Subject is pulling away from baseline
Index vs control · period 0 = 0 · 6 periods
Source: Illustrative product analytics (fictional)Data Visualisation lab by Gagan Malik — judgment over catalogue: which comparison to draw (level, rate, share, index), how to show fog (confidence bands, forecast fans), when to use small multiples, waffles and per-capita maps instead of raw choropleths, and what not to trust (radar rankings, overstuffed donuts, area-as-people). Forced-dark SVG on site --chart-* tokens.
Technology
Data Visualisation
By Gagan Malik
1 min read
1. Pick the comparison
Versus baseline, peer, prior, or target.
Source: Illustrative product analytics (fictional)Most categories moved up
Score 0–100 · before vs after · demo pair
Source: Illustrative experiment (fictional)Referrals lead; Support NPS trails
Score 0–100 · you vs cohort median · Q2 2025
YouCohort median
Source: Illustrative product health benchmark (fictional)You outperform your age group in most categories
Score by category — you vs. cohort average
Language
601
603
Problem Solving
618
482
Memory
624
506
YouAges 40–49
Source: Demo dataThree metrics still short of target
Actual vs gap to target · onboarding funnel · week 32, 2025
ActualGap to target
Source: Illustrative OKR dashboard (fictional)The gap varies sharply by category
Two values compared, sorted by size of gap
FromTo (increase)To (decrease)
Source: Demo dataLeave rises with age
% within age band · Leave vs Remain · Jul 2025 poll
LeaveRemain
Source: Fictional omnibus survey, n=1,800Wednesday saw the sharpest spike in blocked threats
Last 7 days
Daily average:346 blocked threats
Ads and TrackersMalicious websites
Source: Demo data2. Show the exception
Mark the series or period that breaks the pack.
Subject breaks from the pack
Metric by cohort, over time
Source: Demo dataThe 2020 turn reshaped the trajectory
Brand health vs price sensitivity · index · 2018–2024
Source: Illustrative brand tracker (fictional)Which categories moved, and where
Arrow points from before → after, on two measures at once
Source: Demo dataMost points track the line — one doesn't
A vs B, with the flagged outlier called out
Source: Demo dataSubject climbs from 4th to 1st
Rank over time — 1 is top
Source: Demo dataHere's what moved the total
Running total, step by step
Source: Demo dataOne category is pulling ahead of the rest
Index over time
Source: Demo dataWhere cardio fitness stands today
Above Average
37.6 VO₂ max
Jul 2025
High
Above Average
Below Average
Low
Source: Demo data3. Show the fog
Bands, fans, and distributions — not just the point.
Estimate widens as horizon extends
Modelled estimate · 95% CI · t = −2…11
Central estimate95% interval
Source: Illustrative statistical model (fictional)Uncertainty widens after the last observed month
MAU (thousands) · UK + IE fintech · Aug 2025 forecast
ObservedScenario band
Source: Illustrative MAU projection (fictional)Your cardio fitness trended above average
Aug 2024 – Aug 2025
LowBelow AverageAbove AverageHigh
Source: Demo dataA's range is pulling away from B's
Min–max per period, by group
Group AGroup B
Source: Demo dataSpread varies more than the average suggests
Distribution by group
Source: Demo dataWhere the distribution actually clusters
Each dot is one observation
Source: Demo dataSelf-serve clusters fastest; Free tier spreads wide
Days-to-first-value density · H1 2025 cohort · by segment
Source: Illustrative onboarding timing (fictional)Comparing several items across dimensions at once
One line per item, one axis per dimension
Source: Demo data4. Prefer multiples
Shared-scale panels beat spaghetti.
Coast breaks away
Sales index · Jan 2025 = 100 · six UK regions · Jun 2025
Typical regionException — Coast
Source: Illustrative regional sales index (fictional)Metrics at a glance
Inline sparklines
Sessions
Active users, 7-day
12.4k
+1.2kConversion
Checkout completion
3.2%
+0.4%Latency
p95 response time
180ms
-42msBounce
Landing exits
41%
-2.1%Source: Demo dataGrowth differs by category
Index over time
Category A
Category B
Source: Demo dataVolume clusters midweek
Support tickets per day · EU timezone · Jul–Aug 2025
Source: Illustrative helpdesk export (fictional)Activity peaks midweek
Intensity by row and month
Source: Demo data5. People, not area
Waffles and rates — not raw choropleths.
