Information design and infographics

In short

Information design is the practice of shaping data, instructions and complex content into a form a reader can understand quickly and correctly; an infographic is one deliverable of that practice, combining charts, diagrams, type and illustration into a single explanatory graphic. It sits between statistics and visual communication: the numbers must be handled honestly, and the visual encoding must match how the eye and brain actually compare quantities. Indian university Visual Communication papers examine it as theory (Tufte's principles, Bertin's visual variables, chart types, the design process) and as studio work (a data poster, a dashboard, a process diagram). It is employed daily by newsrooms (The Hindu Data Point, Indian Express, Mint, Hindustan Times, IndiaSpend, Factly), by PIB and MyGov for government communication, by agencies and design studios, and by product teams building dashboards for banks, fintechs and SaaS firms. Placement tests typically ask you to pick the right chart for a dataset, spot a misleading graphic, and defend your layout and colour choices.

Explained

What information design is, and where it came from

• Information design: the deliberate organisation of content - data, instruction, navigation - so that a defined audience can find, read and act on it with the least effort and the fewest errors.
• Data visualisation is the sub-set that maps numbers to visual properties; information graphics (infographics) are composed pieces that may combine data, diagram, map, illustration and narrative text.
• Explanatory vs exploratory: explanatory graphics are authored to deliver a known finding to a reader; exploratory tools (dashboards, filters) let the user find findings themselves.
• Key figures: William Playfair (1780s-1800s - invented the line, bar and pie chart); Florence Nightingale (1858 polar-area diagram on army mortality); Charles Joseph Minard (1869 flow map of Napoleon's Russian campaign, six variables in one image); John Snow (1854 cholera dot map); Otto Neurath and Gerd Arntz (Isotype, 1920s-30s - repeated pictograms, one symbol = one fixed quantity, never a stretched symbol); Harry Beck (1933 London Underground diagram - topology over geography, the ancestor of the Delhi Metro and Mumbai local maps); Jacques Bertin (*Sémiologie Graphique*, 1967); Edward Tufte (*The Visual Display of Quantitative Information*, 1983); Cleveland and McGill (1984, experimental ranking of graphical perception); Alberto Cairo, Nigel Holmes, Giorgia Lupi (data humanism) in current practice.
• Three qualities commonly examined (after Cairo): truthful (faithful to the data), functional (easy to decode), insightful/beautiful (rewards attention without distorting).

How people read a graphic: perception and encoding

• Visual variables (Bertin): two planar variables (position on x and y) plus six retinal variables - size, value (lightness), texture/grain, colour hue, orientation, shape.
• Each variable has properties: selective (lets you isolate a group), associative (lets you group), ordered (shows rank), quantitative (lets you read a ratio). Position and size are quantitative; hue is only selective and associative, so never encode a quantity in hue.
• Cleveland and McGill's accuracy ranking (most accurate first): position on a common scale → position on identical non-aligned scales → length → direction/angle/slope → area → volume/curvature → shading and colour saturation. This is why a bar chart beats a pie chart, and a dot plot beats a bubble chart.
• Stevens' power law: perceived magnitude ≈ (actual magnitude)ⁿ, with n ≈ 1.0 for length, ≈ 0.7 for area and ≈ 0.5-0.7 for volume - so people systematically under-read areas and volumes.
• Gestalt principles used to build structure: proximity, similarity, enclosure, closure, continuity, connection (the strongest grouping cue), figure-ground, common fate. Grouping by white space is usually cleaner than grouping by boxes and rules.
• Pre-attentive attributes - colour, size, orientation, position, motion - are detected in under ~250 ms and are the tool for highlighting one bar out of thirty.
• Cognitive load: working memory holds only a few chunks, so a graphic must carry its legend, units and takeaway in place rather than making the reader hold them in mind.
• Shneiderman's visual information-seeking mantra for interactive work: *overview first, zoom and filter, then details on demand.*

