Setting the file. One moment.
Skill 02 · Dreambase Data Stories
Subchapter 2.4
references/storytellers.mdMarkdown8 KBView on GitHub
Distilled working principles from the most revered data-storytelling practitioners, verified against primary sources (July 2026). Read this when deciding how a story should be told, when justifying a bespoke form, or when a stakeholder asks “make it like the NYT/FT.”
Astronomer-turned-designer; Data Sketches (with Shirley Wu) documents her full process across 24 projects. Start from the actual data and goal, not a chart-type catalog; sketch before code — and when the data’s structure is unknowable, “sketch with code” with disposable prototypes; iterate toward a visual metaphor (her royal-genealogy network became glowing star constellations after she noticed the resemblance). Bespoke shapes make stories memorable where generic charts don’t. — visualcinnamon.com · datasketch.es
Her 2017 manifesto’s actual tenets (verbatim section headers): embrace complexity · move beyond standards · sneak context in (always) · remember that data is imperfect (as we are). “Numbers are always placeholders for something else.” Dear Data (with Stefanie Posavec): 52 weeks of hand-drawn personal-data postcards — data as intimate, imperfect, human. Density can be a feature when it invites slow reading. — giorgialupi.com · dear-data.com
“The Pudding explains ideas debated in culture with visual essays.” Their process: pick questions that spark debate among friends; open with a single data point, then zoom out; workshop the arc “like a movie script, not critiquing a bar chart”; visuals carry the narrative with sparse prose; collect original data when none exists (they hand-measured 80 pairs of jeans for the pockets piece). — pudding.cool, process guides “How to Make Dope Shit” parts 1–3
Archie Tse’s three rules (Malofiej 2016, verbatim): (1) “If you make the reader click or do anything other than scroll, something spectacular has to happen.” (2) “If you make a tooltip or rollover, assume no one will ever see it.” (3) Interactivity must work on all platforms or not exist. “Readers just want to scroll”; most visuals are static; animation triggers on scroll; “we are writing and editing a lot more text.” Amanda Cox (reliably reported secondhand): “The annotation layer is the most important thing we do.” — github.com/archietse/malofiej-2016
D3 creator, Observable co-founder, ex-NYT. “Examples are lightweight and informal… capable of expressing big ideas with immediate impact” (For Example, Eyeo 2013). Ship small disposable examples, one idea each; build the repertoire before the abstraction; judge tools as user interfaces (“Programmers are people, too”); make work forkable. His NYT work (rent-vs-buy calculator, 512 Paths to the White House) shows direct manipulation winning when the reader has a genuine personal input. — bost.ocks.org/mike
The COVID trajectory charts: log scale, one saturated line in a sea of grey, doubling-time references, technical choices justified in plain language on the chart. “Text is where people’s attention goes first. If we’re not using that, we’re really missing an opportunity.” Claim-titles that answer the reader’s question; direct labels over legends; iterate in public daily. “Don’t just make charts for chart people. Make stories for all people.” — ft.com/john-burn-murdoch; GIJN webinar coverage 2025
Physical-scale metaphors make abstract numbers visceral (Drowning in Plastic‘s rendered pile of 1M bottles). “That little publish now button is not the end for us” — graphics must survive syndication across clients and formats. Automate the repeatable chart so human craft concentrates on the exceptional story; put chart tools in every reporter’s hands. — reuters.com/graphics
“Knowledge compression”: at-a-glance gestalt, relative-not-absolute figures (the Billion Dollar-o-Gram’s core move), data as creative medium. Take the aesthetics and shareability; heed the documented critique — Stephen Few (2011, verbatim): his visualizations “display information in ways that hide much that’s relevant and essential.” The boundary: area-scaled blobs are perceptually imprecise; when the comparison matters, position/length wins. Beauty earns attention; encoding earns trust. — informationisbeautiful.net
(Site shuttered March 2025; use archives.) The 2020 forecast redesign (Anna Wiederkehr): lead with the distribution, not the point estimate — ballswarms of simulated outcomes, 100-dot scenario arrays; reframe the metaphor (weather, not coin flip); a guide layer teaching readers how to read the graphic inside the graphic (Fivey Fox); modular cards, each one lens on the model; published design rationale. Gelman et al. (2020) honestly note uncertainty-device effectiveness “is hard to judge” — intuition, not measurement. Alumni carrying the practice: Katie Marriner (AP), Gus Wezerek (NYT Opinion), Julia Wolfe (Reuters), Reuben Fischer-Baum (Washington Post).
The synthesis this skill’s workflow is built on. Contested ones are marked — knowing where the masters disagree is part of the craft:
Claims that circulate widely but failed primary-source verification — do not cite them:
Source