A readable word cloud in Power BI, even from free-text comments

In short

What is a word cloud? A word cloud shows a list of words whose size tells their weight: the more often a word is mentioned (or the larger its value), the bigger it is written. You see at a glance what thousands of comments talk about, which topics come back in a survey, which products sell best.

Mini example. In 900 customer reviews, "delivery" appears 410 times, "quality" 360 times and "instructions" 25 times: "delivery" is the biggest word, "instructions" one of the smallest.

• What the visual does. Word Cloud Plume places the words in the middle of a shape (an oval by default), largest first, without overlap and always the same way for the same data. It splits long texts into words by itself, removes English and French stop words (the, and, for, le, de…), takes a measure for the size, a color per word, dozens of Windows and Office fonts, several orientations and 22 free shapes. As premium features (★), it also counts phrases ("customer service"), colors words by a measure (average sentiment), compares two periods (rising ▲ or falling ▼ words), fills the shape with repeated words, offers 131 more shapes or your own SVG, and makes the cloud change over time.
• When to use it. Customer reviews, answers to an open question, complaint reasons, keywords, ticket titles, most frequent products or topics.
• When to prefer something else. To compare close values precisely, a bar chart (Bar Plume) is more accurate: the eye is poor at comparing word sizes.
• What you will be able to do. Prepare the data, insert the visual, choose the size measure, split comments, exclude words, pick a shape, a font and the orientation, filter the page with a click.
• What you need: Power BI Desktop, the Word Cloud Plume visual installed (see the common basics). The sample report PlumeWordCloud-Sample-EN, in your "Downloads" area, shows topics and 900 fictitious reviews (page 1), shapes, fonts and orientations (page 2) and the ★ features (page 3).

The data to provide

The shape. Two cases, which the visual recognizes by itself:

1. One row per word (a topic, a product, a keyword) and a measure that gives its size (number of mentions, sales…).
2. One row per text (a review, an answer, a comment): the visual splits each text into words and counts them. A count measure is still recommended (see the pitfall below).

Topic	Mentions
delivery	980
quality	860
customer service	790
price	720
lead time	610
advice	560
packaging	520
return	470

To try it without a file: Modeling › New table, then paste:

Topics =
DATATABLE (
    "Topic", STRING,
    "Mentions", INTEGER,
    {
        { "delivery", 980 }, { "quality", 860 }, { "customer service", 790 }, { "price", 720 },
        { "lead time", 610 }, { "advice", 560 }, { "packaging", 520 }, { "return", 470 },
        { "store", 450 }, { "welcome", 420 }, { "sofa", 400 }, { "website", 390 }
    }
)

Then the measure (Modeling › New measure):

Number of mentions =
SUM ( Topics[Mentions] )   -- adds up the mentions of the topic in the current context (filters, slicers)

The identical comments pitfall. Power BI groups identical values of a column: two customers who both write exactly "Fast delivery" make a single row for the visual. So that every review counts, drop a count measure into Size:

Number of reviews =
COUNTROWS ( Reviews )   -- number of reviews behind each text (usually 1, 2 or more for identical texts)

Other pitfalls: an empty text is ignored; an empty, zero or negative size removes the word (a cloud cannot draw a negative size).

Step by step

1. Insert the visual. Click the Word Cloud Plume icon in the Visualizations pane. The visual shows "Add fields to get started".
2. Drop the words column. Topics[Topic] in Words or text. Each topic appears, sized by its number of rows (1 here: all the same).
3. Add the size. Number of mentions in Size. "delivery" becomes the biggest word, in the middle; the automatic title becomes "Number of mentions by Topic".
4. Look at the colors. By default, each word takes a color of the style palette, in size order. In Format › Data colors › Coloring, choose Single color then a Color: the whole cloud takes this hue, sizes make the hierarchy.
5. Pick a shape. Format › Shape › Shape: Heart. The words fit inside the heart. Free shapes are the oval, the rectangle and the 22 shapes without ★; the others (★) show a watermark without a license.
6. Pick a font and the orientation. Format › Word text › Font: Georgia (serif). Format › Layout › Word orientation: All horizontal for a very readable cloud, or Horizontal and vertical (default, 25% of the words rotated).
7. Set the spacing and the sizes. Format › Layout › Spacing between words (px) (2 px by default) loosens or tightens the cloud. Format › Word text › Minimum size (px) and Maximum size (px) bound the sizes; with Size fitting on Fill the shape, the words grow until they fill the shape; in a small shape, the least frequent words are left out rather than shrinking everything to the minimum size.
8. Switch to comments. Copy the visual, remove Topic and drop the reviews column (Reviews[Comment]) into Words or text, then Number of reviews into Size. The visual splits the texts into words, removes "the", "and", "for"… and shows the most used words.
9. Exclude useless words. Format › Words › Other words to exclude: type for example thanks ; product (one word per line, or separated by ; or ,). Case and accents are ignored.
10. Filter the page with a click. Click "delivery": the other visuals of the page only show the reviews that contain this word. Ctrl + click adds a word; a click on the background clears the selection. Right-click opens the Power BI menu (Drill through, Include, Exclude).

Troubleshooting

What you see	Likely cause	Fix
Every word has the same size	No measure in Size, and each word appears once	Drop a measure (Number of mentions) into Size
A whole comment shows as a single "word"	The texts are short and automatic splitting keeps them whole	Format › Words › Split text into words: Yes
Words like "very" or "really" remain	They are not in the stop word list	Add them to Other words to exclude, or a text measure in Excluded words
"12 word(s) not placed" note in edit mode	The shape is too small for every word, even at the minimum size	Enlarge the visual, lower the Minimum size, the Spacing or the Maximum number of words, or pick a wider shape
Words are tiny in a narrow shape (star, France)	Once in proportion, the shape leaves little room	Format › Shape › Proportions: Stretch to fill the visual, or fewer words
"Plume Visuals · Premium preview" at the bottom of the visual	A ★ feature is used (★ shape, phrases, comparison, play axis, silhouette…)	Go back to a free feature, or paste your key in Format › Plume Studio License
Identical comments count only once	Power BI groups identical values	Put Number of reviews (COUNTROWS) in Size

Key takeaways

• One column of words or texts, one size measure: the visual does the rest (splitting, stop words, layout).
• The cloud is always the same for the same data; Layout variant offers another arrangement.
• No word is ever dropped silently: the "words not placed" (or "less frequent words not shown") note tells you in edit mode.
• A click on a word filters the page on the rows that contain it.
• ★ shapes, phrases, comparison, repeated words, the play axis, gradient by size, logarithmic and rank scales, merged plurals, "once per row" and the silhouette are premium (★).

Learn more: how-to recipes (phrases, sentiment, comparison, SVG, play axis) · fields and settings · the common basics of Plume visuals · the visual's page · support.