Types of information graphics and how to choose

• Comparison: bar and column charts (ranked bars for categories), grouped and paired bars, dot/lollipop plots, bullet charts, radar (rarely justified).
• Change over time: line charts, area and stacked area, slope charts, candlestick, sparklines (word-sized, label-free trend lines set in running text), horizon charts.
• Part-to-whole: stacked bars, 100% stacked bars, treemaps (nested rectangles for hierarchy + size), sunburst, waffle/unit charts, pie and donut - acceptable only for 2-4 slices with a whole that genuinely sums to 100%.
• Distribution: histogram, box plot, violin plot, strip/beeswarm, population pyramid.
• Relationship: scatter plot, bubble chart (area = third variable), connected scatter, heat map, correlation matrix.
• Flow and process: flowchart, Sankey diagram (flow width = quantity), alluvial, chord, network/node-link graph, swimlane and BPMN-style process diagrams, Gantt chart for schedules.
• Space: choropleth (regions shaded by a *rate*, never a raw count), proportional-symbol map, dot-density map, cartogram, flow map, hex-bin map; schematic transit diagrams.
• Explanatory diagrams: cutaways, exploded views, step-by-step instructional sequences (IKEA-style, wordless), anatomical and technical illustration, timelines, org charts, comparison tables.
• Formats: single data poster, editorial graphic, long-form scrollytelling page, dashboard, report chart suite, animated/looped social graphic, print newspaper graphic.
• Selection framework (Abela's chart chooser): first name the question - comparison, composition, distribution, or relationship - then the number of variables and whether one is time; the chart follows from the question, not from taste.
• Small multiples: one chart form repeated across categories with shared scales - the cheapest way to show many variables without a clever new chart form.

The design process, step by step

1. Brief and audience: who reads it, on what device or paper size, how long they will look, what decision or takeaway it must support.
2. Acquire data: official source in preference to a secondary report; record the publisher, the year, the unit, the geography and the licence.
3. Clean and verify: fix state/district name spellings, handle missing values explicitly (do not plot them as zero), check totals, reconcile against the source table.
4. Normalise and transform: convert to rates (per 1,000 / per lakh / per capita), deflate money to constant rupees, compute shares, index to a base year, aggregate consistently.
5. Explore: make quick, ugly charts of everything; the finding is discovered here, not designed here.
6. Choose encodings using the accuracy ranking; decide what each axis, size, and colour means.
7. Wireframe/storyboard: hierarchy of title → key chart → supporting charts → source line; decide reading order (Z or F path, grid columns).
8. Design: type scale, colour palette, annotation, direct labelling, consistent rounding and number format.
9. Annotate: an informative, findings-based title ("Urban unemployment fell in four of six quarters"), labels on the marks, and notes for caveats.
10. Test: squint test (does the main message survive blur?), five-second test, greyscale print test, colour-blind simulation, read-aloud of every number against the source.
11. Produce and publish: export at the correct resolution/colour space for print or screen, with alt text and a machine-readable data link where possible.
12. Document: keep the cleaned dataset, the source URL and the date of access with the artwork file.

Graphical integrity, Tufte's principles, chartjunk

• Tufte's principles of graphical integrity: the representation of numbers should be directly proportional to the numbers; label clearly to defeat ambiguity; show data variation, not design variation; use standardised units and deflated currency for time series; the number of information dimensions must not exceed the dimensions in the data (no 3-D bars for one variable); never quote data out of context.
• Lie Factor quantifies distortion; an honest graphic has a Lie Factor of 1 (practically 0.95-1.05).
• Data-ink ratio: the share of ink that carries data. Maximise it within reason; erase non-data ink (heavy grids, borders, backgrounds, 3-D shading) and erase redundant data-ink (a bar that is also labelled, gridded and legended).
• Chartjunk: decoration that adds no information - moiré vignettes, gratuitous 3-D, the "ducks" where the graphic becomes an illustration of itself. Note the counter-evidence: Bateman's 2010 study found Nigel Holmes-style embellished charts were recalled better over time, so "zero decoration" is a position to argue, not a law.
• Data density = data entries per unit area; good statistical graphics are dense, and shrinking them (small multiples, sparklines) usually helps.
• Layering and separation: use weight, value and colour so that data sits forward and scaffolding recedes; grey the gridlines, black the data.
• Common distortions to recognise: truncated or dual y-axes, inverted axes, unequal time intervals, area scaled by radius instead of by area, 3-D pies, cherry-picked ranges, raw counts mapped on a choropleth, aggregating away a Simpson's-paradox reversal, and misleading class breaks on a map.
• Axis rules: bar charts encode length, so a bar axis must start at zero; line charts encode position and may be truncated if the range is labelled. Dual axes are best replaced by indexed lines or two panels.
• Banking to 45° (aspect ratio): in a line chart, choose the width-to-height so that the average line slope is near 45°, where slope differences are judged most accurately.
• Limitations of infographics: they compress and therefore simplify; a single graphic cannot carry uncertainty, caveats and methodology well; they are expensive to produce and go stale; a visually strong graphic lends false authority to weak data; accessibility for screen-reader users requires separate text or a data table.

