ToolboxHub's CSV preview displays at most 500 matching data rows at a time. That limit applies to the table on screen; it does not mean the loaded data stops there, and it does not define what the Export CSV button downloads.
Read the count before diagnosing a missing export
Clear the search box and compare the displayed count with the number of records expected from the source. The first parsed row becomes the header, so the preview's data-row count excludes that row.
For example, a workshop system exports 1,240 registration records plus a header. A successful load of that simple dataset can report 1,240 rows while showing only the first 500, followed by a notice about the remaining rows. Scrolling to the bottom of the table does not load another page of results. Repeatedly reopening the same file will not remove the cap.
First confirm that 1,240 really describes this export. A source dashboard might count cancelled registrations as well as active ones, while an export contains only the active group. Record the export scope and time alongside its expected count. A mismatch between two differently scoped reports is not evidence that the browser lost rows.
Open the file with CSV / Excel Online Preview and note its name, the count with search cleared, and a few known record IDs. Keep the original file unchanged while investigating. If the parsed count itself is unexpectedly low, the display cap alone does not explain the problem.
Search for a known record beyond the initial 500
Search runs across the loaded data rows before the tool takes the first 500 matches for display. A sufficiently specific record ID can therefore reveal a row that was outside the initial visible sample.
Use one known ID near the beginning, one around the middle, and one near the end of the source export. In the registration example, these could be R0002, R0620 and R1239. Enter each separately, read the returned fields, and clear the query between checks. Use synthetic IDs like these for a demonstration rather than exposing attendee details.
Matching is case-insensitive and uses a substring found in any cell. Searching for R12 can return several IDs; it is not an exact-match query against a designated ID column. A match in a notes field can also make a record appear. Read the actual identifier and its companion fields instead of accepting the result count as proof that you found the intended person.
The numbered column starts at 1 again for the displayed results. A record shown as row 1 after a search is the first match in that view, not necessarily the first record in the original file. Avoid writing instructions such as “fix row 1” based on a filtered screenshot; record the stable ID instead.
Check the selected sheet and the header shape
For a workbook with multiple sheets, counts and searches concern the currently displayed sheet. Select the intended sheet explicitly and clear any previous search before comparing counts again.
The component turns a worksheet into arrays of cell values using SheetJS. Its header: 1 option corresponds to the array output described in the SheetJS arrays-of-data documentation. ToolboxHub then uses the first parsed row as the visible column headings; it does not ask you to designate a different header row.
That matters if an export begins with a report title rather than field names. A heading row with fewer cells than later records can leave some values outside the rendered columns. A row might even match a search through a cell you cannot see in that narrow table. Return to the producing application and obtain a clean data export with the intended header rather than treating a plausible-looking preview as a complete view of every field.
For a plain CSV there is no multi-sheet workbook to navigate. If the expected record is still missing, check that you opened the right export and compare its source scope before assuming a worksheet-selection issue.
Export CSV includes more than the search results
The download uses the full current parsed sheet, including its first row. It does not use the active search filter or the 500-row rendered sample.
Suppose a search shows three workshop registrations. Clicking Export CSV still produces a derived file from all loaded rows in the current sheet. Do not send that download to a colleague expecting a three-record sample: it may include the rest of the roster. Create an authorized subset in the source application, save it separately, and verify that separate file before sharing.
Likewise, a download containing more than 500 rows is not inconsistent with the preview. The two actions use different scopes by design. Reopen a derived download separately and compare its expected record count and known IDs, but remember that checking it through the same parser is not an independent validation of the source.
The viewer does not edit the original CSV or XLSX. Its output is a newly serialized CSV, not an original-byte backup or a preserved workbook. Keep the producing system's export and any source workbook; do not overwrite them merely because the browser offers a download.
Decide whether you have enough evidence to continue
A matching count and successful searches support a limited spot check. They do not establish that every field in every record is intact, especially when an import has consequences beyond displaying a table.
Before a full transfer, reconcile the source and destination counts under the same scope, check duplicate and missing identifiers, and inspect rejected records through the destination's approved process. Include a sample with unusual punctuation or empty fields. If values or counts disagree, stop and investigate the source export or receiving application's interpretation.
For a small, non-sensitive excerpt, Text Diff can help compare two text versions. It does not reconcile entire workbooks or understand that differently formatted values mean the same thing. Keep the comparison small enough to inspect and use the actual data system for a complete audit.
Frequently asked questions
Does the tool discard everything after row 500?
The rendering code limits the visible table to 500 matches. That is separate from loaded data. Check the count with search cleared and search for a known later record; neither step alone proves the source export is complete.
Can I click a next-page button to see the rest?
The current component has no pagination control. Use a specific search to locate a record, or use an approved data application when you need to inspect the entire dataset.
Why does searching change the row count?
The displayed count becomes the number of matching loaded data rows. Clear the query to return to the unfiltered count; the header is not included in that data-row figure.
Will the download contain only my search results?
No. Export CSV serializes the full current parsed sheet. Prepare and verify a separate subset outside the viewer if you intend to share only selected records.
Can I use this preview to prove a large import succeeded?
No. It can locate records and expose suspicious structure, but a full import needs reconciliation against the destination's stored data and rejected-row report. A clean preview is only one check.