A scatter plot maker approach is the combination of input format, axis behavior, and verification steps you adopt for a specific dataset, and choosing the right one begins with understanding how the tool parses your rows, scales your axes, and reports what was kept or skipped. The XY Scatter Plot Maker at Lizely turns strict x,y numeric pairs into a vector scatter plot with independent linear axes, so the approach you choose has to fit both your data shape and your goal: small experiments, paired observations, calibration samples, classroom demonstrations, rough correlation checks, or quick visual inspection. Because every non-empty row must contain exactly two comma-separated numbers and only the first 200 are actually checked, the format you paste decides what survives. Because the smallest x maps to the left and the largest to the right on an independent linear scale, unequal numeric gaps stay unequal. Because the tool does not include zero automatically, the labels expose the displayed range. Picking the right approach is therefore a matter of matching your data shape to the parsing rules and then verifying three separate counts before exporting.
Match the Approach to Your Data Shape
The first decision is what shape of data you actually have, because the tool's behavior is shape-dependent. A scatter plot maker approach is not one-size-fits-all: it shifts based on whether your x and y values vary, stay constant, or include outliers. Once you can name your shape, the rest of the workflow becomes mechanical rather than guesswork.
| Data shape | What the tool does | Right approach |
|---|---|---|
| Single valid point | Centers it using constant-axis expansion on both axes | Treat it as a quick reference marker; verify both axis labels |
| All x values identical | Expands the x domain symmetrically around the constant | Do not pre-flatten to zero; trust the centered layout |
| All y values identical | Expands the y domain symmetrically around the constant | Apply the same rule independently to the y column |
| Wide numeric spread on both axes | Linear scaling, unequal gaps preserved | Paste raw values; the gaps are the signal you want to see |
| Tight cluster with small range | Axis labels may not include zero | Read both axis ranges before judging density or correlation |
| Over 200 non-empty rows | Only the first 200 parsed; the rest counted as over-limit | Trim or split the dataset before pasting |
| Mixed invalid rows | Skipped and counted separately from over-limit | Fix formatting issues; never assume silent truncation |
Naming your shape up front is what separates a chart that argues your case from one that quietly distorts it.
Respect the Strict Parser and Row Limits
The second decision is whether your pasted text will survive parsing. The tool is intentionally strict so it can distinguish real data from accidental text, which means the format you paste decides what survives. Every non-empty row must contain exactly two comma-separated numbers. Signed values, decimals, zero, and scientific notation are accepted. Empty components, extra commas, unit text, hexadecimal values, Infinity, NaN, and absolute magnitudes above 1e12 are invalid. Blank rows are ignored. Negative zero is normalized to zero.
This strictness matters because the tool's summary always reports three separate numbers: valid points, invalid rows, and over-limit rows, including zero. Invalid rows do not become zero points or jump to an axis; they are skipped and counted. Rows past the boundary are not mixed into the invalid count because they were deliberately not parsed. If you paste a single block of 5,000 rows, only the first 200 are checked, the remaining 4,800 are counted as over-limit, and the generated summary shows that distinction on the next click. Nothing is silently truncated or represented as if all pasted rows were used.
The input box itself has a hard ceiling of 50,000 UTF-16 code units, and inputs above that limit are rejected before parsing begins. For very large datasets, trim to your most relevant subset first rather than expecting the tool to silently page through the rest. The right approach is to size the paste to the check, not the check to the paste.
A related decision is text safety. The title is XML-escaped before it enters the SVG, which means ampersands, angle brackets, quotes, and apostrophes are converted to entities and code points forbidden by XML are replaced with the replacement character. A title containing closing tags or script-like text therefore stays visible text and cannot create SVG elements in the generated markup, so you can paste descriptive labels without worrying about the chart.
Choose the Approach and Build the Chart
Once your data shape and format are settled, the chart itself takes only a few deliberate steps. The Scatter Plot Maker walks you through the same workflow whether you are plotting calibration samples, paired observations, or classroom data, so the difference between a clean export and a wasted click is mostly about what you do before you click Generate.
- Enter a title in the title field. Keep it descriptive so the exported SVG carries context with the chart, and remember that closing tags or script-like text are rendered as visible text, not as markup.
- Paste your data into the input area with one strict x,y numeric pair per non-empty line, in a form like 1,3 or -2.5,4.8e2. Blank rows are ignored.
