Making Web Analytics Understandable
The Problem
Google Analytics is powerful, but for non-technical website owners, understanding what actually changed often requires navigating complex reports, metrics, filters, and dimensions.
The challenge was to turn this complexity into something a business owner could understand in seconds, without sacrificing statistical accuracy or trust.

The Approach
I designed PlotMe around four simple questions:
What changed? → Why does it matter? → Is it significant? → What should I do next?
Instead of presenting users with another dashboard full of charts, Plot me translates analytics data into plain-English insights, while keeping the underlying mathematics transparent.
Every insight can be opened through “Show the Maths”, allowing users to see the underlying sample sizes, confidence intervals, statistical tests, and calculations.
Key Design Principles
1. Narrative over dashboards
Replace metric-heavy screens with concise explanations of what changed and where to focus.
2. Trust over AI-generated interpretation
Numbers never originate from an LLM. Statistical conclusions are generated deterministically from the underlying data.
3. Explainability by default
Every significant claim has a transparent mathematical explanation rather than asking users to simply trust the system.
4. Context over correlation
The interface deliberately avoids implying causation. Insights use language such as “accounts for” and “coincides with” when describing relationships in the data.

The Statistical Engine
Under the interface sits a deterministic TypeScript analytics engine that evaluates:
Historical baselines and Z-scores
Weekday seasonality
Conversion-rate changes
Channel distribution shifts
Anomaly detection
Confidence intervals
Benjamini–Hochberg false-discovery correction
This allowed the product to communicate insights without introducing unsupported conclusions.
Visualization System
Rather than relying on large charting libraries, I built a lightweight visualization system for:
Distribution charts
Range comparisons
Proportional bars
Heatmaps
Statistical comparisons
The result was a responsive visualization layer with semantic markup and accessibility built into the components.
Experience Design
The product was designed to feel less like an analytics console and more like a guided conversation with your data.
The landing experience uses interactive scrollytelling and WebGL motion to introduce the product concept, while the application progressively reveals the most important information instead of overwhelming users with every available metric.
Outcome
Under 2 min
Time for a new user to connect a website and reach their first digest.
154 tests
Statistical and product logic validated through automated regression testing.
~41 KB
Initial application entry bundle.
0 KB charting dependency
Custom SVG visualizations instead of heavyweight chart libraries.
100% explainability
Qualitative testing showed non-technical users could explain what changed and understand the confidence associated with an insight.

The Takeaway
Plotme wasn't about creating another analytics dashboard.
It was about designing a trust layer between complex data and human decision-making.
Good analytics shouldn't make people better at reading dashboards. It should make them better at understanding what their data means.