Craft: colour, type, layout, accessibility

• Palette types: sequential (one hue, light→dark, for ordered magnitudes), diverging (two hues around a meaningful midpoint such as zero or the national average), qualitative (distinct hues for unordered categories - keep to about 7). ColorBrewer and Viridis are standard safe sets; Viridis is also perceptually uniform and greyscale-safe.
• Colour-vision deficiency affects roughly 8% of men and about 0.5% of women worldwide, with Indian studies reporting broadly the same order for men; red-green is the common form, so never rely on red vs green alone - add lightness difference, direct labels, or pattern.
• Contrast: WCAG 2.1/2.2 asks for 4.5:1 for normal text, 3:1 for large text (≥18 pt, or ≥14 pt bold) and for graphical objects and UI components. Level AAA asks 7:1.
• Type: one or two families; a clear hierarchy of title / standfirst / chart title / axis label / annotation / source; tabular (monospaced) figures for aligned numbers; avoid rotated y-axis labels where a horizontal bar chart would remove the need; set a minimum of about 8-9 pt for source lines in print.
• Numbers, Indian convention: Indian digit grouping (12,34,567), lakh and crore for domestic audiences with the international equivalent in a note for export work; ₹ symbol before the figure; state the unit and the year on the chart, not only in the caption.
• Layout: a modular grid, consistent gutters, a single dominant element, generous margins, and a fixed position for the source and credit line. Reading distance governs size - a poster read at 2 m needs far larger cap heights than a report chart read at 400 mm.
• Deliverable specs: screen graphics in sRGB at 1× and 2× pixel density; print in CMYK at 300 ppi for images and vector for all type and charts; large-format posters often at 150 ppi at final size.
• Accessibility and web practice: alt text that states the finding (not "bar chart"), a linked data table, no information carried by colour alone, keyboard-reachable interactive charts. Government of India websites follow GIGW 3.0 (2023), which is built on WCAG 2.1 Level AA.
• Ethics and law: facts and raw data are not copyrightable in India, but the expression of a chart or an illustration is protected under the Copyright Act, 1957; always credit the source and check the licence. Most datasets on data.gov.in are released under the Government Open Data Licence - India, pursuant to the NDSAP (2012). When visualising personal data, remember the DPDP Act, 2023; under the DPDP Rules, 2025 the Data Fiduciary duties (notice, consent, breach intimation, erasure) commence on 13 May 2027, with the earlier IT Act s.43A and the SPDI Rules applying until then. Aggregate and de-identify before publishing.

Indian data sources and practice

• Primary sources: Census of India (last completed 2011; the next Census is under way, with the reference date 1 October 2026 for snow-bound areas and 1 March 2027 for the rest of the country), MoSPI/NSO (PLFS for employment, CPI with base 2024=100 from January 2026, IIP and GDP on 2022-23 base, WPI on 2022-23 base), RBI Database on the Indian Economy, NFHS, NCRB *Crime in India*, UDISE+ and AISHE for education, Election Commission results, NITI Aayog SDG India Index, data.gov.in and the National Data and Analytics Platform.
• Caveats to annotate: changed base years break a time series; district and state boundaries change (new districts, J&K/Ladakh reorganisation in 2019, Telangana in 2014) so maps need a vintage; revised vs provisional estimates; population denominators projected forward from 2011.
• Map data: use boundary files consistent with Survey of India depictions; show Jammu & Kashmir and Ladakh as per the official Indian depiction for work published in India.

Key points

• Information design - organising content so a specific audience can find, read and act on it accurately.
• Infographic - a composed graphic combining charts, diagrams, type and illustration to explain one idea.
• Bertin's visual variables - position (x, y) plus size, value, texture, hue, orientation, shape; only position and size read quantitatively.
• Cleveland-McGill ranking - position > length > angle > area > volume > colour saturation, in decoding accuracy.
• Stevens' power law - areas and volumes are under-estimated by viewers; scale circles by area, never by radius.
• Lie Factor - shown effect ÷ actual effect; 1 is honest.
• Data-ink ratio - data ink ÷ total ink; raise it by deleting grids, borders, 3-D and duplicated labels.
• Chartjunk - decoration carrying no information; the Holmes/Bateman memorability findings are the standard counter-argument.
• Bar axes start at zero; line charts may be truncated if the range is labelled.
• Choropleth maps show rates, not counts - otherwise they map population.
• Sequential, diverging, qualitative - the three palette classes, chosen by whether data is ordered, centred, or categorical.
• Red-green alone fails ~8% of men - add lightness, labels or pattern.
• WCAG contrast - 4.5:1 body text, 3:1 large text and graphical objects.
• Small multiples and sparklines - high data density without inventing a new chart type.
• Shneiderman's mantra - overview first, zoom and filter, details on demand.
• Isotype rule - repeat a pictogram to show more; never stretch one pictogram.
• Informative title - state the finding, not the variable names; direct-label marks and drop the legend where possible.
• Always carry source, unit, year and note - an unsourced graphic is not publishable.

