Data Studio (formerly Looker Studio) chart types — and how to choose the right one

Data Studio has twenty documented chart types, plus a newer Histogram and community visualisations, and the shape of your data picks the right one: how many dimensions and metrics you have, whether one is a date, how many values the dimension has, and whether the parts add up to a real whole. Most good reports use five of them. The rest earn their place on specific questions, and every one has a setting that makes it honest or misleading.

Below: a chooser table, then each of the twenty types with what it’s for, short setup steps and when it misleads, plus the “Others” row that quietly changes your numbers. The screenshots are mine, on Google’s GA4 demo account (Merchandise Store), and every setting was checked against Google’s chart references on 19 October 2026 or in my own editor.

First, the naming

Google renamed the tool Looker Studio, then back to Data Studio. Same product, same charts. Looker, without “Studio”, is Google’s separate enterprise BI platform with its own visualisations; if a page mentions LookML, it isn’t about this tool.

The chart chooser: find your data’s row

Match your fields to a row and start with the first choice; the runner-up is for when the page is already busy. Chart names jump to their setup steps below.

If you remember two rows, make it these. A pie needs parts of a real whole; the slice count comes second. And “one metric spread over many rows” is the one most reports skip: average order value shows up everywhere, the spread behind it almost never.

Start with the shape of your data, not the chart

“Decide what you want to show” is the usual advice, and it skips a step. Your field list settles most of it first:

• How many dimensions? None, one or two. (A dimension is a field you group by: channel, page, month.)
• How many metrics? One, two on different scales, or many. (A metric is a number you add up: sessions, revenue.)
• Is one dimension a date? Then you’re probably showing change.
• How many values? A handful of channels and a few hundred landing pages won’t fit the same chart.
• Do the parts add up to a real whole? Device categories do. Your top ten pages don’t.

Google’s chart references are written in the same terms. Chapter 9 of my ebook, Data Storytelling in Data Studio, starts from these five questions and the perception research behind them, and the rest of Part IV takes each chart family in turn.

All twenty chart types at a glance

Google’s list: scorecard, table, pivot table, time series, bar and column, pie, combo, geo, Google Maps, area, scatter, bullet, gauge, tree map, Sankey, waterfall, boxplot, candlestick, timeline and funnel (Types of charts in Data Studio). Several have variants in the same menu: the donut is a pie, the line chart is a combo, the bubble chart is a scatter.

Add a chart, switch its type, find its settings

Select any chart and the properties panel has two tabs. Setup holds the data source, dimensions, metrics, sort, date range and row limit: everything that changes the numbers. Style holds colours, labels, axes and variants: how the same numbers look. My test: if changing it would alter an exported figure, look on Setup. The editor’s layout is in the Data Studio interface, explained.

1. In edit mode, click Add a chart, pick a type and draw it on the canvas.
2. To change a chart’s type, select it and open Chart types at the top of the properties panel.
3. Set the fields on Setup first, then Style.
4. Title it in Style › Chart title. Charts have a title and no subtitle field, so anything longer goes in a text box.

The Chart types list also has a Histogram group. Google hasn’t documented it yet, so the histogram recipe here is the one you build from a column chart, choosing the bands yourself.

Numbers and tables

Scorecard: one number, against something

For: a headline KPI that needs a verdict. Takes: one metric, no dimension.

1. Add a scorecard and set its metric.
2. Open Setup › Comparison type. Period (the default; “Full stop” in UK English) compares with the previous period. Value compares with a number you type into Target value; Metric reads the target from a field.
3. For a target, turn on Show as progress, on Setup just under Target value.
4. On cost metrics, swap the change colours in Style › Comparison fields, so a falling CPA shows green.

Misleads when: it’s a bare number, a row mixes comparison bases, or a cost’s rise shows green. A tile can’t carry a sparkline and progress at once. Every setting, plus two comparisons stacked on one tile, is in my scorecards guide.

Table: which rows, and by how much

For: long ranked lists, several metrics, anything someone will export. Takes: up to 10 dimensions and 20 metrics on GA4 or Google Ads; up to 100 of each on Sheets or BigQuery (Table reference).

