Dashboards

A dashboard is a page of charts, tables and headline numbers built on your datasets. Jason builds it for you from a description. Each card runs a saved query against the dataset’s active version, so the dashboard stays current as pipelines publish new versions and nobody has to edit it.

This page covers finding, creating and reading dashboards. To rearrange one, change a card, add sections or set up filters, see Editing dashboards.

The Analytics page

Analytics in the sidebar opens All Analytics, the list of every dashboard in the workspace.

Each card shows:

• Miniature charts. Up to three of the dashboard’s charts, drawn small. They load after the list, so each card fills in on its own.
• Name and subtitle. The subtitle lists the charts by name (“Products tracked · Price changes · 30 days · +3 more”). A dashboard with nothing on it reads “No charts yet”.
• Where the data comes from. The icons of the pipelines whose datasets the dashboard reads. Hover an icon for the pipeline’s name.
• Failing charts. “2 of 9 failing” in amber when some queries fail against the current data. Open the dashboard and ask Jason to fix them.
• Build state. A blue card with a spinner and Building while Jason is still making the dashboard. An amber card reading Build didn’t finish means the build ended before anything was drawn.
• Timestamp. When the data behind the dashboard last changed, not when its layout did.
• Menu. Pin to sidebar (or Unpin from sidebar) and Delete. See The sidebar.

The controls above the cards work the same way as on the Pipelines page.

Option	Values	Default	Notes
Search	Free text (placeholder “Search dashboards…”)	Empty	Matches dashboard names.
Sort	Last Updated, Name	Last Updated
View options	Large grid, Small grid, List	Large grid	Remembered in your browser.
New folder	A folder name		Drag cards into a folder, or select several and use Move to folder.

Delete is in each card’s menu. You can also select several cards and press Delete in the selection bar. Both ask first: “The dashboard will be permanently deleted.” Deleting a dashboard does not touch the datasets it reads.

With no dashboards yet, the page reads “No analytics yet” with a New Dashboard button.

Creating a dashboard

Press New Dashboard at the top of the list, or the Add dashboard tile at the end of it.

1. Describe what you want to see. The box reads “What do you want to see? Type @ to pick the data it should use.” Say what the dashboard is for, the numbers you care about, and how you want them broken down.
2. Link the data. Type @ for a list of pipelines and ready datasets, or press Link to tick them in a picker. Each linked item shows as a chip under the box; its cross removes it. If you link nothing, Jason picks the datasets from your description.
3. Attach files if they help. Attach, or drop files on the box: a mock-up, a report you want to reproduce, a list of metric definitions. Jason reads them.
4. Or start from a suggestion. The Suggestions pills are written from the columns in your datasets and the dashboards you already have. Clicking one fills the box. Edit it before sending.
5. Press Start building (or Cmd/Ctrl+Enter). A new conversation with Jason opens full screen with your request in it.

Option	Values	Default	Notes
Request	Free text	Empty	Required. Start building stays disabled until you type something.
Link	Any pipelines and ready datasets in the workspace	None	Tells Jason which data to use.
Attach	Any files (CSV, PDF, images and so on)	None	For Jason to read. They are not added to a dataset.

A dashboard for the buying team: products tracked, price changes over the last 30 days, coverage by retailer, and a table of today’s biggest price moves with a link to each product. Use @Live product offers and @Price observations.

Jason reads the columns and a sample of the rows, writes and tests a query for each card, then draws the page. In the conversation you can see its steps, such as Create dashboard, Save dashboard query, Draw dashboard and Check dashboard. The new dashboard appears in All Analytics straight away as a blue Building card, and Jason links it in the conversation when it is done.

You can also ask for a dashboard in any conversation with Jason, or from a pipeline’s Analytics section (see below).

While a dashboard is being built

Opening a dashboard that has nothing to draw yet shows one of three messages:

• Jason is building this dashboard. It appears as soon as it is ready. The page checks back on its own; Open the conversation takes you to the build.
• This dashboard wasn’t finished. Jason’s build ended before it drew anything. Open the conversation to see why, or ask again there.
• Nothing on this dashboard yet. Nobody has asked for it to be built. Ask Jason from the sidebar.

The dashboard page

Dashboards always use a dark surface, whatever the app’s theme. From top to bottom:

• Breadcrumb. Analytics, then the dashboard’s name. The name is a dropdown of every dashboard in the workspace, so you can switch without going back to the list. When you reached this dashboard by following a link from another one, the dashboards you came through appear in the breadcrumb too, and each one takes you back.
• Title and “Updated”. The dashboard’s name, and when its data last changed if the dashboard shows it.
• Entity chips. On an entity page only; see Entity pages.
• Latest data. Which version of the data the page is reading; see Looking at older data.
• Filter row. One chip per filter, when the dashboard has filters; see Filters.
• Edit dashboard. Turns on edit mode. See Editing dashboards.
• Sections and cards. Cards sit on a twelve-column grid, grouped under section headings.

The page loads as a whole: a placeholder shows until the first data has arrived, then every card appears at once. When Jason changes the dashboard in a conversation, the page reloads with the change when his turn ends.

Reading a card

Every card has a title, and most have an ⓘ button at the right of the title.

