How to design dashboards that help decision-makers

Introduction

Pharma companies generate enormous amounts of data: Production data, quality data, clinical trial results, supply chain metrics, financial performance. But data alone does not create value. It is the insights that drive decisions.

A good dashboard transforms data into actionable insights. A bad dashboard is at best ignored, at worst it drives the wrong decisions based on misunderstandings.

At Hyperbolic, we design dashboards for both pharma and general software. Here are our principles for dashboards that are actually used.

The Three Types of Decision Makers

Different roles have different needs. A dashboard designed for the VP of Operations is useless for a batch supervisor.

Strategic (C-level, VPs):

• Need: High-level trends, KPIs, comparison to targets

• Time Horizon: Months to years

• Action: Strategic planning, resource allocation

Tactical (Managers, Directors):

• Need: Performance tracking, problem identification, resource optimization

• Time Horizon: Weeks to months

• Action: Process improvement, team management

Operational (Supervisors, Specialists):

• Need: Real-time status, alerts, detailed diagnostics

• Time Horizon: Hours to days

• Action: Immediate problem-solving, daily execution

Common Mistake: Attempting to build one dashboard for all. Result: Too much info for strategic, too little detail for operational.

Best Practice: Layers of dashboards with drill-down capability. VP sees trends, can click for manager view, further to operational detail.

The Five Principles for Effective Dashboards

Principle 1: Start with the decision, not the data

Do not ask "What data do we have?" Ask "What decisions need to be made?"

Example - Manufacturing:

• Decision: Should we increase staffing on production line A?

• Data Needed: OEE trend, planned vs actual output, defect rate, backlog

• Not Needed: Historical maintenance logs, detailed material traceability (unless OEE indicates a problem)

Principle 2: Less is more

Cognitive overload is real. Humans can process 5-9 metrics at a time. More than that, and everything becomes noise.

Bad Dashboard: 30 metrics, 15 graphs, 5 tables on one screen. Nothing stands out.

Good Dashboard: 5-7 key metrics, clear visual hierarchy, the rest accessible via drill-down.

Principle 3: Make deviations visible

Decision makers need to quickly see "Is something wrong?"

Techniques:

• RAG (Red/Amber/Green) status: Immediate visual signal

• Exception highlighting: Show only what is outside of the normal range

• Trend arrows: ↑↓ show direction, not just absolute value

• Benchmarking: Comparison to target, previous period, or peer group

Principle 4: Context is critical

A number without context is meaningless. "65" means nothing. "65% OEE (target: 75%, last month: 70%)" drives action.

Always provide:

• Target/benchmark: What is good?

• Trend: Is it improving or getting worse?

• Comparison: Vs. last period, vs. peers, vs. plan

Principle 5: Design for mobile

Decision makers are not always at their desks. A dashboard that only works on a 24" screen is not used.

Mobile Considerations:

• Responsive design

• Prioritize metrics for small screens

• Touch-friendly interactions

• Offline capability for factory floor

Pharma-specific dashboard use cases

Production Performance Dashboard

Target User: Plant Manager (Tactical)

Key Metrics:

• OEE (Overall Equipment Effectiveness): Combined metric of availability, performance, quality

• Planned vs. actual output: Are we on track?

• Right-first-time rate: Quality metric

• Changeover time: Efficiency of line transitions

• Deviation count: Quality/compliance indicator

Visual Design:

• Top: OEE gauge (target 75%, current 68% in amber)

• Middle: Line-by-line output vs. plan (bar chart)

• Bottom: Trend of deviations (spike last week triggers investigation)

• Drill-down: Click line for detailed downtime analysis

Quality Metrics Dashboard

Target User: Head of Quality (Tactical/Strategic)

Key Metrics:

• OOS (Out of Specification) rate: Trending down?

• Deviation backlog: Aging analysis

• CAPA effectiveness: Are issues being resolved?

• Audit findings: Trend over time

• Batch release time: Quality impact on business

Design Considerations:

• Aging analysis of open deviations (>30 days highlighted)

• CAPA effectiveness: % of repeat issues (should be low)

• Audit findings: Categorized by severity, with drill-down to action plans

Clinical Trial Dashboard

Target User: VP Clinical Development (Strategic)

Key Metrics:

• Enrollment rate: Vs. plan by study

• Site performance: Which sites are lagging?

• Data quality: Query rate, protocol deviations

• Timeline status: On track for milestones?

