Open your datasets as a sheet

Pivot a dataset in a live grid, keep the views you need as tables that refresh, type in your own targets, and ask about any of it in a conversation.

Sometimes you want to look at the data yourself: buyers by channel and month, laid out as a grid you can regroup in a click. A sheet is that grid. You open one of your datasets as a sheet and pivot it, so you can group rows, split columns, filter and sort. Then you keep the views you’ll want again as tables of their own. Every figure is counted on the full dataset, not a sample, so a grid over millions of orders shows the same numbers a query would.

What you’ll need

A plan that includes sheets — see pricing. Without it, a dataset has no Open as a sheet button, and Sheet under an answer is greyed: clicking it says “Sheets are not included in your plan. Ask your administrator about a plan that includes it.”
A dataset in Data home: an upload, a connected source, a table a conversation made, or a table in your own data lake that you’ve linked (see Read your own data lake). See Data home.

Open a sheet and pivot it

From a conversation: under the latest answer with a chart, once it has finished, click Share this as → Sheet. The chart’s data opens as a sheet beside the conversation. Sheet shows only when the chart has a dataset behind it. Or type it: “open this as a sheet”.

From your data:

In Data, open a dataset and click Open as a sheet. If you already have a sheet on it, that sheet opens; otherwise a new one does. The grid shows every row of the dataset, numbered, and the address in your browser is the sheet’s own, so you can come back to it. The sheet’s tables are tabs along the bottom, like a workbook’s.
The pivot panel is open on the right (click Pivot to show or hide it). Drag columns into the row groups, the column split, the filters and the sort. For example: group by month, split by channel, and show customer_id counted uniquely. You’ll see unique buyers for each month and channel.
Read the totals. Each total in the grid is counted again from the rows, not added up from the cells above it. A customer who bought by email and again through an ad is one buyer in the total, even though they’re in both channels’ cells.

Heads up: a unique count, a rate or a share is never added up, so the grid won’t offer to sum one. It offers a minimum, maximum or average instead, and says why if you ask for a sum.

Chart it

Click Chart and the view you set up is drawn as bars: group by day and show daily_events, and you see a bar a day. Pick another kind at the top of the pivot panel: lines, areas, scatter, a heatmap, a treemap or a sunburst. Click Grid to go back to the rows. A chart is a way of looking: it isn’t kept when you save a table, which keeps the rows. Map charts draw their points without a background map.

Add a column

You can add a column in two ways:

Write a formula in the pivot panel (New column). Formulas use the table’s own columns, numbers and text, + - * /, comparisons, and/or/not, and a short list of functions: round, abs, coalesce, nullif, lower, upper, length, if, year, month, day, and date_trunc for a day, week or month. For example, date_trunc('month', ordered_at) gives each order’s month as its first day.
Ask for it in words. Click Add a column, describe it (“revenue per order”, “orders over $100 as yes/no”) and click Write it. BumbleB Crunch™ shows you the formula and a plain sentence saying what it computes. Nothing is added until you click Add column. A column that divides is marked “Never summed: it divides.” A total (“total revenue per channel”) isn’t a column, and BumbleB Crunch says so instead of guessing.

Keep a view as a table

Set up the view you want, then click More → Save as table, name it (“Buyers by channel and month”), and click Save.
It appears as a tab next to your dataset, marked ·saved. Its counts are named for what they hold (unique_customer_id, sum_revenue), and its rows are in order of its groups.
A saved table remembers how it was made. Click More → Refresh and every saved table is made again from the data as it is now. If you uploaded a new version of the file, the saved tables follow it.

Type in a small table

Targets, a mapping of campaign codes to names, a few notes: some numbers aren’t in any dataset. Click More → New typed table, name it, give each column a name and a type (text, number, whole number, date or yes/no), and type the rows. Click Save table and it appears as a tab marked ·typed. A cell that doesn’t fit its column’s type (“seventy” in a number column) is refused, naming the row and the column. Click Edit rows to change it later; each save is a new version.

Correct a value

Sometimes a row in your data is simply wrong: an order recorded at $9,999 instead of $99.99, a customer tagged with the wrong channel. You can correct the value in your sheet without touching the data itself.

Open the dataset’s tab and use a plain view: no row groups and no column split, with a column that names each row once (like order_id) showing.
Click the cell you want to correct. A line appears above the grid: “Correct revenue for order_id o000123: 9999 →”. Type the right value and click Correct. A value that doesn’t fit the column’s type (“seventy” in a number column) is refused.

The correction is kept beside your data, never written into it. From then on, the grid counts the corrected value, and so does every table you save afterwards, every refresh of them, and BumbleB Crunch when you ask about the sheet in a conversation (it tells you its figures include your corrections). Above the grid you’ll see “Includes 1 corrected value”. Click Corrections to see each one, who made it and when, and Undo it.

When your data changes, each correction is checked again. If its row is gone, or the value in your data is no longer the one you corrected, it isn’t applied, and Corrections lists it as “no longer matches”, never dropped without telling you.

Open a table from your data lake

If you’ve linked a table from your own data lake, it’s listed in Data home with your other datasets, and Open as a sheet works the same way. Its rows stay in your lake: the grid reads them there, nothing is copied in first.

