← Back to all work

Front-end

User Data Analytics Platform

Front end and tracking layer for an in-house analytics platform deployed across four products and hundreds of branded sites.

Cropped traffic flow, realtime audience and trend views with demo data and product identifiers removed.
Year
2022
Role
Front-end Engineer, front end and tracking layer (Neutec Ltd.)
Status
Internal company system
  • React
  • styled-components
  • Redux
  • GraphQL
  • ECharts
  • Jest
  • Vite
Products using it
4
Plus several hundred branded sites
Events tracked a day
Millions
Requests on a typical page
3 to 4 → 1
Data layer moved from REST to GraphQL

Overview

Built the front end and tracking layer of an in-house alternative to Google Analytics that keeps user data within the company for security. It was deployed across 4 products and several hundred branded sites, tracking millions of events a day.

  • Built an attribute-based event-tracking SDK that lets product teams track clicks by adding an HTML attribute to a link or button. It also supports page views, scrolling and custom events, with shared handling for batched uploads, payload-size thresholds and pausing and resuming after collection errors.
  • Built the ECharts reporting dashboards that product managers, marketing and operations used to compare clicks, dwell time, navigation paths, response times, devices and countries.
  • Worked with the back-end team to migrate complex reporting queries from REST to GraphQL, consolidating a typical page’s 3 to 4 requests into one.

The SDK listens for clicks at page level and walks from the clicked text or icon to its parent tracking attribute, giving products a consistent way to mark events. Scrolling uses a 500 ms debounce, and events are sent in batches every 30 seconds by default, with a configurable interval. Real-time reports use polling and cancel requests and clear timers when conditions change or a view unmounts.

Product managers, marketing and operations used the dashboards to optimise marketing and retention.

The visuals use demo data with product identifiers removed.