Enterprise Search: Decoupled Rules Engine

Tech Lead / Architect 2024 2-3 months 3 Devs (Lead + 2) 3 min read

Enabled instant content updates via CMS, bypassing CI/CD bottleneck and driving measurable lift in user conversions (NABs).

Overview

Architected a 'Rules Engine' that decoupled business data from the application logic. By moving the mapping of search terms and component triggers into a CDN-hosted JSON configuration, the Angular frontend became a dynamic delivery vehicle. This allowed the Product team to update 'upsell' content and run A/B tests instantly without waiting for code deployments.

Gallery

Problem

Our Jenkins CI/CD pipeline was a massive bottleneck; a single UI change required a multi-stage gauntlet of unit tests, E2E tests, and security scans that could take days to clear. This made it impossible for the Product team to iterate on high-traffic search results or react to market trends in real-time.

Constraints

  • Backend API was not yet ready to return specialized marketing metadata.
  • Must maintain performance and light payload.
  • Required 100% WCAG accessibility compliance and custom analytics integration.
  • Solution needed to be a flexible 'shim' for future backend integration.

Approach

Instead of hard-coding search result variations into the app, we built an Angular service that acted as a matching engine. This service fetched a remote JSON file from the CDN on initialization. When a user performed a search, the engine performed a direct match between the query and the JSON keys to determine which high-impact components (FAQs, CTA banners, or products) to render. I developed the initial POC and established the service architecture, then led a small team to build out the modular component library.

Key Decisions

Leverage Bloomreach WCM for JSON-as-a-Service

Reasoning:

By hosting the 'Rules Engine' in the CMS, we bypassed the Jenkins pipeline entirely. Content editors could update search mappings and deploy to the CDN in minutes, providing total business autonomy.

Architected a Flexible Data Interface

Reasoning:

Since the API roadmap was still in flux, I designed the frontend service to consume a normalized JSON structure. This acted as a 'shim,' allowing us to deliver business value immediately. I prioritized making the matching logic modular so that when the API eventually came online, we would only need to update the data-fetcher service rather than the entire component architecture.

Parallelized Development Lifecycle

Reasoning:

After establishing the core service pattern and accessibility standards, I broke the project into modular tasks, allowing the team to build components and tracking services simultaneously to meet the 8-week launch window.

Conditional Asset Loading

Reasoning:

To protect page performance, I ensured that 'upsell' assets (images/icons) only triggered loading after a successful query match was confirmed by the service.

Tech Stack

  • Angular & TypeScript
  • Decoupled Architecture
  • Bloomreach WCM
  • System Design
  • Custom Analytics Integration

Result & Impact

  • Reduced from 48 hours to minutes
    Time-to-Market
  • 100% (No engineering required for content updates)
    Product Autonomy
  • Significant lift in NABs through rapid A/B testing
    Business Impact

This architecture bridged the gap between a lagging API roadmap and an urgent business need. By building a stable frontend interface for our JSON 'stub,' we decoupled the UI development from the backend's timeline. This allowed us to launch a high-impact feature months early, with the understanding that the system was built to be easily integrated with a live API once the backend schema was finalized.

Learnings

  • Engineered a way to resolve cross-team dependencies (API/CMS/Design) to clear a technical bottleneck.
  • Demonstrated how a POC can establish standards that allow a small team to work effectively in parallel.
  • Solved for the immediate deployment friction while keeping long-term maintenance in mind.

The Impact

By moving the search criteria into the CMS as a CDN-hosted JSON file, we decoupled the search data from the application code. This allowed the front-end and search query service to perform real-time comparisons against live data, giving the Product team the ability to update additional search results content in minutes. This bypassed the days-long deployment cycle, enabled real-time A/B testing, and ultimately helped us launch the feature ahead of schedule.

All projects