Visualizing data with AI

Important

This feature is in beta and subject to change. Reach out to SquaredUp Support with any issues.

Have you ever found yourself sitting on a mountain of data, struggling to weave a narrative and unsure where to even begin? You're not alone. Most people building a dashboard for the first time spend more time staring at the chart settings than they do making sense of their actual data.

The good news is, thanks to SmartViz, dashboarder's block is a thing of the past.

How it works

When you turn on SmartViz, it looks at your data and automatically builds what it thinks is the best chart for it, picking the chart type, wiring up the axes, and choosing sensible groupings without you touching a single setting.

From there, you refine it by describing what you want changed, and SmartViz updates the chart to match, so you go from a rough first guess to the chart you actually meant with a few short prompts.

SquaredUp has two AI assistants in the tile editor. SmartAssist writes the query that fetches your data, and SmartViz turns that data into a chart. This article covers SmartViz. Both are in Beta.

Before you start

  • AI features must be enabled for your tenant. An admin can turn these on in Settings > AI.
  • For help writing the query itself, see SmartAssist. SmartViz picks up once your data is on screen.

How to use SmartViz

There are three ways to get a chart out of SmartViz, depending on where you’re starting from: a blank tile, an existing SmartViz chart, or a classic visualization you want to bring over.

Automatically generate a chart

Here a tile of alert counts by service goes straight from raw table to finished chart.

Loom video thumbnail

To add a new tile using SmartViz:

  1. Add a new data tile to your dashboard.
  2. Configure your data stream and shape your data in the tile editor.
  3. Click Use SmartViz when prompted to select a visualization chart type. The best chart type and settings are configured based on the current shape of the data.
  4. Review the output, accept it as-is, or continue shaping it with a prompt.
  5. Click Save when you’re happy with the result.

Edit a chart through prompts

Here a few short prompts take one bar chart through several shapes, ending with a color rule that flags any service above 50 alerts.

Loom video thumbnail

To edit an existing SmartViz chart:

  1. Open a tile with an existing SmartViz chart.
  2. Type what you want into the message box, for example “change this to a line graph”, “color the bars by team”, or “hide the legend”.
    Note

    Prompts shape the chart itself. To change which rows reach it, use the Filter | Group | Sort step in the tile editor before prompting.

  3. Review the result. SmartViz applies it to your existing chart rather than starting over, and logs it as a new step in the conversation.
  4. Hover an earlier step to preview it, then click its reset icon to roll back if a change doesn’t work out, or use the Data Mapping section below the conversation to adjust something by hand instead.
  5. Click Save when you’re happy with the result.

Convert a classic chart to SmartViz

Here a classic stacked bar of deployments by team is rebuilt as a SmartViz chart, keeping its axes, grouping, and colors.

Loom video thumbnail

To convert a classic visualization:

  1. Open a tile that already has a classic visualization.
  2. Click Use SmartViz. It reads your existing configuration, including the axes, grouping, and colors, and rebuilds it as a SmartViz chart.
  3. Keep prompting or edit the mapping by hand, it behaves like any other SmartViz chart from here.
  4. Click Save when you’re happy with the result.

Tips to get a good result

  • Load real data first: SmartViz shapes the chart around the columns your data stream returns, so a representative dataset gets you a far better first suggestion than an empty one.
  • SmartViz reads your columns, not your rows: “make anything over 50 red” works because you supplied the threshold. “Show me the top 5” doesn’t.
  • Timeframe belongs to the tile, not the prompt: asking for “the last 7 days” changes nothing. Set the timeframe on the tile, and SmartViz charts whatever comes back.
  • Describe the result, not the setting: “sorted bar, highest first”, “split by team”, “breakdown by environment” and “show the trend” all land, because they name an outcome rather than a control to toggle.
  • Change one thing at a time: each prompt rebuilds the whole chart with only your requested change applied, so short follow-ups stack up rather than starting over. Bundle several changes into one prompt and some get dropped.
  • Experiment freely: every step is reversible, so try the odd idea and hover back through the conversation to compare.

Prompt examples

Every prompt below was run against SmartViz and produced the chart listed beside it. Start from the one closest to what you want to see, then swap in your own column names.

Chart prompts

These prompts set the chart type. SmartViz reads the columns in your data stream and decides for itself which one belongs on each axis.

Prompt
Chart you get
Data it needs
“Show this over time”
A time column plus one or more numbers
“Plot latency against error rate”
Two numeric columns
“Show the distribution of response times”
One numeric column, 20+ rows
“Change this to a health history chart”
A time column plus a status column
“Compare deployments by team”
A category column plus a number
“Show actual against target”
A value plus a target value
“Compare these scores as a spider chart”
Three or more numeric columns
“Show alerts by hour and day”
Two category columns plus a number
“Show the quartiles”
Pre-computed low, Q1, median, Q3, high
“Percentage breakdown by status”
A category column plus a number, 2 to 20 categories
“Show where the space is going, by size”
A label plus a numeric size, 10 to 200 items
“Show the drop-off through each stage”
Stage names plus a count, 3 to 15 rows
“Show each service as a percentage ring”
One numeric value per row
“Show the flow from source to destination”
Source, target, and a numeric value per row
“Show this as a big number”
One or a handful of values
“Show utilization as a dial”
A single numeric value with a natural maximum
“Show health as status blocks”
A label column plus a status column
“Table with status indicators”
Mixed-type columns

Refinement prompts

These prompts adjust a chart you already have. Each one changes the single thing you asked for and leaves the rest of the chart as it was.

Prompt
What you get
Chart types
“Split by team”
One series per team, with a legend
Line, bar, scatter
“Sorted bar, highest first”
Horizontal bars, so long category names stay readable
Bar
“Stack the bars, broken down by team”
Series stacked into one bar per category
Bar
“Show the trend”
A dashed regression line over each series
Line, bar, scatter
“Add a mark line at 15”
A fixed reference line across the chart
Line, bar, scatter
“Make anything over 50 red”
A color rule that recolors matching bars or points
Line, bar, scatter
“Show the values”
Numeric labels drawn onto each bar, point, slice, or cell
Most chart types
“Make the text larger”
Bigger labels, for a chart being read from across the room
Most chart types
“Color production red and staging blue”
Fixed colors pinned to named series, slices, or blocks
Line, bar, pie, treemap, funnel, status grid, health history
“Add a legend at the bottom”
A legend in the position you name, hidden again with “no legend”
Line, bar, scatter, funnel, radar, pie
“Sort highest first”
Segments or blocks reordered by value
Funnel, status grid
“Make it a donut”
A hole through the middle, with the total in the center
Pie
“Show five per row”
A set number of blocks or rings on each row
Status grid, ring grid
“Use 20 bins”
Narrower buckets, for finer detail in the distribution
Histogram
“Color the rows by status”
A tinted row background driven by a column value
Rich table
“Use a blue color scale”
A sequential ramp in place of the default palette
Heatmap

Was this article helpful?

Have more questions or facing an issue?
Submit a ticket