Quebec SaaS Company

Scaling a Real-Time Platform Without Losing Responsiveness

We documented the technical uncertainties, experiments, and knowledge gained that led to a successful SR&ED claim.

Recovered

Company At A Glance

Business Type

SaaS Software Company

SR&ED Category

Experimental Development

Team Size

24

Scope

Real-Time Collaboration, Scaling, Latency Reduction

Province

Quebec

The Challenge

The company operated a real-time collaboration platform where users could work together through shared boards, live cursors, presence indicators, and instant updates. The platform worked well at normal traffic levels, but as usage increased, the system began experiencing delayed updates, inconsistent cursor movement, dropped events, and performance spikes during busy sessions. 

The challenge was that standard scaling approaches did not solve the issue. Adding more servers, replicas, and caching improved some areas but did not reliably reduce latency or maintain consistent state across users. The team had to determine whether the platform architecture could be redesigned to support high volumes of concurrent activity while preserving real-time responsiveness and session accuracy. 

What We Doucumented

01

Uncertainty

Uncertainty

The uncertainty was whether the platform could maintain low-latency real-time updates while supporting a much larger number of concurrent users across different regions and usage patterns.

02

Iterations

Iterations

The team tested different sharding strategies, message-routing models, in-memory coordination methods, batching approaches, compression techniques, and load-testing scenarios. Some designs reduced latency but created state consistency problems. Others improved reliability but increased processing load or delayed updates during burst traffic. 

03

Knowledge Gained

Knowledge Gained

The company gained new architectural knowledge about how to structure real-time data flow, coordinate active sessions, reduce unnecessary update traffic, and maintain consistent user state under high concurrency. 

Results

Why This Qualified

  • Required experimentation beyond standard scaling, caching, and server replication
  • Existing architecture could not maintain real-time performance under higher concurrency
  • Latency, state consistency, and message-routing behaviour changed unpredictably under load
  • Multiple architecture, sharding, and coordination models were tested and measured
  • New knowledge was gained about scaling real-time collaboration workloads