Text Boxes, Shapes and Smart Narratives in Power BI - Making Reports People Actually Read

There is a particular kind of Power BI report that gives me a headache the moment it loads. Twelve visuals crammed onto one page, no titles that mean anything, no explanation of what you are looking at, and a colour scheme that suggests the author discovered the theme palette five minutes before the deadline. Every chart is technically correct. Every number is right. And nobody in the business reads it, because reading it is work.

The gap between a report that is correct and a report that gets used is almost always about presentation and context. Not the data. The data is usually fine. What is missing is the stuff that tells a human what matters, where to look, and what to do about it. In Power BI that job falls to the humble supporting elements: text boxes, shapes, and the newer smart narratives feature. They are not glamorous. They are also the difference between a dashboard that informs a decision and one that gets opened once and forgotten.

Microsoft's reference for adding text boxes, shapes and smart narratives lays out the mechanics. I want to talk about how we actually use these things on client work, and where each one earns its place.

Text boxes do more than titles

The obvious use for a text box is a title, and yes, please put a title on your report. A shocking number of reports arrive with no heading at all, or a heading that says "Report 1". But text boxes earn their keep well beyond that.

The best use I have found is a short context line at the top of a page that tells the reader what they are looking at and, just as importantly, when the data is from. Something like "Sales performance, financial year to date, refreshed daily at 6am AEST". That one line stops half the questions a report generates. People want to know if the numbers are current and what period they cover, and if you do not tell them, they will assume the worst and email you to check. A text box answers it before they ask.

Text boxes also take dynamic values. You can drop a measure straight into the text so it reads "Total revenue this quarter: $1.4M" and that figure updates with the data. This is genuinely useful for the headline number you want people to see first, the one that should not require them to hunt through a chart to find. Used well, a couple of these at the top of a page act like a summary someone can read in three seconds before deciding whether to dig deeper.

One warning. It is tempting to write paragraphs. Do not. A text box full of dense explanation is a text box nobody reads. Keep it to a line or two. If you need to explain the report at length, that belongs in documentation, not on the canvas competing with the visuals for attention.

Shapes for structure, not decoration

Shapes in Power BI are rectangles, lines, and the like. The instinct is to treat them as decoration, and that is exactly the wrong instinct. Their real job is structure. A subtle rectangle behind a group of related visuals visually says "these three things belong together" far more effectively than white space alone. A thin line between sections tells the eye where one topic ends and another begins.

This is basic visual grouping and it is what separates a report that feels organised from one that feels like a pile of charts. When we build reports for clients, we spend real time on this layer, because a page that has clear visual regions is a page people can scan. The reader's eye should move through it in an order that makes sense: headline first, then the supporting detail, then the fine print. Shapes are how you enforce that order without the reader ever noticing you did it.

Where shapes get misused is when people go decorative. Drop shadows, gradients, borders on everything, a different accent colour for each box. It looks busy and it reads as noise. The best report design is quiet. The shapes should organise the page so calmly that nobody comments on them, because good structure is invisible. If someone notices your rectangles, they are probably too loud. This is the same principle we bring to any interface work, and it is a big part of why we treat report design as a proper discipline rather than an afterthought in our Power BI consulting engagements.

Smart narratives, the feature I have mixed feelings about

Smart narratives is the AI-driven one. You add a smart narrative to a page and Power BI generates a written summary of the data in plain English, updating automatically as the data changes and as users filter the report. In principle it is lovely. Instead of asking every reader to interpret a chart, the report writes its own commentary: "Sales increased 12% compared to last quarter, driven mainly by the New South Wales region."

When it works, it is genuinely good. For an executive who wants the story rather than the chart, a well-tuned smart narrative can be the most valuable thing on the page. It surfaces the movements that matter and states them in words a non-analyst understands immediately. For accessibility it is a real win too, because a screen reader can read the narrative aloud in a way it cannot with a bar chart.

Now the honest part. Out of the box, smart narratives can be underwhelming. The auto-generated text is sometimes bland, occasionally states the blindingly obvious ("the highest value was the highest value"), and does not always pick the insight you would have chosen. Left completely to its own devices it produces competent filler. The trick is that smart narratives are editable. You can rewrite the generated text, mix in your own sentences, and embed dynamic values so your wording updates with the data while keeping your framing. That combination, your judgement about what matters plus Power BI's ability to keep the numbers current, is where it actually delivers. Treat the auto-generated draft as a starting point, not the finished article.

I would also temper expectations. Smart narratives is good at describing what changed. It is not analysis. It will tell you sales dropped in a region. It will not tell you why, or what to do about it. That interpretation is still a human job, and anyone selling it as automated business analysis is overselling. It is a summarising tool, and a decent one, as long as you know its limits. Understanding where this kind of AI genuinely helps versus where it is window dressing is exactly the sort of thing we work through with clients in our AI strategy work, because the same pattern shows up across a lot of AI features: impressive demo, useful in reality only once a human shapes it.

Putting it together

A report page that works usually has all three of these doing quiet jobs. A text box up top with the title and a context line, maybe a dynamic headline figure. Shapes organising the visuals into clear regions so the eye knows where to go. And, where it suits the audience, a smart narrative giving the plain-English story for people who want words rather than charts. None of these are the data. All of them are what make the data land.

The mistake I see most often is teams pouring effort into the modelling and the DAX, getting the numbers perfectly right, then throwing the visuals onto a page with no thought for how a human reads them. The report is an A-grade underneath and a C-grade on the surface, and the surface is all the business ever sees. Presentation is not fluff. It is the part that determines whether your correct numbers actually change a decision.

If your reports are technically sound but nobody seems to use them, the problem is often this layer, and it is very fixable. We help Australian businesses turn accurate-but-ignored dashboards into reports people genuinely rely on, and a lot of that work happens in exactly these unglamorous supporting elements. If that sounds like your situation, get in touch and we will take a look at what your reports are doing and, more importantly, what they are failing to say.