• Budget vs. actual: Cost control

Design:

• Heatmap of site performance (enrollment vs. quality)

• Gantt-style timeline with milestone status

• Traffic light status per study

• Geographic map of active sites

Supply Chain Dashboard

Target User: Supply Chain Director (Tactical)

Key Metrics:

• Inventory levels: Days of supply by SKU

• Stockout risk: Predictive, highlight high-risk items

• Supplier performance: On-time delivery, quality

• Production schedule: Upcoming batches vs. inventory need

• Expiry risk: Products nearing expiry

Design:

• Inventory heatmap: Green (healthy), amber (monitor), red (action needed)

• Supplier scorecard with trends

• Predictive alerts for potential stockouts

Technical Implementation

Tool Selection:

BI Platforms (Power BI, Tableau, Qlik):

• Pros: Powerful, flexible, good visualizations

• Cons: Licensing costs, potential performance issues with large datasets

• Best for: Strategic and tactical dashboards

Custom Development (React + D3.js, Python Dash):

• Pros: Full control, can optimize for performance

• Cons: Development time, maintenance burden

• Best for: Operational real-time dashboards, unique requirements

ERP/MES Built-in:

• Pros: Integrated, no extra systems

• Cons: Limited flexibility, often poor UX

• Best for: Standard operational reports

At Hyperbolic, we typically choose Power BI for executive/tactical and custom React for operational real-time dashboards.

Data Pipeline:

A dashboard is only as good as its data. A robust data pipeline is critical:

1. Data Extraction: From source systems (MES, ERP, LIMS, etc.)

2. Transformation: Cleansing, aggregation, calculations

3. Loading: To data warehouse or dashboard database

4. Refresh Frequency: Real-time, hourly, daily? Based on use case

GxP Considerations:

• Audit trail of data transformations

• Validation of calculations

• Access control on dashboards

• Data integrity (ALCOA+)

Common Design Mistakes

Mistake 1: Chart Junk

3D charts, excessive colors, animations, gradients, etc. add no value and distract from the data.

Better: Flat, simple designs. Edward Tufte's principle: Maximize data-ink ratio.

Mistake 2: Wrong Chart Type

Pie Charts: Generally avoid. People are bad at comparing angles. Use bar charts.

Line Charts: For trends over time. Not for categorical comparison.

Bar Charts: For comparison among categories.

Heatmaps: For two-dimensional patterns (e.g., site performance vs. time).

Mistake 3: No Actionability

A dashboard that just shows numbers without guiding to action is a waste of time.

Better:

• Include recommended actions on deviations

• Links to detailed reports or investigation tools

• Clear escalation paths

Mistake 4: Static vs. Interactive

PDF reports distributed monthly are not dashboards. Real business benefit comes from interactive, frequently updated dashboards.

Better: Web-based, auto-refreshing, with drill-down capabilities.

Case: Production Performance Transformation

Before: Plant produced monthly Excel reports with 50+ metrics. Took 2 weeks to compile, was outdated at distribution, no one used them.

Intervention: We designed layers of dashboards:

• Executive: 5 key metrics, monthly trends, RAG status

• Manager: Line-by-line performance, daily granularity, drill-down to issues

• Supervisor: Real-time line status, current batch progress, alerts

Technology: Power BI connected to MES via automated ETL pipeline. Auto-refresh every 15 minutes for operational, daily for tactical.

Results:

• Adoption: 100% of managers check the dashboard daily (vs. <20% read old reports)

• Decision Speed: Issues identified and addressed the same day vs. weeks later

• Performance Improvement: OEE increased by 8% over 6 months (visibility drives accountability)

User Adoption Strategies

Even the best dashboard fails if no one uses it.

Pre-launch:

• Involve Users in Design: Workshops with actual users, not just management

• Prototype Early: Show mockups, get feedback, iterate

• Communicate Value: What's in it for them?

Launch:

• Training: Hands-on sessions, not just "here's the link"

• Champions: Identify power users to evangelize

• Support: Dedicated support in first weeks

Post-launch:

• Usage Analytics: Track who's using what, identify gaps

• Regular Reviews: Quarterly check-ins with users for feedback

• Continuous Improvement: Dashboard is never "done"

Conclusion

A good dashboard transforms how decisions are made: faster, data-driven, more confident. But it requires:

1. Start with the decision, not the data

2. Design for the user, not for yourself

3. Simplicity over comprehensiveness

4. Actionability as a central principle

5. Continuous iteration based on feedback

At Hyperbolic, we design dashboards that are actually used. We combine UX expertise, data engineering, and domain understanding to deliver insights-driven tools.

Contact us to discuss how we can transform your data into actionable insights.