Analysis is only useful when it reaches the people who need to act on it. This guide covers what data storytelling is, what separates a good data story from a collection of charts, and how to build and publish one using Data Canvas in AnalyZ Solutions.
Most analysis does not fail at the point of collection or computation. It fails at the point of communication. Findings get produced, filed, and forgotten, not because they lack value, but because they are never translated into a form that non-technical audiences can engage with. Data storytelling is the discipline that closes that gap.
Data storytelling is the practice of combining data, visualisation, and narrative to communicate findings to a specific audience in a way that drives understanding and action. It is not the same as reporting, which typically compiles outputs comprehensively for a record. It is not the same as a dashboard, which lets an audience explore data on their own terms. A data story is authored: someone has decided what the data means, who needs to hear it, and how to sequence the evidence so the argument is clear.
A data story has three components that work together. The data provides the factual basis. The visualisations make patterns visible without requiring the audience to interpret raw numbers. The narrative explains what the patterns mean, why they matter, and what follows from them. Remove any one of the three and the story breaks down: data without narrative is uninterpreted, visualisation without narrative is decorative, and narrative without data is assertion.
For programme teams, M&E practitioners, and researchers in the development and humanitarian sectors, this matters for a practical reason. Most of the people who need to engage with your findings are not analysts. They are programme managers, donors, community representatives, and decision-makers who have neither the time nor the background to interpret a regression output or a co-occurrence matrix. A data story meets them where they are.
Good data stories share a set of qualities that distinguish them from slide decks with charts or reports with summaries.
It makes one argument. The most common failure in data communication is trying to say too many things at once. A good data story is built around a single central claim and selects evidence that supports it. Findings that do not contribute to the argument are left out. Not because they are unimportant, but because including them dilutes the story. Every slide should make a point that advances the argument.
The headline is the finding, not the label. "Food security by region" is a label. "Southern regions reported the lowest food security scores across all three survey rounds" is a finding. A finding headline tells the audience what to conclude from the visual before they look at it. Audiences who have to draw their own conclusions from unlabelled charts often draw the wrong ones.
The narrative explains significance, not method. Narratives should answer "so what?", not "how did we calculate this?". Readers do not need to know that a logistic regression was used; they need to know that people who feel capable of acting environmentally are three times more likely to do so. Method language belongs in a technical annex, not in a story slide.
Visuals and text carry equal weight. The visual should show something that is hard to convey in words. The narrative should say something the visual alone cannot. When both work together, the reader processes the evidence faster and retains it longer.
It is written for one audience. A story written for a programme team and one written for a donor audience may contain the same findings but will differ in tone, assumed knowledge, and the conclusions they draw. Writing for "a general audience" usually means writing for no one in particular.
Data Canvas is a module in AnalyZ Solutions that puts these principles into practice. You collect the findings that matter: charts, tables, infographics, qualitative outputs. You add them to a canvas, and generate a story that arranges them into structured slides with AI-written headlines and narratives. Each slide can be refined with a plain-language instruction, edited for style, and then published to a shareable link or downloaded as a PDF.
The generated story is a draft, not a finished product. The AI structures findings into slides and writes narratives based on what you have collected and the context you provide. You review each slide, make adjustments where needed, and publish when the story reflects what you want to say.
A canvas is the container for a Data Canvas story. Each canvas has a title, an optional description, and a collection of findings drawn from your analysis. You can have multiple canvases, one per report, presentation, or audience. Canvases are independent: what you add to one does not affect any other.
A finding is any output you have produced in AnalyZ Solutions and added to a canvas. It could be a chart from the Data Visualisation module, a descriptive or inferential analysis result, an infographic, or a qualitative output such as a co-occurrence table or thematic summary. Each finding is tagged with the module it came from and a headline you provide. Findings of different types can be combined in the same canvas.
A story is the AI-generated sequence of slides that AnalyZ produces by reading all the findings in a canvas. Each slide has a headline, a narrative paragraph, and one or more visuals drawn from your findings. The AI assigns each finding to the slide where it fits most naturally, applies a layout appropriate for the visual types involved, and writes a narrative that explains what the findings show. You can regenerate the story at any time, or refine individual slides without touching the rest.
Each slide in a story uses a layout that matches the visual content available. The AI selects the layout automatically based on the types of findings involved. Available layouts include single and paired charts, chart with a text panel, stat callout cards for key numbers, and full-text slides when no visual is assigned. Charts that work best at full width such as heatmaps always render full-width regardless of other slide content.
| Layout | What it shows | When it is used |
|---|---|---|
| Single chart | One chart or infographic at full slide width. | When the finding is a single strong visual that does not need a paired element. |
| Two charts | Two visuals side by side. | When two related findings benefit from direct visual comparison. |
| Chart and text | A chart on one side, a narrative or key point on the other. | When a visual finding needs a written explanation or a supporting observation. |
| Chart and quote | A chart alongside a highlighted qualitative excerpt. | When a quantitative finding is accompanied by a direct quote that adds meaning. |
| Stat callout | A designed card highlighting two to four key numbers. | When the finding is a table or summary with a small number of headline figures. |
| Full text | A narrative slide with no visual. | When no unassigned visual remains, typically for introductory or concluding slides. |
Before generating a story, you fill in a short context form with four fields: what the study is about, who the audience is, an optional key message you want the story to land on, and a presentation duration (5, 15, or 30 minutes). The AI uses these to calibrate the tone, depth, and structure of the slides. A story written for a programme team in an internal review session reads differently from one written for a donor audience at a presentation. Answering these fields accurately produces a more appropriate first draft.
