Optimizing an Email Retargeting Platform for eCommerce Cart Abandonment Recovery

The Scenario

An email retargeting platform founded in 2011 set out to solve one of the most persistent problems in eCommerce: shopping cart abandonment. The platform — built by practitioners who had personally encountered the problem while working at a SaaS company — gave eCommerce stores of all sizes the ability to recover lost potential revenue by sending customized, cart-specific email campaigns to shoppers who left without completing a purchase.

The average recovery rate for customers using the platform sits between 10% and 20%. Depending on a store's monthly checkout volume, average order value, and abandonment rate, that translates to anywhere from a few thousand to several hundred thousand dollars in recovered revenue per month — a meaningful return for a relatively lightweight intervention.

Over time, the platform attracted high-profile retail and consumer brands including MasterCard, Hallmark, Dreamworld, AMF Bowling, Liftopia, and VTech. But rapid growth exposed serious technical debt that was constraining the platform's ability to scale.

The Challenge

The platform's engineering team faced a set of compounding problems common to early-stage MarTech startups that gain traction faster than their infrastructure can keep pace with:

• Performance issues in data reporting — slow queries and lag in report rendering degraded the user experience for customers who relied on near-real-time visibility into campaign results.
• Limited technical resources — the internal team lacked the capacity to simultaneously maintain stability, develop new features, and address the underlying architectural issues.
• Deployment friction — releasing new features was slow and error-prone, with deployments taking anywhere from 30 to 60 minutes.
• No dedicated testing environment — testing was coupled to local development environments, making it difficult to catch regressions before they reached production.
• Manual billing operations — account billing required hands-on work that didn't scale with the customer base.

The strategic decision made early in the process was significant: rather than rebuilding the platform from scratch, it was more financially viable and technically feasible to refactor and optimize the existing codebase. A greenfield rebuild would have introduced risk and cost that the business couldn't absorb at that stage.

The Approach

The work proceeded across several parallel tracks:

Code Refactoring

A systematic refactor of the codebase improved code quality, reduced complexity, and made the platform significantly easier to test and extend. The refactored codebase lowered the risk associated with adding new features and made ongoing maintenance more tractable.

Platform Optimization

Stability improvements were made alongside the refactor, reducing the frequency of unexpected behaviour under load.

Database Schema Redesign

Changes to the database schema resulted in faster queries and more efficient data retrieval. The direct user-facing impact was the elimination of lag in report display — users could view reports instantly rather than waiting for slow queries to resolve.

Continuous Integration (CI) Server

A CI server was configured to create a dedicated testing environment, independent from local development setups. This allowed bugs to be caught earlier in the development cycle, protected existing platform logic from regressions, and ensured new code was compatible with the existing codebase before reaching production.

Automated Deployment

Deployment was automated, reducing release time from 30–60 minutes to a matter of seconds (for smaller changes). This eliminated repetitive manual steps, reduced human error in deployments, and enabled a higher frequency of feature releases.

Infrastructure Redesign and Hosting Migration

The platform's infrastructure was redesigned and migrated to a new hosting provider with a better-organized staging environment. The most consequential change here was the introduction of database replication, which created redundancy: if the master database were to go down, a replica could take over, increasing overall platform availability.

Automated Billing

The billing process was automated, eliminating the manual account management work that had previously been required for each billing cycle.

Implementation Considerations

Several factors shape whether a refactor-and-optimize approach succeeds in a scenario like this:

Extensibility over purity. When the goal is to stabilize a live platform without halting feature development, a complete rewrite is rarely the right call. Incremental refactoring — applied strategically to the highest-impact areas — produces meaningful gains with lower risk and cost.

Schema changes require care. Database schema changes on a live platform carry real risk. The benefit here — faster queries and instant report rendering — only materializes if the migration is executed without data loss or downtime. This is an area where having prior experience with analytics-adjacent platforms is particularly valuable.

CI and deployment automation compound over time. The immediate benefit of reducing deployment time is measurable (minutes to seconds), but the longer-term benefit is cultural: teams deploy more frequently, catch bugs earlier, and accumulate less fear around releases. These changes make a platform easier to evolve.

Database replication is a threshold improvement. Moving from a single-database setup to a replicated one isn't just a reliability improvement — it changes the risk profile of the entire operation. The platform can sustain a primary database failure without going dark.

Outcomes

By the time these improvements had been implemented and the platform had matured, the platform's customers had collectively generated over $56 million in cumulative revenue through recovered cart abandonment.

The technical improvements positioned the platform to serve an increasingly high-profile customer base without the performance and reliability constraints that had limited earlier growth. The combination of schema optimization, automated deployment, CI infrastructure, and database replication transformed what had been a brittle early-stage product into a scalable, maintainable MarTech platform.

For eCommerce platforms in a similar position — strong product-market fit, growing customer base, but mounting technical debt — this case illustrates that a well-executed optimization program can deliver the scalability of a rebuild without the timeline and financial exposure of starting over.