Rethinking Data Visualization: A UX Framework for Dashboards That Drive Decisions

Source note: This article is adapted from Meriem Benhabiles' "Rethinking Data Visualisation: A UX Approach To Dashboards That Actually Drives Decisions," with edits for clarity.

Charts Are Not the Whole Truth

Francis Anscombe's 1973 quartet demonstrates that four datasets with identical summary statistics (mean, variance, correlation, regression line) produce radically different scatter plots: one shows a stable linear relationship, another a curved trend, a third is distorted by outliers. Visualization reveals what tables hide.

Figure 1: Anscombe's quartet — four datasets with identical summary statistics generating four completely different scatter plots

Visualization also communicates with the reader. Chart choice, information layout, and visual hierarchy determine whether the audience leaves with a pile of numbers or a repeatable judgment. Visual Capitalist's "History of Pandemics" uses proportionally sized bubbles on a shared timeline; before reading labels, viewers instantly grasp the Black Death's magnitude dwarfing all other pandemics.

Figure 2: History of Pandemics — a proportional bubble chart

Edward Tufte's data-ink ratio urges that every visual element must explain data, not decorate. But charts are never read in a vacuum; they are viewed by a specific person, in a specific context, under specific pressure. Stripping a chart to its barest form can remove the very context decision-makers need. Simplicity is not the goal; appropriate complexity is.

The 80% Before the Chart Appears

The chart is the final output; the high-leverage 80% happens before any tool is opened, before data is extracted, before any design choice. It reduces to three questions:

Context: What do we want to show with this data?

Audience: Who is the target audience? How do they think?

Insight: How does the data guide business change?

Context: Start from the Problem, Not the Data

Many data-intensive projects invert the process: they pull already-tracked metrics from internal tools and build visualizations around them. The question the data should answer is either assumed or never asked. Teams treat available data as the boundary of possible questions.

A vague prompt like "Tell me how the product is performing" yields a cluttered dashboard. A specific question — "Which features drive retention among Q1 sign-ups?" — simultaneously surfaces the metric, the cohort, and the implied product action. Once the question is concrete, data selection, comparisons, and noise reduction become clear.

Example: A UX team fixing an e-commerce checkout funnel. A data-first dashboard surfaces everything: clicks, scroll depth, device type, channel, page paths — leaving everyone asking "What do we actually change?" A context-first approach starts with "At which checkout step do users drop off?" Filtering out ~90% of irrelevant data produces a simple funnel chart that immediately highlights the payment-page bottleneck and points to the redesign target.

Audience: Design for Familiarity and Responsibility

Designing for the audience hinges on two factors: familiarity (data literacy) and responsibility (accountability).

Familiarity: Can the audience intuitively read the chart? Will a complex visualization add cognitive friction? Handing the same dense, multi-layered dashboard to a sales lead and a senior analyst is like giving the same map to someone who navigates by landmarks and someone who reads latitude/longitude — only the latter truly understands it.

Responsibility: A 12% decline means budget, targets, and organizational pressure for an executive; for an analyst tasked only with reporting, it is a result to be explained. Same chart, completely different weight.

This translates to a practical question: how much can you put in front of a person? Analysts need high-density environments for diagnostic exploration, willing to drill down. Executives need highly synthesized views to quickly judge growth drivers, risks, and resource allocation.

Figure 3: Left — high-density user journey exploration for analysts; Right — summarized executive overview for rapid budget and direction decisions

Insight: From Information to Action

Information answers "What does the data show?" Insight answers "How should we judge, adjust, and where do we go next?" Many projects assume that accurate, clear charts naturally yield insight. Reality differs: if the design process doesn't clarify the desired business change upfront, the dashboard remains passive reporting.

Example: Booking rate drops 15%. A dashboard that only "shows the drop" sounds an alarm. Leadership immediately suspects product failure, process error, or UX breakdown — triggering an expensive, possibly misdirected fire drill. An insight-driven dashboard places key variables side by side: booking rate, traffic sources, campaign launch timing, ad spend changes. The real cause surfaces: the app is fine, core user conversion is stable; overall metric dilution comes from a newly scaled campaign bringing low-intent clicks. The action shifts from product redesign to pausing the inefficient campaign and adjusting acquisition strategy.

Every visualization implies a next step — even if that step is "no change needed right now."

Case Study: Enterprise B2B SaaS Talent Management Platform

The platform collects massive daily user behavior data. Initial request: "We have huge user activity data; present it to enterprise teams." Building charts for all collected data would leave users with just data, not utility.

1. Context

First, define "performing well" in business terms. Which variables indicate progress vs. mere usage? Time-on-platform is a universal, easy-to-show proxy, but it's indirect: it shows presence, not value. More meaningful are competency scores, certification completion rates, and performance trends over time. Combining usage time with performance metrics adds an explanatory layer: identifying underused modules and linking underuse to low competency scores. Usage time becomes a supporting signal, not the core judgment.

Second, determine granularity. The same metric means different things to different viewers. Individual contributors check personal completion rates to gauge pacing; managers need team aggregates to pinpoint who needs support. This distinction dictates subsequent data presentation and filtering. Clarifying this structure early defines what to show and to whom.

2. Audience

The naive solution: same charts, just personal view vs. team aggregate — many dashboards call this "personalization" with minor scale adjustments. True audience-centric design reveals deeper differences: individuals and managers need different narrative structures and design principles.

Individual interface: A highly personalized mirror for self-assessment, providing granular, authentic, personally relevant information.

Manager interface: Puts individual details secondary; first shows macro team progress and persistent skill gaps. Prioritizes overall team status with clear, intuitive drill-down paths for when managers need to inspect individuals and address daily collaboration issues.

3. Insight

Focus shifted to pacing the narrative arc, not producing a high-density passive data log.

For individuals: Core value is autonomous planning. Ideal state: Monday morning glance yields this week's clear priorities.

For managers: Goal is to fundamentally change when operational conversations happen. Interface must supply judgment basis so managers intervene before skill gaps become critical project risks; visualization must surface nodes suitable for human follow-up, shifting management from reactive post-mortems to proactive guidance.

A "comparison tool" — never explicitly requested — became a standout feature: placing two team members side-by-side on the same metrics. It originated from a design hypothesis and ended up as the most resonant function.

The project's most critical decisions lay outside the original requirements frame. Collecting data customers didn't know to ask for, and organizing it to answer questions they couldn't articulate, is the core differentiator of user-centered data strategy.

Outcomes: Post-launch, platform analytics weekly active usage rose markedly. Managers shifted from monthly static report exports to weekly Monday planning sessions. Over the next two quarters, revenue and user growth trended positively, and churn dropped to a low point. Attribution to the dashboard alone is difficult; the strongest evidence is user feedback. Managers reported they previously only reviewed poor performers at month-end; now they instantly identify underperformance and schedule brief supportive check-ins before things worsen.

Summary

When visual presentation participates in upstream architectural decisions rather than downstream layout, data design reaches its full potential. Bringing structured UX thinking into data transforms visualization from passive display into an active decision-driving engine, ensuring every chart, report, and metric directly serves human purpose.