Findings are added from within the analysis module where you produced them. After running an analysis or viewing a chart, look for the + Add to Canvas button in the result panel. Clicking it opens a short form where you can write a headline for the finding and select the canvas to add it to. If no canvas exists yet, you can create one from the same form.
Navigate to Data Canvas from the main menu. The canvas list shows all your canvases with the number of findings in each and the date last modified. Click any canvas to open it. If a story has already been generated for a canvas, the Story tab is available alongside the Findings tab.
From the Findings tab, click Generate Story. A context form opens with four fields: what the study is about (required), who the audience is (required), an optional key message you want the story to land on, and a presentation duration 5 minutes for key messages only, 15 for balanced coverage, or 30 for a full treatment. Fill in the required fields, then confirm. The AI reads all the findings in the canvas together with your answers and generates a sequenced set of slides. This typically takes 20 to 40 seconds depending on the number of findings.
Once generation is complete, the view switches automatically to the Story tab. Each slide shows its headline, narrative, and one or more visuals drawn from your findings. Slides are ordered to form a coherent narrative arc. The AI assigns a layout to each slide based on the visual types involved.
Click Preview to open a full-screen viewer of the story as it will appear to readers. The preview shows the slides in their published layout, including all visual sizing and narrative text. Use the preview to check that each slide reads clearly before refining or publishing. The preview opens in a right drawer without leaving the canvas.
Each slide has a Refine button. Click it and type a plain-language instruction describing what to change. Instructions that address content adding a finding, changing which visual is shown, adjusting the narrative emphasis are handled by the story generation route. Instructions that address appearance colour, font size, layout arrangement are handled by the styling route. You do not need to specify which type of instruction you are giving; the system detects the intent and routes accordingly.
After typing your instruction, click Apply. The slide updates immediately. Other slides are not affected. You can refine the same slide multiple times until it reflects what you want.
For direct control over the headline or narrative text on any slide, click Edit text to open the text edit drawer. This panel gives you separate editable fields for the headline and narrative, along with formatting controls: bold, italic, underline, font size (XS through 2XL), font family, and a colour picker with presets. Changes are applied directly to the slide and saved to the story. Use the text editor for final wording adjustments rather than regenerating the slide.
Each chart or infographic on a slide has height adjustment controls: a minus button and a plus button that change the visual height in 80-pixel increments. Use these when a chart is cropped or has excess whitespace. Height adjustments apply to the current session; they are not saved between sessions unless explicitly noted in the slide.
When the story is ready, click Publish. A short form opens where you can add an optional note for readers: a sentence or two explaining what the story covers or who produced it. After confirming, the story is assigned a public link that anyone with the URL can view. The published view shows the story in its slide layout with your visuals and narratives, with a PDF download option in the footer.
To update a published story after making changes, click Re-publish. This replaces the previous published version with the current state of the slides. The public link remains the same.
To restrict access, use the Public / Private toggle at the top of the canvas. Setting the story to Private makes the public link inactive without deleting the story. The story can be made public again at any time.
Watch how to create charts and publish a dashboard in AnalyZ Solutions.
Building and publishing a Data Canvas in AnalyZ Solutions.
The headline you write when adding a finding is the single most important input the AI has for that slide. A headline written as a statement "Participation rates were 22 percentage points higher in urban areas" tells the AI what the finding means. A label "Urban vs rural participation" tells it only what variables are involved. Statement headlines produce slides with clearer narratives and more relevant layouts.
The context questions before generation shape the tone and depth of the story. If you answer "programme team" for a story you will actually present to donors, the narrative will read as an internal technical discussion rather than a polished external brief. Take a moment to answer the context questions accurately; it saves a full regeneration later. The Sense Making guide explains how context shapes an interpretation in the same way.
If a single slide needs adjustment (a different visual, a shorter narrative, a change to the infographic appearance), use Refine on that slide. If the overall sequence is wrong, if you have added or removed findings, or if you want to try a different context, regenerate the full story. Regeneration replaces all slides; refining affects only the one you target.
The story is not automatically updated when you add a new finding or remove an existing one. If you add a chart after an initial generation, regenerate the story to produce a version that includes it. Slides generated from an earlier set of findings will not reflect changes to the canvas until the story is regenerated.
Before clicking Generate Story, review the Findings tab to confirm that the right outputs are included. Each finding shows its headline, source module, and a preview of the visual. Remove any outdated or duplicate findings before generating; the AI will include everything present in the canvas, and redundant findings produce redundant slides.
When the narrative on a slide is almost right but needs a specific word changed, a sentence removed, or a figure corrected, use the text editor rather than a refine instruction. Refining re-generates the slide narrative from scratch; editing gives you direct control. Reserve regeneration for structural or content changes, and use the editor for final polish.
Explore AnalyZ Solutions. Free to sign up, free to analyze.
Get started