Only 37 of 100 completed signup
People per 100 · UK bank app adopters · Mar 2025
Source: Illustrative adoption survey (fictional), n=2,400Ranking flips when you rate per capita
Event share vs per 100 people · five UK regions · 2024
#1 North#5 West
Source: Illustrative public-health events (fictional)Bottom 40% earn ~14% of income
Cumulative income share · household deciles · FY2024
Source: Illustrative inequality snapshot (fictional)Exposure is concentrated in a few regions
Share of total exposure · schematic regions · demo
LowHigh
Source: Teaching toy (not Natural Earth)6. What not to trust
Forms that usually fail the job.
Chart forms that lie
Pick a better form when you see these
- Radar charts for ranking
- Donut charts with 8+ categories
- Raw choropleth for people counts
- Spaghetti lines without small multiples
Source: Data Visualisation lab — judgment notesYour brain in a nutshell
How you're performing in each category, 0–100
ScoreBaseline
Source: Demo dataOne segment leads, the rest trail off
Share of total, %
Segment A (42%)Segment B (28%)Segment C (18%)Segment D (12%)
*Percentages may not sum to 100 due to rounding.
Source: Demo data7. Rates and shares
When the mix is the story.
Opinion shifted sharply after the event
% responding, by period
Before
55%
25%
20%
After
78%
12%
10%
SupportUnsureOppose
Source: Demo pollingWhere prospects drop off
Count at each stage
Source: Demo dataWhere time in the app actually goes
Most used
Entertainment
23m
Productivity & Finance
15m
Social
11m
Creativity
7m
Other
48s
Source: Demo dataA few products make up most of the total
Rectangle size = share of total
Source: Demo dataHow the whole breaks down, two levels deep
Inner ring = category, outer ring = sub-item
Category A (40%)Category B (35%)Category C (25%)
Source: Demo dataMost of the variable segment sits in one group
Value-sized quadrants — box area is proportional to value
Fixed segmentSegment CSeries ASeries B
Source: Demo dataWhere the flow leaks
Stage-to-stage volume
Source: Demo dataUptake is accelerating, and deepening
Share reporting each frequency, %
DailyMonthlyYearly
Source: Demo dataComposition ebbs and flows over time
Stacked with a wiggle baseline, not a flat one
Source: Demo dataNet change has swung both ways
Gains above the line, losses below
Source: Demo data8. Maps
When place is causal.
Opportunity is concentrated in a handful of major markets
Composite index by country, 0–100
0100
Loading map…
IndiaChinaUnited StatesJapanGermanyUnited KingdomFranceAustraliaBrazilCanadaRussiaSouth Africa
Source: Demo dataIndia composite index
Index 0–100, trailing 5 periods
0100
Loading map…
Source: Demo dataIndia, state by state
All states and union territories
Loading map…
Source: Demo dataWhere it's concentrated
Value by location — the highlighted point is the one that matters
LowMediumHighHighlighted
Source: Demo dataThe route it took to get here
Path over time, by stop
Source: Demo dataShare of population in high or extreme poverty
By state, % of population
0100
Loading map…
Source: Demo dataAppendix
Strategy diagrams.
Expected value guides decisions under uncertainty
Decision tree with chance nodes — outcomes weighted by probability
*EV = (payoff × probability) + (payoff × probability). Choose the branch with the higher expected value.
Source: Demo dataThe threat to fight only deters entry if it’s credible
Sequential game tree — the entrant moves first, the incumbent responds
Entrant payoffIncumbent payoff
*Payoffs shown as (Entrant, Incumbent). Backward induction: the incumbent accommodates rather than fights (2 > –2), so entry is rational and the incumbent’s threat is not credible.
Source: Demo dataNeither player benefits from defecting alone
Payoff matrix — (Player A, Player B), classic Prisoner's Dilemma
Player A payoffPlayer B payoff
Source: Demo dataMultiple equilibria — no single dominant strategy
Stag Hunt: high payoff if both cooperate, but risk of mutual loss
Source: Demo dataFirms set quantities based on each other's output
Best-response curves — Firm B's optimal quantity given Firm A's choice
The Nash equilibrium occurs where both firms' best-response curves intersect. Each firm is maximizing profit given the other's choice.
Source: Demo dataOptimal bid trades off winning against profit margin
Bidder payoff by bid level (sealed-bid first-price auction)
The optimal bid balances the probability of winning (higher bid = higher P(win)) against payoff if winning (bid = lower profit margin).
Source: Demo dataCooperation breaks down under repeated play
Prisoner's Dilemma — share of rounds both players cooperated
Both players start cooperating, but incentives to defect build. Once defection begins, cooperation rarely returns.
Source: Demo dataHow the pieces connect
Node size = importance, line = relationship
Source: Demo dataWork together?
View plans, ask a question, or book a call.
