, and the file’s name stays there too. The Network tab of the browser’s developer tools shows that no request carries the data.
Does it draw a trend line or R squared?
No trend-line option, and no R squared printed on the chart: a scatter chart here draws the points of the columns you tick. The Math dialog comes close in two ways. “Correlation (r)” over the plot gives Pearson’s r and r² of two columns as numbers. Polynomial fit, with degree 1 for a straight line, adds the fitted values to the plot as one more series, drawn on a scatter chart as dots along the line.
How many points can it plot?
There is no fixed limit on rows; the device’s memory is the limit. A chart holding more rows than it has room for is an overview of real rows, about four per pixel of its width at most, and says so in its corner. Zoom in until the range fits and every point in it is drawn. Our tests do this on a scatter chart of 1,000,000 rows.
Can I make a scatter plot from an Excel or TXT file?
Yes. An .xlsx or .xlsm workbook opens on its first sheet with data, and TSV and TXT files are read like a CSV, with the separator detected. JSON, MATLAB, Parquet, HDF5 and LabVIEW TDMS files open here as well.
Related tools and guides
- Plot a large CSV fileWhy thinned-out charts can hide spikes
- Plot multiple CSV files on one graphSeparate runs on one chart, labelled by file
- Excel to CSVDownload any sheet of a workbook as a CSV
- How we check our mathsOur maths checked against NumPy and SciPy
Scatter charts are drawn in the browser. The CSV, and its name, stay on the device.