1. Add a table with the dimension (Landing page) and metrics (Clicks, plus a change or a rate).
2. In Setup › Sort, sort by the metric that matters, descending (up to 10 sort fields).
3. In Style, set the main metric’s column type to Bar; Show target adds a target line.
4. Add a Conditional formatting rule: change less than 0, red fill.
5. In Setup › Number of rows, switch Pagination to Top N for a fixed list; only then does Group others appear.

Misleads when: it has more columns than anyone reads, an alphabetical sort, a rate with no volume beside it (a page with a handful of impressions tops a CTR list), or a heatmap on totals, which simply shades the largest rows darkest.

Pivot table: which combination stands out

For: two dimensions at once, such as channel by month. Takes: row dimensions, up to 2 column dimensions, up to 20 metrics (Pivot table reference).

1. Add a pivot table with the row dimension, the column dimension (Month) and the metric.
2. With several row dimensions, turn on Expand-collapse, ordered general to specific.
3. In Style, set the metric to Heatmap with one colour, or add Conditional formatting with Colour Scale (Gradient).

Misleads when: the shading sits on totals across rows of different sizes, so the biggest category is dark everywhere. Shade a rate or a share. Pivots cap at 500,000 cells and five per page, don’t paginate and don’t take metric filters.

Change over time

Time series: what changed, and when

For: sessions, clicks, spend or revenue over weeks and months. Takes: a date dimension and well over five metrics (Google’s line chart reference gives up to 20, and a time series took 13 when I checked), or one metric split by a breakdown.

1. Add a time series with a monthly date (Year Month) for anything read monthly, and the metric.
2. Set a Comparison date range of Previous year; last year draws as a second line.
3. For a channel split, add a Breakdown dimension and set the Number of series.
4. In Style, grey every series, then give the one that matters your accent colour.
5. Set Missing data to Line breaks. The default, Line to zero, draws a crash on any day with no rows.

Google’s time series page still says five; the screenshot is my October 2026 check. Trend lines, Smooth and Cumulative live in Style (Time series reference). Mark a target with a reference line, a launch day with an X-axis date line (a time series takes one, and it prints in the PDF), or a year of events with annotations.

Misleads when: there are more than four lines, daily noise sits in a monthly report, or a partial last period looks like a collapse. Which date setting wins is in my date range controls guide.

Combo and line chart: did one measure follow another

For: two related measures, usually a volume as bars and a rate as a line. Takes: any dimension type (a time series needs a date; a line chart doesn’t) and up to 20 metrics with one dimension (Line chart and combo chart reference).

1. Add a combo chart with the date and both metrics.
2. In Style, set each series’ Series type (Line or Bars) and, on combo charts only, its Axis.
3. Title both axes and fix their minimum and maximum.
4. Better: a column chart of the first metric with a copy pasted directly below at the same width, changed to a line of the second. One timeline, no second axis.

Misleads when: the lines cross. Where they meet depends on where each axis starts, so the crossing means nothing.

Area chart: is the total growing, and from where

For: total and contribution over time (stacked), or the shifting mix (100% stacked). Takes: a date, a breakdown dimension and one metric (Area chart reference).

1. Pick Stacked area or 100% stacked area from the chart menu.
2. Use a breakdown with four values or fewer; Number of series and “Others” control the bands.
3. Use Breakdown dimension sort to put the band that matters at the bottom.

Misleads when: readers judge a middle band, which has no straight edge, or a 100% chart hides growth. Stacked area takes a constant-value or parameter reference line; 100% stacked takes none.

Timeline: what ran at the same time

For: campaign flights, launches, migrations. Takes: no metric: a text Row label, a Start time, and an End time or a Duration, not both (Timeline chart reference).

1. In a Sheet, one row per campaign: Team, Campaign, Start, End. Connect it.
2. Add a timeline: Row label Team, Start time Start, End time End.
3. Set Campaign as the Bar label and turn on Show bar label in Style.

Misleads when: it’s read as a time series and someone hunts for a value on the vertical axis. Title it as a calendar.

Categories and parts

Bar and column chart: which is biggest

For: ranking and comparing categories. Takes: one dimension with up to 20 metrics, or a dimension plus a breakdown and one metric (Bar chart and column chart reference).