- Click Generate to parse the input. The summary will report valid points, invalid rows, and over-limit rows as three separate numbers, including zero.
- Verify the axis ranges shown in the summary and the five reference labels drawn on each axis before reading anything from the chart.
- Inspect representative points in the on-page preview and open the SVG to read each circle's text title containing its exact formatted x and y.
- Download the standalone SVG. The downloaded file mirrors the preview, since both reuse the same percent-encoded SVG data URI derived from one generated string.
If the summary shows zero valid points, generation reports an error and clears any older chart, so a clean run always starts from a clean input.
Read the Axis Behavior the Tool Actually Uses
A scatter plot maker approach also has to account for how the tool actually places points, because a chart that hides its own scaling rules will quietly mislead. Both axes use independent linear scales. The smallest x maps to the left plot boundary and the largest x maps to the right; the smallest y maps to the bottom and the largest y maps to the top. Values between those limits are placed in direct proportion to their numeric distance. Five reference lines and labels expose each displayed range.
This is the reason unequal numeric gaps remain unequal on the chart, and the reason a tight cluster can look misleading without zero on the axis. The axes do not auto-include zero, so a point that looks "high" may simply sit near the top of the displayed range rather than the top of a zero-to-max range. Inspect the labels before interpreting the apparent position or spread of a cluster. For more on what the spacing actually means, see Are X Positions Equally Spaced in a Scatter Plot Maker?
Constant axes get special handling because their minimum and maximum would otherwise be equal, and division by zero would make coordinates invalid. If every x is identical, the tool expands the x domain symmetrically around that constant and centers every point horizontally, using ten percent of the absolute constant or one unit, whichever is larger. The same rule applies independently to a constant y. A single valid point therefore appears in the center of both axes instead of failing.
Coordinates are rounded to three decimal places in the SVG, and axis labels use practical significant digits or scientific notation for very large and small values. Calculations use JavaScript double-precision floating point, so this is a visualization, not a precision-preserving data format. Keep the original rows for analysis, uncertainty, reproducibility, and exact values, and treat the chart as a visual record rather than a numeric one.
Verify Three Counts Before You Trust the SVG
The fourth decision is verification. Before exporting the SVG, confirm three things in the generated summary: the valid point count matches what you expected, the invalid row count is zero or fully explained, and the over-limit row count is zero or acceptable. Editing the title or the data clears the generated string and disables the old download, so a clean re-run starts only after you confirm the summary.
Overlapping points can hide one another, especially when identical pairs repeat. Read several point titles in the SVG-aware preview to make sure the points you expect are the points you actually see. If your data is dense or categorical, the tool is not designed for color, accessible labeling, logarithmic scales, or publication styling, so a dedicated plotting package is the better choice for those cases. The chart does not calculate correlation, regression, a trend line, confidence intervals, error bars, density, clusters, or statistical significance; it only plots supplied pairs on linear axes and makes no statistical claim. For an after-the-fact checklist that complements this one, see Check a Scatter Plot Result After You Generate It.
All parsing, layout, XML creation, preview, and download preparation happen locally in the current tab, so no points or title are uploaded to Lizely.
When the Right Approach Is a Different Tool
The right approach is sometimes to use a different tool, and pretending otherwise wastes a generation. The XY Scatter Plot Maker is built for small, strict, linear, two-axis plots. It accepts up to 200 checked rows, refuses magnitudes above 1e12, ignores categorical color, and produces a single series of blue circles. If your goal is multi-series comparison, categorical color, accessible labels, logarithmic axes, regression overlays, statistical inference, or formal publication styling, use a dedicated plotting package and retain the source dataset. If you only need bars or lines instead of points, a Bar Chart Maker or Line Graph Maker will save you the friction of forcing the wrong shape of data into a scatter chart.
Within its scope, however, the approach is simple: name your data shape, format strict x,y rows, generate, verify three counts, and download the same SVG you just previewed. The output is a standalone 800 by 500 SVG with a white background, title, grid, numeric axis labels, and point markers. Per the W3C SVG 2 specification, vector formats like this scale cleanly without re-encoding, which is what makes the standalone SVG a portable end product for slides, worksheets, or a static report image. The on-page preview and the downloaded file are not reconstructed separately; both use the same percent-encoded SVG data URI derived from one generated string, so what you see is what you save.