Delbridge Solutions
Delbridge Solutions

Engineering a Faster, Leaner E-Commerce Experience with MongoDB Atlas

Talk to Delbridge

How One E-Commerce Platform Optimized Atlas to Boost Speed and Cut Costs

The Challenge: Sluggish Search, Soaring Costs, and a Platform Under Pressure

When customers across Europe logged into their favorite online electronics store, they weren’t expecting delays. But instead of a seamless shopping experience, they faced sluggish searches, lagging filters, and slow-loading pages. Finding the right product became frustrating. Conversions dropped. Complaints rose. 

Behind the scenes, the platform was straining under the weight of rapid growth. 

  • Infrastructure costs were ballooning 
  • Servers were overloaded 
  • Search performance was degrading fast 

This mid-sized, multi-tenant e-commerce provider served over 250 business clients across the UK, France, Germany, and the Netherlands. With millions of products and 13 supported languages, performance wasn’t just a technical concern, it was directly tied to revenue. 

 

The company’s newer platform, Prisma 2, was particularly vulnerable. Despite using MongoDB Atlas Search, the configuration was flawed. Indexes were bloated. Pagination relied on skip/limit, degrading user experience with every scroll. Read costs surged. Memory spikes triggered node restarts and replica elections. 

The platform wasn’t broken, but it was stuck in first gear.  That’s when Delbridge stepped in.

A Precision Approach to Performance at Scale

Delbridge began with a comprehensive audit of the MongoDB Atlas environment. The findings told a familiar story: a modern tech stack with plenty of potential but misconfigured, under-optimized, and driving up costs. 

 

Key issues identified: 

  • Multilingual search was underpowered, relying on default analyzers that ignored language differences and lacked synonym support 
  • Pagination used skip/limit, generating increasingly expensive queries as users scrolled 
  • Indexes were redundant or misaligned, failing to reflect actual query patterns 
  • COLLSCAN operations were rampant, pushing CPU to 100% and memory usage to 84% 
  • Maintenance scripts and bulk writes ran during peak hours, compounding performance issues 

The diagnosis was clear: the platform didn’t need a rebuild, it needed a rethink. 

The Fix: Five Smart Improvements to Unleash Atlas Performance

1. Rebuilding Atlas Search for Multilingual Relevance

Delbridge restructured the Atlas Search index to support: 

  • Multilingual analyzers 
  • Dynamic synonym collections, auto-generated using Node.js and AI models (OpenAI & Grok) 
  • Special character filtering for consistent indexing 

Example synonym document: 

2. Cursor-Based Pagination

Skip/limit pagination was replaced with cursor-based logic using _id, createdAt, and searchAfter. 
This dramatically improved response times and reduced query load. 

Example: 

3. Index Optimization Using ESR Strategy

Indexes were redesigned using Equality, Sort, Range (ESR) principles: 

  • Prioritized high-cardinality fields (like productId) 
  • Auto-generated based on schema and live API traffic 
  • Monitored using native index stats 

Command to identify index usage: 

4. Storage & Cost Efficiency

To reduce cloud spend and improve resource usage, the team: 

  • Reviewed and trimmed pre-provisioned IOPS 
  • Compacted collections to reclaim disk space 
  • Dropped temporary structures no longer in use 

Example cleanup commands: 

5. Real-Time Monitoring & Search Node Isolation

To improve observability and reduce performance risk, Delbridge: 

  • Introduced dedicated search nodes to isolate workloads 
  • Enabled real-time tracking of index size, latency, and memory 
  • Updated drivers and aligned with the latest cluster version 

The Impact: Faster, Leaner, and AI-Ready 

Once the improvements were deployed, the results were immediate: 

Platform Outcomes at a Glance:

  • Search performance improved by up to 5× 
  • Pagination delays dropped from 18 seconds to near-instant
  • Cloud infrastructure costs have reduced significantly 
  • Multilingual UX enhanced with synonym-aware search 
  • Query load balanced across the system, eliminating outages 

And most importantly, the platform is now AI-ready—capable of powering smart recommendations, predictive search, and personalized ranking models without a full-stack rebuild. 

Why It Matters

For product leaders and founders, this isn’t just a technical win, it’s proof that MongoDB Atlas, when fully optimized, can become a powerful engine for scale. 

With the right partner, a sluggish stack isn’t a setback, it’s an opportunity. 
You don’t need to rebuild. 
You need to reimagine what your database can do. 

Delbridge can help you get there.