1. Add a bar chart with the dimension (Item category) and the metric (Revenue). Bars for long labels, columns for short ones or dates.
2. In Setup › Sort, choose the metric, descending.
3. In Setup › Display, set Top N and the Number of bars.
4. In Style, grey every bar and colour the one that matters.
5. For a breakdown, add a Breakdown dimension (grouped), then Style › Stacked bars, and 100% stacking for shares. Secondary sort then orders the series.

Misleads when: it’s alphabetical; the axis starts above zero, so a small gap looks huge; a stack runs past three segments; an “Others” bar is read as a category. Stacked bars take a constant-value or parameter reference line; 100% stacked take none (reference lines and bands).

Pie and donut: roughly how a real whole splits

For: “about half, or about a quarter?”, never “47.3%?”. Takes: one dimension, one metric, 1 to 10 slices (Pie chart reference).

1. Add a pie (or donut) with a dimension whose values make a real whole, such as Device category.
2. In Setup › Display, set the number of slices. Keep the “Others” slice so the pie still adds up.

Misleads when: the whole is false. A pie of your top landing pages shows each page’s share of those pages, not of all your clicks, so the biggest one looks far larger than it is. Also past four slices. A sorted bar is the better default.

Tree map: what each category is made of

For: a hierarchy with lots of parts that someone will browse, not rank. Takes: dimensions parent first, one metric for size (Treemap reference).

1. Add a tree map with the dimensions in hierarchy order (category, subcategory) and one metric.
2. Set Levels to show to 2, or turn on Drill down.
3. Colour follows the metric by default; change Color by if it should mean the category.

Misleads when: the reader needs to rank. Area is hard to compare. Put a sorted bar of the top level beside it.

Row limits and the “Others” row, which quietly change your numbers

This is the setting most likely to make a chart wrong rather than ugly. When a dimension has more values than the chart’s display limit, Group the rest as “Others” decides what happens to the rest. On bars, pies, tree maps and a time series’ breakdown it’s on by default, rolling everything past the limit into one bar, slice or series labelled Others.

• An “Others” row is not a category. It’s the tail of your data wearing a label, and on a client report somebody will read it as a channel. Say what it holds (“rest of site”).
• Turning it off doesn’t show you more data. It shows you less. The values past the limit simply disappear, so the chart no longer adds up to your total; Google’s docs warn it may not display all the data.
• It’s incompatible with comparison metrics. If you want a comparison metrics calculation such as Percent of total on that chart, you have to disable Others grouping, which brings the point above into play. The limit is on those calculations, not on a previous-period comparison date range.
• You can’t cross-filter by “Others” (Chart cross-filtering), and an average reference line is worked out from the rows the chart displays, so the limit moves it too.

The habit: when a chart doesn’t tie back to a scorecard on the same page, check the row limit before you suspect the data source. It’s the row limit far more often than the connector.

Relationships and spread

Scatter and bubble: which rows break the pattern

For: two measures per row, and the exceptions. Takes: a dimension and two metrics; a bubble chart adds one for size. At most 1,000 bubbles (Scatter chart reference).

1. Add a scatter with Query as the dimension, Metric X Average position and Metric Y Site CTR, from Search Console.
2. Filter out brand queries, so they don’t squash everything into a corner.
3. In Style, add a trend line and fix both axes’ min and max, so the chart doesn’t rescale monthly.
4. For bubbles, add Bubble size and lower the bubble opacity.

Misleads when: correlation is read as cause, or a thousand dots arrive with no sentence on what to look at. Send a stakeholder the outliers as a table.

Boxplot: what hides behind the average

For: the spread behind an average that drives a decision. Takes: a dimension and exactly five metrics in order: minimum, 25th percentile, median, 75th percentile, maximum. You supply all five values; the chart only plots them (Boxplot chart reference).

1. Use a source with one row per order, such as a Sheets or BigQuery export. Percentiles can’t come from daily totals.
2. Create the five calculated fields (Google suggests MIN, MAX, MEDIAN and PERCENTILE). I use PERCENTILE(Order value, 5) and PERCENTILE(Order value, 95) for the ends, so one odd order doesn’t stretch the whiskers.
3. Add a boxplot with Channel as the dimension and the five fields in order, plus a key: “box = middle half of orders”.

Misleads when: the reader has never seen one, or the sample size is hidden. Too much for the audience? Put the median in a scorecard, or build the histogram below.