ⓘ opens a dialog named after the card:

• Data. The exact rows the card’s query returns right now, in a searchable, sortable table of ten rows per page.
• SQL. The query itself.
• Definitions and caveats. Lines above the rows saying what a measure means, such as what counts as a price change, and, in amber, anything to keep in mind when reading it.

Use it when a number surprises you. A card with no recorded query says “No query is recorded for this widget.”

Asking Jason about a value

Click a bar, a point or a slice in a chart. A small box opens where you clicked, labelled “Ask Jason about this value”, with the value’s fields underneath. Type a question and press Enter. The question goes to Jason in the sidebar along with the card and its queries, so Jason knows which number you mean.

When the chart links to another dashboard, the box also offers Open {name} →, which opens that dashboard’s page for the value you clicked.

Why did Best Buy’s coverage drop on the 14th?

Clicking the empty parts of a card, its axes or its text does nothing. Escape or a click elsewhere closes the box.

Tables

Tables show 25 rows per page with Previous page and Next page controls and a count (“26–50 of 312”). A table with more rows than fit on one page has a Search box beside the card’s title; it matches any column. Click a column heading to sort ascending, again for descending, and a third time to clear the sort. Empty values sort last either way.

Click a row to open it. What happens depends on the table:

• A linked table opens the linked dashboard’s page for that row. The last column then has a Show this row button for looking at the row itself.
• Any other table opens the row in a dialog with every field in full. Ask Jason about this row closes the dialog and puts the row in the Jason sidebar with the start of a question for you to finish.

Some tables are built to open a row’s source. For those, the row opens in the standard row dialog on its Activity tab, showing how Jsonify found that row. See Row detail and provenance.

Filters

When a dashboard has filters, they sit in a row under the title, for example Retailer: All and Category: All. A filter narrows every card that reads a dataset with that column, so one choice changes the whole page.

• Pick values. Click the chip for a list of the column’s values with a row count beside each. Tick as many as you like. A search box appears when there are more than eight values, and “Showing the most common values” means the list is cut short. Clear unticks everything.
• Min – max. Numeric filters can instead take a range. Type a Min, a Max or both and press Apply. The placeholders show the column’s current lowest and highest values.
• Clear filters. Appears once any filter is set, and resets them all.

The chip shows what it is set to (“Amazon”, “Amazon +2”, “≥ 100”). The other filters narrow each filter’s list, so with Retailer: Amazon set, Category lists only the categories Amazon has. Filter values are kept in the address, so a link you send opens with the same filters. Anyone who can open the dashboard can set filters; adding, changing or removing them is done in edit mode.

Looking at older data

The Latest data button in the header says which version of the data the page is reading. Click it to open Data as of.

For each dataset the dashboard reads:

Option	Values	Default	Notes
Dataset version	Latest, or one of the dataset’s 50 most recent versions	Latest	Latest follows the dataset as new versions are published (“Currently v7”). An older version lists its date, time and row count.

Press Apply. Every card re-runs against the versions you chose, and the button reads “As of Sep 14”, the date of the oldest version you picked. Back to latest in the dialog puts every dataset back on Latest. The choice lasts until you leave the page; it is not saved and nobody else sees it.

This is how to see what a dashboard showed last week, or compare before and after a change. See Versions.

Entity pages and links

Some dashboards are about one thing at a time: one product, one retailer, one venue. Jason builds these as entity pages and links them from overview dashboards.

• Following a link. Clicking a linked table row, or Open {name} → on a chart value, opens the entity page with its value in the address, for example /dashboards/marketdash03?product_id=MR-100001. The breadcrumb names the dashboard and the entity (“Product detail: AirPods Pro”).
• Picking a value. Opening an entity page without a value shows Choose a product with a searchable list. Pick one to load the page.
• Switching. On an entity page, the entity chip in the header (Product: AirPods Pro ▾) opens the same list, so you can move to another product without going back.

Back and Forward in the browser step through the entities you have viewed. Filters you set carry over when you switch entity.

Add a product page to this dashboard: price history per retailer, current promotions and stock, and link each row of the price-moves table to it.

Dashboards on a pipeline page

A pipeline’s Analytics section shows the dashboards that read the pipeline’s datasets, embedded in the page. It sits between Latest outputs and Sources.

• One dashboard fills the section.
• Several get a list on the left; pick one to show it. The list shows up to 50, with View all linking to All Analytics.
• None shows “No dashboard yet” and Add dashboard. Pressing it opens Jason on a new conversation asking for dashboards over the pipeline’s main dataset. When the dashboard is ready it loads here.

Filters, Data as of, ⓘ and edit mode all work inside the section. Following a link to an entity page opens that dashboard on its own page. See Pipeline pages.

Sharing a dashboard

When the workspace has link sharing turned on, dashboards are included along with everything else in the workspace. Visitors can read cards, set filters, open ⓘ and follow links. They do not get Edit dashboard. See Sharing.

Home page analytics

The Analytics section of the workspace home page is not a dashboard. It charts pipeline runs (rows processed, pipeline runs, failed runs, success rate) for one pipeline or all of them over the last 24 hours, 7, 30 or 90 days. Dashboards are about your data; the home page charts are about your pipelines.

What’s next