One version while you work. Your lake’s table can change while you have the sheet open. The grid keeps reading the version it opened with, so scrolling and pivoting never mix old rows with new ones. Click Refresh to move to the latest version.
Saved tables are made from the version the grid was reading when you saved them.
Publishing copies the table at that version for your readers, so they never reach your lake or its credentials. A very large table (see Current limits) isn’t published; the sheet says “Not published:” with the table and its row count.
No corrections here. A linked table’s values are corrected in your lake, so the sheet doesn’t offer to correct them.
Pin a saved table, not the linked one. A linked table can’t be pinned to a dashboard as it is: BumbleB Crunch says “A linked warehouse table is read in place; pin a table saved from it instead, which refreshes from the warehouse.” Save a table from it (More → Save as table) and pin that.

Work it out by hand in a calc

Sometimes you want a spreadsheet: a quick =SUM over a few rows, a ratio you’re testing before you ask for it properly. A calc takes the view you’re looking at into a spreadsheet with cell formulas, right beside the sheet.

Set up the view, then click Calc. Its rows open as a spreadsheet: one header row, then a row per grid row. A view with more rows than a calc takes (see Current limits) is turned down; filter it first.
Type formulas in any cell, like =SUM(B3:B6), with a label in the cell to its left.
Click Save calc. It’s kept on the sheet as a button next to its tables, with a dashed border; click it to open it again. A sheet keeps a few calcs.

The calc says “Worked by hand — not counted from the data” at its top, and that’s how BumbleB Crunch treats it. When you ask about the sheet in a conversation, BumbleB Crunch counts its answer from the sheet’s tables, never from your calc. If one of your hand figures says something different, BumbleB Crunch says so and gives both, naming yours as worked by hand in that calc. The calc lists the hand figures BumbleB Crunch is told about, under “Hand figures chat is told of (as worked by hand)”.

A calc is only for you and your editors. It’s never published, never saved as a dataset and never pinned to a dashboard, and readers of the sheet don’t see it.

Click Share and choose who can read it: people by email, or everyone in your organization. The dialog lists every table and its columns, because readers see whole tables.
Click Publish and share. Each table of the sheet, as it stands now, is copied for your readers, and each person you added gets an email and a notice in BumbleB Crunch with the sheet’s link. Someone without a BumbleB account isn’t emailed; the dialog says so, and you can send them the link once they join.

The dialog also shows the sheet’s link (…/data/p/…), the same kind of link a story or dashboard has: copy it and send it yourself. Whoever you shared with opens it signed in; anyone else is told it isn’t available.

Your readers open the sheet and see “Published … · read-only”. They can pivot, filter, sort and chart every table, and add a column for their own view, but they can’t save, type or correct tables, or refresh. They see the sheet as you published it: what you change afterwards stays yours until you click Publish again. Readers need a plan that includes sheets; their pivots are counted on their own plan.

For someone outside BumbleB Crunch, make a link anyone can open without signing in. In Share, under Anyone with the link, pick when it expires and click Make a link. Copy it then: it’s shown only once.

The link shows a copy of the published sheet, frozen when you made it: the first rows of each table, up to a set number of tables (see Current limits). It opens as the same grid, with the pivot panel and charts, so readers can pivot, filter, sort and chart it in their browser; nothing they do is kept. It doesn’t refresh on its own. To show a newer version, publish again and make a new link; Turn off ends it. If your organization has an owner or admin, they must first allow sharing the sheet outside the organization.

Take the rows away

You can export the grid’s rows to a file, or copy them, from the grid itself, up to the export limit. Past that, it says “That is more than 10,000 rows at once: filter the view, or save it as a table, before exporting it.” Text that a spreadsheet would run as a formula (a cell starting with =, say) is exported with a leading ', so it opens as plain text rather than running.

Use it elsewhere

On a dashboard: open a table’s tab and click More → Pin to a dashboard. Pick a dashboard or click New dashboard. The table becomes a tile, and the dashboard’s Refresh makes it again from the latest data. See Publish a dashboard.
In a conversation: type @ and the sheet’s name, then ask about one of its tables by name. For example: “@Orders, using the Buyers by channel and month table in it, which channel grew the most from January to August?” BumbleB Crunch answers from that table’s own rows and names it. Ask about a table the sheet doesn’t have, and BumbleB Crunch says it isn’t there rather than give you a number for it.

How to tell it worked

Your sheet keeps its tabs: the dataset, each ·saved table and each ·typed table. Reload the page and they’re all still there. You’ll also find the sheet under Published → Made by me, marked Sheet. After Refresh, a saved table’s counts match the grid’s on the same view.

Good to know

Who can do what. Anyone who can open a sheet can browse it and ask for a column. Saving a table, typing one, editing rows, calcs and pinning a table to a dashboard take an editor. Only the owner can refresh, correct values, share and publish.
Limits. How many corrections a table takes, how big one view, a formula and a typed table can be are in Current limits. For a bigger typed table, upload a file.
Correcting a cell again keeps the value it first had, so Undo always takes it back to your data.
Refresh counts toward the daily refresh limit, the same as a story or dashboard.
One sheet per dataset, unless you ask. Open as a sheet reopens the sheet you already have on that dataset. To start a separate one, click New sheet next to it.
Not yet: a reader can’t ask about a shared sheet in a conversation, and can’t save their own pivots.

Publish a dashboard — tiles, filters and sharing.
Data home — your datasets and where they came from.
Upload an Excel, CSV or Parquet file — bring in the data to pivot.