Formulas and examples

Example 1 - pie angles and shares. A college's enrolment: Arts 1,250; Science 900; Commerce 750; Engineering 500; Others 200. Total = 3,600.

• Arts: 1,250/3,600 × 360° = 125.0°, share 34.72%
• Science: 900/3,600 × 360° = 90.0°, share 25.00%
• Commerce: 750/3,600 × 360° = 75.0°, share 20.83%
• Engineering: 500/3,600 × 360° = 50.0°, share 13.89%
• Others: 200/3,600 × 360° = 20.0°, share 5.56%
• Check: 125 + 90 + 75 + 50 + 20 = 360°; shares sum to 100.00%. With five slices a ranked horizontal bar chart would be read more accurately; keep the pie only if the 100% whole is the point.

Example 2 - Lie Factor from a stretched pictogram. Scooter sales rise from 18 lakh to 27 lakh units. Actual effect = (27 − 18)/18 × 100 = 50%. The designer scales a scooter icon from 20 mm high × 10 mm wide to 30 mm × 15 mm (both dimensions ×1.5).

• Areas: 20 × 10 = 200 mm² → 30 × 15 = 450 mm².
• Shown effect (the eye reads area) = (450 − 200)/200 × 100 = 125%.
• Lie Factor = 125/50 = 2.5 - the graphic exaggerates two and a half times. Fix: repeat identical icons (Isotype) or use a bar.

Example 3 - truncated axis. Two colleges' placement rates: 96% and 100%. Actual effect = (100 − 96)/96 × 100 = 4.17%. The y-axis is drawn from 95 to 101, so bar heights are 1 unit and 5 units.

• Shown effect = (5 − 1)/1 × 100 = 400%.
• Lie Factor = 400/4.17 = 96 (to two significant figures). Fix: start the bar axis at zero, or use a dot plot with a labelled, non-zero range.

Example 4 - area-true bubbles. District A has 40,000 beneficiaries, District B 160,000, i.e. 4×. If A's circle has r₁ = 10 mm:

• Correct: r₂ = 10 × √(160,000/40,000) = 10 × 2 = 20 mm; areas 314 mm² and 1,257 mm², ratio 4.0 - honest.
• Wrong (radius scaled 4× to 40 mm): areas 314 mm² and 5,027 mm², ratio 16. Shown effect = 1,500%, actual effect = 300%, Lie Factor = 5.

Example 5 - rates, class breaks and contrast. Cases in five states: 12, 18, 35, 44, 90 (per lakh).

• Equal-interval, 3 classes: width = (90 − 12)/3 = 26 → 12-38 {12, 18, 35}, 38-64 {44}, 64-90 {90}. Quantile classes would instead place roughly equal counts per class and reveal more mid-range variation; state which scheme you used in the legend.
• Converting a count: 1,400 cases in a state of 3.5 crore = 350 lakh → 1,400/350 = 4.0 per lakh, i.e. 4.0 per 100,000.
• Label colour check: #595959 grey on white → each channel s = 89, c = 0.349, c_lin = ((0.349 + 0.055)/1.055)^2.4 = (0.3829)^2.4 ≈ 0.0998, so L ≈ 0.0998. Contrast = (1.00 + 0.05)/(0.0998 + 0.05) = 1.05/0.1498 ≈ 7.0:1 - passes AA (4.5:1) and AAA (7:1).
• Print spec for an A1 poster (594 × 841 mm) at 300 ppi: 594/25.4 × 300 ≈ 7,016 px by 841/25.4 × 300 ≈ 9,933 px.

