Nikhil Gupta, founder of LimeChat, aimed to revolutionize e-commerce connections. But his journey wasn’t without challenges. A major scalability issue forced LimeChat to confront the reality of rapid growth. In this article, Nikhil shares how his team turned a potential disaster into a growth opportunity while scaling product management at LimeChat.
Highlights
Building Products is HARD. Scaling Products is HARDER.
After stepping off my flight in San Francisco, I was greeted not by the excitement of a new year but by a barrage of alarming notifications. LimeChat was down globally. Requesting a large report had frozen our systems, a reality far from the optimistic scenario I had envisioned for our technology and team. This stark wake-up call plunged us into the hard truths about scalability, marking the beginning of a journey from breaking point to a breakthrough in scaling product management.
Our thinking in building LimeChat is simple: Value is everything for the customer. Features help extract value. The more features, the more value. That is how we will win the market.
We focused solely on launching feature after feature for the first three years. The development speed was astonishing, as we prioritized rapid deployment over investing in thorough testing and a quality assurance (QA) team, which we believed would only slow us down or incur extra expenses.
My message to the team was clear:
"Let's take it live and stay alert. In the worst case, we can revert."
Pros and Cons of Rapid Development
Pros:
- A small team can accomplish a great deal quickly
- A competitive edge and the ability to win more customers
- High momentum within the engineering team
- Faster learning of juniors
Cons:
- Accumulation of technical debt
- Features break as systems scale
- New features take priority over maintenance
- More features become exponentially harder to maintain
Slowly, the downsides started to outweigh the benefits. Scaling product management became crucial to navigate these challenges.
Observations
Over 50% of our development time was consumed by maintenance. In hindsight, we had developed many features that should have been declined. These features, appealing to only a fringe group of clients, proved time-consuming due to less testing on edge cases, limited team focus, and a higher likelihood of issues.
Most concerning was the realization that our systems were failing under increased demand. Even basic dashboard reports took TWO days to refresh as databases hit capacity. Messaging services experienced downtime due to festivals and seasonal spikes in the DTC ecosystem. Certain features lack optimization and consume excessive resources, leading to skyrocketing infrastructure costs.
Simple queries such as counting rows were timing out! How can any application run with basic functionality and APIs refusing to run?
How We Responded
In the past year, we’ve made several significant changes:
Improved Feature Planning. We’re now more critical about the long-term value of new features, leading to a decrease in unnecessary rework.
Shared Testing Responsibility. Testing has become shared across the team, with a new emphasis on unit testing and addressing technical debt. This shift towards a test-driven culture has been a priority in 2023.
Architectural Overhaul. We’ve revisited and refined our core architecture to identify and implement scalable solutions, including adopting microservices and new infrastructure components to support long-term growth. Scaling product management was vital to achieving these improvements. Some examples include:
- Scaled analytics by shifting to Clickhouse
- Consolidated all messaging in a central microservice
- Passed internal APIs through a central authentication system to improve speed and security
- Optimized writing of new integrations with platform redesign
These changes required substantial time and effort and presented challenges we had yet to anticipate. The shift in focus led to a temporary slowdown in our development momentum, but we recognized the importance of these changes for providing our clients with the best possible experience.
Outcomes
Now in 2024, the improvements are evident. Our analytics load nearly instantly, a change that feels like magic. The query that brought our systems down in 2023 now runs within 0.3 seconds!
Introducing thorough testing has boosted our team’s confidence in deploying code more aggressively. In places with the redesign, we have been able to ship code in nearly 20% less time with 70% less rework!
By optimizing our queries and managing load spikes more effectively, we’ve cut infrastructure costs by over 30%, primarily by downsizing our core database. These changes were crucial for scaling product management efficiently.
Our time to build new integrations has decreased by over 50%, and the complexity of the work has reduced significantly, allowing junior developers to contribute to the codebase.
After our optimizations, our customer NPS jumped by 22 basis points in a span of 6 months, exceeding our expectations. Our lesson was clear: It’s not new features that the client needs to feel happy; they need their day-to-day to run as smoothly as possible.
This experience has also heightened our team’s appreciation for good architectural design, influencing the incorporation of better design practices right from the start of new projects.
What’s Left
There’s still much room for optimization. As we aim to scale by 10x, we’re mindful of the impact of emerging AI technologies on system development and are preparing to integrate these advancements to stay ahead. Emphasizing a clean core architecture positions us to be AI-ready and represents a sound investment for the future. Scaling product management is a continuous journey.
With the rapid pace of technological change, we’re excited to see what the year brings and how we can continue to innovate and scale effectively.
Balancing Speed with Stability in LimeChat’s Success
Nikhil’s experience shows the importance of balancing speed with stability. LimeChat improved performance by focusing on better architecture and thoughtful feature planning. The key lesson? Success isn’t just about adding features; it’s about ensuring smooth operations. This story reminds us that sometimes, the right changes can lead to significant progress.
Highlights
Read the Chinese version here.