Candlestick: how far each period swung

For: open, close and range per period, such as weekly CPC. Takes: a dimension and four metrics in Google’s order: high, opening, closing, minimum (Candlestick chart reference).

1. Add a candlestick with Week as the dimension and the four metrics in that order.
2. Add a key: filled when the close is above the open, hollow when below.

Misleads when: readers lack the finance conventions. A line of the close with pale high and low lines usually reads faster.

Targets, flows and bridges

Bullet chart: are we on track

For: one number against a target, with ranges for “good”. Takes: one metric, a target, up to three ranges (Bullet chart reference).

1. Add a bullet chart and set its metric.
2. Turn on Show target and enter the Target value.
3. Set the three Range limits. Out of the box they’re 1, 2 and 3, which fits almost no real metric; 0 removes a range.
4. Name the ranges in a text box beside it.

Misleads when: the ranges are unlabelled, or a stretch target keeps it red all year. Bullets can’t be cross-filtered.

Gauge: where on a fixed scale

For: a score on a genuinely fixed scale, such as 0 to 100. Takes: one metric; optional target, comparison and up to five ranges (Gauge chart reference).

1. Add a gauge and set its metric.
2. Turn on Range limits (off by default) and add up to five.
3. Set Axis min and Axis max to the real ends of the scale.

Misleads when: someone picks the maximum. Set a revenue gauge’s maximum high and the month looks dire; low, and it looks healthy. Without a natural 0 to 100, use a bullet.

Waterfall: how we got from A to B

For: the gap between two totals: a revenue bridge, budget variance. Takes: one dimension (the steps) and one metric (each change) (Waterfall chart reference).

1. In a Sheet, one row per step: Step and Change. Add a waterfall with those fields.
2. Set Total calculation to At End (or monthly, quarterly, annual subtotals); Custom aggregation sets the starting baseline.
3. Sort by the metric, descending. The default sorts by the dimension.

Misleads when: the totals look like two more steps, or alphabetical steps hide the biggest mover in the middle. The report theme’s positive and negative change colours set its default colours.

Funnel: which step loses the most

For: ordered stages, from product view to purchase. Takes: a dimension of steps, one metric and a sort field (Funnel chart reference).

1. Add a funnel with Event name, filtered to your steps (view_item, add_to_cart, begin_checkout, purchase).
2. Set Event count as the metric and sort by it, descending.
3. In Style, pick smoothed bar, stepped bar or inverted triangle.
4. Show labels as Number + Percentage, with the percentage as % of previous, not % of max.

Misleads when: every step is a share of the first, so the step that got worse vanishes into a smooth taper, or one step counts users and the next counts events. The event fields are in connecting GA4.

Sankey: where each source ends up

For: a few sources flowing into a few targets. Takes: a source and a target dimension and one metric; Google warns the source and target data must be unique (Sankey chart reference).

1. Group the long tail first: a CASE field keeping the top four channels and labelling the rest “Other”.
2. Add a Sankey: source, then target, and Sessions as the weight. Row limit goes up to 5,000.
3. Copy it as a pivot table with a heatmap, for the exact numbers.

Misleads when: it has more than five or six nodes a side and turns into a tangle, or implies a journey when the data is only pairs.

Place

Geo chart: where the rate is higher

For: regional patterns, as long as the number is a rate. Takes: one geographic dimension and one metric; the first 5,000 rows (Geo chart reference).

1. Create the rate: Purchase revenue / Sessions * 1000, written directly. GA4, like Search Console and Google Ads, sends metrics already totalled, so SUM() around them fails (why).
2. Add a geo chart with Region and the rate; set Zoom area to the country.
3. In Style, set max, mid and min colours as one hue, dark to pale, and put a sorted table beside it.

Misleads when: it shades raw counts, which mostly maps where people live.

Google Maps: exactly where

For: a handful of named places, or thousands of points where the clustering matters. Takes: a location field plus a metric for size, colour or weight (Google Maps reference).

1. Choose the bubble, filled, heat or line map.
2. Set the location field. Filled maps don’t take latitude, longitude or addresses; international postcodes work in bubble maps only.

Misleads when: map tiles lend authority the data doesn’t have. Most field types cap at 3,500 bubbles or areas.

