A trustworthy CSV map starts before the first marker. You need to know which field is latitude, which is longitude, what coordinate system the numbers use, which rows are invalid, what attributes the export preserves, and whether the preview sends locations to somebody else.
Start with separate latitude and longitude columns in decimal degrees. A useful row might contain Site,Latitude,Longitude,Region,Volume. Keep identifiers, labels, categories, and values in their own columns so you can decide deliberately how each should appear. Do not combine coordinates with prose such as 13.7563 N, 100.5018 E near office unless you normalize it first.
This workflow is scoped to WGS 84 geographic coordinates. It does not convert UTM eastings and northings, Web Mercator metres, local survey grids, plus codes, addresses, or place names. A number can look plausible while belonging to the wrong coordinate reference system, so confirm the source documentation before plotting it. If the source does not state a system, ask its owner rather than treating a successful parse as proof.
Latitude measures north and south and must be between -90 and 90. Longitude measures east and west and must be between -180 and 180. Negative latitude is south of the equator; negative longitude is west of Greenwich. A swapped pair may still fall inside both legal ranges, so range checks alone cannot catch every reversal.
The CSV can place the columns in either order because the map asks you to select each role. Export formats then apply their own order. RFC 7946 defines GeoJSON positions in WGS 84 longitude then latitude order. The OGC KML test suite describes coordinate tuples as lon,lat[,height]. Never fix an apparent reversal by renaming headers only; inspect known landmarks and correct the actual mapping.
Treat blanks, words, infinity, degree symbols, and out-of-range values as invalid for a decimal-degree tool. Keep the source row number beside each finding so the correction happens in the original data. The map should skip a point only after recording why; quietly dropping rows makes the picture look more complete than it is.
Then review geographic plausibility. A customer list expected in Thailand should not contain a valid point in the Atlantic. Repeated coordinates may represent a shared building, rounded data, a default centroid, or a genuine duplicate. Zero latitude and zero longitude are legal and meet in the Gulf of Guinea, so do not classify zero as missing automatically. Complete-file validation matters even when the page shows only a bounded issue ledger for performance.
Choose a categorical field such as region, status, or site type for color. Keep the category names in a visible legend and in an exact point table because color alone is not an accessible identifier. If a file has many categories, group the overflow visually instead of pretending a long palette is readable.
Use a numeric field for relative point size only when the values have a shared meaning. Area-proportional symbols require more careful scaling than raw radius; a compact review map can use square-root scaling to avoid enormous markers, but it remains an exploratory cue rather than a statistical publication. Label only a selected point or a collision-free subset. The complete point list should remain available to keyboard and screen-reader users even when labels are bounded.
Location rows can reveal homes, clinical visits, incidents, infrastructure, customers, or valuable assets. A local file parser is not fully private if the preview then requests remote tiles around a file-derived viewport or calls a geocoder with source values. The csvtodashboard map uses a checked-in land outline and local SVG graticule. It sends no coordinate, bounding box, zoom, category, label, or map interaction to a provider.
The schematic basemap comes from Natural Earth's 1:110m land polygons. Natural Earth places its raster and vector map data in the public domain. That scale is appropriate for a compact world overview, not street, parcel, boundary-dispute, navigation, or survey decisions. Use a destination GIS only after reviewing the export and its sharing rules.
Choose GeoJSON for web maps, JavaScript tooling, and GIS workflows that accept a FeatureCollection. Each valid row becomes a Point with [longitude, latitude]; the original row becomes its properties object.
Choose KML for Google Earth and other KML-aware applications. Each valid row becomes a Placemark with a point tuple in longitude, latitude, altitude order and source fields in ExtendedData.
Choose SVG when you need the current schematic view in a report, slide, design tool, or repository. It preserves the visible map, legend-independent marker colors, labels, and selected-point state as editable vector markup.
GeoJSON and KML intentionally preserve original properties for valid points. That is useful downstream, but it can carry names, notes, identifiers, or internal classifications. Remove fields that the recipient does not need, then open a representative export and verify several points before sharing the whole set.
Google's official My Maps import help accepts CSV, TSV, KML, KMZ, GPX, XLSX, and Google Sheets. For tabular imports, the first row should contain titles and a column must provide latitude-longitude information, addresses, place names, or WKT geometry. Google also documents a 2,000-row import limit.
Use the local preview to catch coordinate order, skipped rows, and accidental far-away points before the upload. Choose direct CSV when My Maps should read the original table, or KML when you want reviewed placemarks from this workspace. My Maps is a destination upload: after you choose it, Google's service receives the imported data. That handoff is separate from the private preparation step.
Load or paste the CSV, choose separate latitude and longitude columns, review every invalid row, optionally choose label, category, and size columns, then inspect the local point map or download GeoJSON, KML, or SVG.
The workspace accepts WGS 84 decimal degrees: latitude from -90 to 90 and longitude from -180 to 180. It does not geocode addresses or convert projected coordinate systems.
The CSV may keep them in separate columns in either order because you map them explicitly. GeoJSON output stores positions as longitude then latitude, and KML writes longitude,latitude,0.
Blank, non-numeric, and out-of-range latitude or longitude values are reported with source row numbers and skipped. Every row is validated even when the visible issue table is bounded.
Yes. Choose a categorical column for color and a numeric column for relative radius. A visible legend and exact point table keep category identity available without relying on color alone.
No. The preview uses a checked-in simplified Natural Earth land outline, local SVG graticule, and browser-rendered points. It makes no tile, geocoder, viewport, or file-derived request.
Every original source property on each valid point is preserved. That is useful for downstream inspection but may include sensitive data, so review the exported file before sharing it.
No. Parsing, mapping, validation, rendering, selection, and export happen in the browser tab. Audience measurement receives the pathname only, never map data or interaction payloads.