Where it is used

• Newsrooms and data journalism: The Hindu Data Point, Indian Express, Mint, Hindustan Times, Business Standard, IndiaSpend, Factly, Scroll - roles as graphics reporter, data visualiser, interactive developer.
• Government and public communication: PIB and MyGov explainers, Union Budget graphics, Economic Survey charts, Census and NSO report design, Swachh Bharat and health-mission dashboards, NITI Aayog SDG index visuals.
• Product and analytics teams: dashboards and reporting screens at banks, NBFCs, fintechs (UPI/payment analytics), SaaS and ed-tech firms - roles as information designer, UI/data-viz designer.
• Agencies, consultancies and annual reports: investor-presentation and annual-report chart suites, ESG and CSR reports, pitch decks; large consulting and research firms keep in-house graphics desks.
• Healthcare, pharma and science communication: patient-information leaflets, dosage and procedure diagrams, hospital wayfinding content, research-paper figures and conference posters.
• Instructional and technical documentation: wordless assembly and safety sequences for appliances and two-wheelers, railway and metro schematic maps, museum and exhibition panels, textbook and competitive-exam diagrams.

Interview questions

• Q: When would you choose a bar chart over a pie chart? A: Almost always for comparing category values, because length on a common scale is decoded far more accurately than angle or area; a pie is justified only for a true 100% whole with two to four parts.
• Q: What is the Lie Factor and what value is acceptable? A: The ratio of the effect shown in the graphic to the effect in the data; it should be 1, and roughly 0.95-1.05 in practice.
• Q: Must a chart's y-axis always start at zero? A: Bars must, because the length itself carries the value; line charts encode position, so a truncated but clearly labelled range is acceptable and often more informative.
• Q: Why is a choropleth of raw counts wrong? A: Large or populous regions dominate, so the map ends up showing population or area; shade a rate, share or per-capita value instead and use a proportional-symbol map for counts.
• Q: How do you scale a circle to a value? A: By area - radius proportional to the square root of the value; scaling the radius directly exaggerates the difference by the square.
• Q: What is the data-ink ratio and is maximising it always right? A: The share of ink that carries data; raising it by deleting grids, borders and 3-D usually helps, but stripped-bare charts can be less memorable and harder for novices, so judgement beats dogma.
• Q: How do you make a chart accessible? A: Don't carry meaning in colour alone, meet 4.5:1 contrast for text and 3:1 for graphical elements, direct-label marks, write alt text that states the finding, and provide the underlying table.
• Q: Difference between a sequential and a diverging palette? A: Sequential runs light to dark in one hue for ordered magnitudes; diverging uses two hues either side of a meaningful midpoint such as zero, the mean or a national average.
• Q: How would you present fifteen metrics for twelve states in one page? A: Small multiples with shared axes and sorting, or a heat map of standardised values - not a single multi-line chart, and not fifteen separate pies.
• Q: A client wants the chart to make a 2% rise look dramatic. How do you respond? A: Decline the distortion and offer honest alternatives - annotate what the 2% means in absolute terms, show it against a benchmark or the longer series, or lead with an indexed line chart and a findings title.
• Q: What should every published graphic carry? A: An informative title, units and time period, axis labels, a source line with publisher and year, and any methodological note or caveat.
• Q: Where would you get reliable Indian data, and what pitfalls follow? A: data.gov.in, MoSPI/NSO, RBI DBIE, NCRB, NFHS, Census; watch for changed base years (CPI is now 2024=100, GDP and IIP on 2022-23), provisional versus revised estimates, projected population denominators, and district or state boundary changes that need a dated map.

Common mistakes

• Picking the chart first and fitting the data to it, instead of writing the question and letting the encoding follow.
• Plotting raw counts where a rate is needed - on maps especially - or comparing figures across years without deflating money or re-basing an index.
• Encoding quantity in hue, or using a rainbow palette for ordered data, so the reader cannot read magnitude.
• Scaling icons and circles by radius or by both dimensions, producing a Lie Factor of 2 to 5 without realising it.
• Decorative 3-D, drop shadows, heavy grids and framed boxes that lower the data-ink ratio and hide the data.
• Leaving out units, year, source or sample size - a guaranteed mark deduction in university vivas and an instant rejection in a portfolio review.
• Checking colour only on a bright calibrated screen: no greyscale print test, no colour-blind simulation, and no check that type survives at final print size.

Related topics

Branding and communication design · Publication design and layout · Environmental graphics and signage · Print production and prepress · Design software: Adobe tools · Typography and grid systems · Statistics and data handling