Community visualisations: beyond the built-in twenty

A community visualisation is a third-party chart added from a gallery; Google labels the feature Developer Preview and Beta (Community visualizations). Click Community visualizations and components in the toolbar, choose one, add its fields and consent. It needs owner’s credentials on the data source and an editor’s consent, so check the built-in route first:

The gallery, the credentials trap and the PDF problem are in my community visualisations guide.

Variants and charts you build yourself

• Scorecard with a sparkline: number plus trend, in scorecard space.
• Table with bars or a heatmap: a bar chart that exports, or a dense table made scannable.
• Small multiples: one time series filtered to one channel, axis min and max fixed by hand, copied into a grid with only the filter changed. Fixed axes are the point.
• Slope chart: a line chart (not a time series) with a two-value Period (“Sep LY”, “Sep TY”) from a Sheet and the item as breakdown.
• Histogram: a CASE field that bands each row (CASE WHEN Order value < 20 THEN "£000–020" … ELSE "£200+" END), then a column chart of the band against Record Count, sorted by band. Try two band widths.

The head-to-heads

Line or bar? Line for time, bar for categories. A line implies continuity: true of sessions over days, false of sessions by channel. With four quarters or fewer, columns read better.

Bar or column? Same chart, rotated. Horizontal bars when labels are long (page paths, queries), columns when labels are short or the axis is time.

Pie or bar? Bar, almost always. A pie earns its place only when the parts make a genuine whole, there are about four of them at most, and the point is “roughly half”.

The charts that are usually a mistake

• The many-sliced pie. Ten slices is six more than anyone can rank by eye.
• The decorative gauge. A quarter of a row for one number a scorecard shows in a fraction of the space.
• The dual-axis combo with unrelated metrics. Any two lines can be made to converge by the axis scaling.

All three come from choosing charts for variety. Variety isn’t a design goal. Clarity is.

Formatting that does more than the chart choice

• Set a limited palette once in the theme (colour palette guide).
• Don’t let colour carry meaning alone. Colour blindness affects about one in twelve men (Colour Blind Awareness), and red/green is the usual problem.
• Sort by the metric, and use one accent colour. If everything is highlighted, nothing is.

Starter sets, by what you’re building

A reliable page: three or four scorecards, one time series, one or two breakdowns, and a text box saying what changed. In-house reporting upward, lean on scorecards and one time series (my executive summary page guide). At an agency, settle a house pattern and copy it per client.

Meta, LinkedIn and TikTok have no native connector, so a cross-channel page needs a paid one: a separate decision.

How many charts should a page have?

Fewer than you think. Read the page as its recipient and ask what decision it helps them make; charts that don’t help are decoration. More in dashboard design principles.

Frequently asked questions

How many chart types does Data Studio have?

Twenty in Google’s documentation: scorecard, table, pivot table, time series, bar and column, pie, combo, geo, Google Maps, area, scatter, bullet, gauge, tree map, Sankey, waterfall, boxplot, candlestick, timeline and funnel. Plus an undocumented Histogram group, and community visualisations from third parties. Most good reports use about five.

How many metrics can a line or time series chart show?

Well over five. Google’s line chart reference gives up to 20 with one dimension, and a time series took 13 when I checked in October 2026. A breakdown dimension cuts it to one. Past about four lines, it stops being readable.

Why does my chart show an “Others” row?

The dimension has more values than the display limit, and “Group the rest as Others” is on by default. Raise the limit or filter, but don’t just switch it off: the values past the limit then disappear.

Why can’t I add Percent of total to this chart?

Comparison metrics, calculations such as Percent of total or Difference from max, aren’t compatible with “Group the rest as Others”. Disable it to use them, at the cost of the chart no longer showing all your data. Google’s limit is on those calculations, not on a previous-period comparison date range.

Does Data Studio have a histogram?

Yes: the Chart types list has a Histogram group, though Google hasn’t documented it yet. To choose the bands yourself, build one from a calculated field that bands each row, then a column chart of the bands against Record Count.

Can I use a chart type Data Studio doesn’t have?

Yes, through community visualisations. They need owner’s credentials and an editor’s consent, so check whether a built-in chart can do the job first.

Where to go next

• Designing scorecards that communicate at a glance
• Reference lines: targets, averages and dates on a chart
• Community visualisations, and the built-in alternatives
• Date range controls and comparison periods
• Dashboard design principles for readable reports
• Create your first report