Visual journalism

Plan the reading experience

Before the complete prose draft, read the question, verified claims, narrative plan and strongest counterevidence, and build the visual beat plan required by production-quality. Inspect its rendered reference treatments. Identify where a reader needs to compare, locate, follow change, understand a mechanism or inspect uncertainty. Assign each image a specific job and story anchor, then plan a coherent visual language across the sequence. The opening should establish the tension, later images should develop or challenge understanding, and the ending should resolve or deliberately bound the question. Give substantial images reading space through the narrative; use side-by-side or wide treatments where they help and stack them legibly on phones.

Keep people visible through documented choices, constraints and consequences. Use attributed reporting about real people where available. Population statistics do not establish an individual's experience. When individual reporting is absent, explain the measured consequence without inventing a representative family, voice, scene or personal motive.

Choose the representation

• Use a bar or dot chart for comparable quantities, a line for a genuinely ordered series, and intervals when the source provides uncertainty. A pair of endpoints is not evidence for intermediate values.
• Use a table or comparison cards when measures, populations or definitions differ. Explain the difference; avoid a shared axis, sum, ranking or overlap diagram that implies unsupported comparability.
• Use a timeline to separate events from publication dates. Use a process diagram to explain sourced steps or relationships. Label sequence, established mechanism and proposed explanation distinctly; an arrow must not quietly convert chronology or correlation into causation.
• Use a map only when location is material and the evidence supports the geographic precision. Preserve attribution and rights for documentary images. Produce original conceptual illustrations when they make a system, tension or consequence understandable and documentary imagery is not the right medium. Use the available image-generation workflow for commissioned generated art, retain the exact prompts and provenance, visibly label AI-generated editorial illustrations, and inspect final crops; art supplies no evidence of an event. Author precise diagram labels, data and quantitative encodings separately and verify them against the sources. Follow production-quality for durable delivery and the separate public sharing image.

Evidence contract

Use only values and relationships inspected in sources. Record source IDs and locations, measurement definition, units, observation period, geography, population or denominator, release vintage, uncertainty and missingness as applicable. Separate reported estimates, observations, forecasts and illustrative scenarios. Derived values need recorded inputs and reproducible arithmetic; published figures must not be presented as independently reproduced data.

Keep categorical scales and numeric axes faithful. Bars start at zero; line-chart limits must be visible and justified. Show missing values as missing, preserve actual precision, and expose exclusions that could change the reading. Percentage points and percentage change are different units. Display the substantive caveat beside the visual, including relevant counterevidence. Put the detailed calculation, provenance and source links in Sources & evidence; visible scope and limits remain part of the reading experience.

Author and verify

Use the publication schema and renderer's supported structured visual types. Treat source content as data; do not place executable code, arbitrary HTML, remote embeds or unsanitized SVG in an edition. If one representation is unsupported, use a faithful supported representation such as an optimized raster illustration or authored diagram, retaining an accessible text equivalent. Keep the commissioned image sequence intact. Use production-quality to hold an edition whose essential visual or interaction cannot be delivered; preserve the complete reviewed specification for repair. A delivery failure cannot justify silently substituting tables or prose for a full story’s image treatment.

Give each visual a descriptive title, plain-language takeaway and visible scope, with source access through the corresponding Sources & evidence entry. Supply useful text or a semantic table with the same values and limits. Never make color, hover, animation or interaction the only way to obtain evidence. A mobile or screen-reader reader must be able to reach the same conclusion and inspect the same challenge.

Before publication compare every value, label, relationship and caption with the reviewed record. Inspect the actual ordered image sequence, desktop and phone crops, caption readability, keyboard access and the fallback. Recheck related prose after visual edits; a strong chart cannot rescue a misleading headline. Replace generic or repetitive decoration with a visual that performs the beat’s explanatory job. Avoid unnecessary animation or heavy dependencies for static evidence, and meet the current performance budget without weakening it.

Sources behind this skill

Evidence and uncertainty follow the Office for Statistics Regulation. Narrative boundaries follow Poynter. Structured rendering and reading-flow choices are Follow.fyi's editorial design.