Capture a cell range instead of the whole web table

Full-table capture is a habit from a time when tables were small. A modern page will give you 40 columns because someone in product added badges, sparkline hooks, an empty “select all” checkbox, and three columns that only populate for premium rows.

I do not want those columns in my file. I also do not want the first twelve rows if they are a frozen summary. I want the block I came for.

Cell range mode is the unglamorous feature that does that. You start it, you drag, you inspect the selection, you export. It feels like selecting cells in Excel, which is the entire compliment.

When a range is the right unit

A league table where I need positions 1 to 20, not the play-off footnotes.

A price list where I need SKU, name, and net price, not the “compare” widgets.

A PDF schedule where the useful grid sits under a title block that the extractor also thinks is a table.

A research page where one wide table holds three topics side by side.

An admin grid where the first column is a checkbox and the last column is a trash icon.

If I capture everything, I will spend the repair pass deleting. I can skip that by not capturing it.

How I drag without lying to myself

I zoom the page until the cells are easy to hit. I start on the header I want, not on the decorative title above it. I end on the last data row I trust, not on the “Load more” control.

If the table is wider than the window, I scroll horizontally before I start so the first drag includes the columns I care about. A half-drag that stops at the overflow is how you get a file missing unit_cost.

I inspect the range in the editor before I download. The drag is a proposal. The editor is the decision.

Table Capture Chrome has range mode on web tables and on PDF captures. You can still run Capture Score on that slice, which is the part people skip. A small range can still have a bad type.

A range that saved a meeting

A procurement page showed a master table of 80 SKUs. I needed the twelve rows in the “filters” category. Filtering the page was flaky. It kept resetting when the session token refreshed.

I sorted the on-page table by category, dragged those twelve rows plus the header, and exported XLSX. The file had four columns. Nobody asked why the checkbox column was missing. It was missing because I never selected it.

If I had captured all 80 rows, someone would have filtered in Excel and then mailed the unfiltered file “just in case”.

Ranges on PDFs

PDF extractors love the whole page. Titles become rows. Page numbers become cells. A dragged range around the actual grid cuts that noise.

On multi-page PDFs I range each page if the table repeats with a new header. Then I stack the files, or I recapture page by page into the same workbook after export. It is not elegant. It is correct.

If the PDF used a double header (group names on row 1, field names on row 2), I include both rows in the drag and fix the header in Table Studio. Dropping row 1 too early loses the group context.

Ranges on DIV grids

Card layouts make dragging feel less precise because the hit targets are large. I still try. If the drag keeps including a neighbouring card, I widen the window and switch to a list breakpoint if the site has one.

Sometimes a range is the wrong tool and a full capture plus column delete is faster. I do not force it.

What I do in the editor after a tight drag

I still check types. A twelve-row slice can contain the only POA in the column.

I still rename headers if the site used Amt and I want amount_gbp.

I still drop a totals row if I accidentally included it because it looked like data.

Then I export. CSV for a script. XLSX for a person. Clipboard if Sheets is already open.

Partial capture is not incomplete work

People treat a subset as cheating. In practice, a subset is closer to the question you asked. “What do the mid-tier plans cost?” does not require the enterprise row with a blank price and a phone number.

The file you did not create cannot be leaked, mis-sorted, or sent to the wrong channel.

If you later need the rest of the table, recapture. Do not keep a bloated “archive” extract that you do not trust.

Drag the block. Look at it. Save that. Leave the rest of the page where it is.

A worked range on a comparison grid

A SaaS pricing page put four plans in columns and a feature list in rows. I did not need Enterprise. I needed Starter and Team, plus the rows for seats, storage, and SSO.

I dragged a rectangle that started at the feature names and ended at the Team column. The editor showed three columns. SSO was No and Yes. Seats were numbers. Storage was 5 GB and 50 GB, which I split later if I am modelling cost per gigabyte. For a meeting, I left the units in the cell.

If I had captured the whole page I would also have imported a FAQ accordion that the extractor treated as a two-column table. That accordion had nothing to do with the quote.

Sticky headers and the false first row

Some grids keep a sticky header that is a duplicate of the real header. A drag that starts too high captures both. You get two header rows and Excel treats the second as data.

I look at row 1 and row 2 in the editor. If they match, I delete row 2. If row 1 is a group label (Pricing, empty, empty) and row 2 is the real fields, I merge the meaning into new header names (starter_gbp, team_gbp) and drop the group row.

Frozen summary rows at the top are the same family of problem. A dashboard table often shows “This week” totals above the line items. Those totals look selected because they sit in the same visual grid. They are not line items. I start the drag below them.

Keyboard habits that help

I use the browser zoom, not the Windows display scale, so the cells get larger without changing the layout breakpoints as much.

I turn off the page’s own “compact density” if it has one. Compact density is pretty and hostile to a precise drag.

If the table virtualises rows, a range of “everything I can see” is the honest range. I scroll, recapture the next slice, and stack files. Pretending I selected 2,000 rows when 40 exist in the DOM is how you get a confident, wrong workbook.

Naming the file after the range

orders.xlsx is a lie if the file is “orders for region North, August, columns A to F”. I use orders-north-2026-08-cols-af.xlsx or a short equivalent. The next person should not have to open the file to learn what was excluded.

If the range is a one-off for a ticket, I put the ticket id in the filename. Then the capture has a home and I can delete it when the ticket closes.

Range mode is a way of making the question smaller until the table fits the question. That is the whole technique.