The internal case study conducted by Amazon has sparked both interest and skepticism within the developer community,  and here at Product Science. The case study showed, for their Prime Video application at least, that transitioning a backend system from microservices to a monolithic architecture can significantly reduce costs – by a whopping 90% in this case. The irony is not lost on other customers of AWS, whose own guidance recommends microservices, due to their scalability and reliability.

While this controversy is interesting, it’s understandable that companies focus on optimizing their backend systems, especially when looking to increase profitability. As Amazon’s case study demonstrates, backend improvements are crucial to reducing costs and improving efficiency. Indeed, this is a message we see time and again. Heard of data-driven purchasing, or “cloud repatriation”?

However, it's essential to recognize another significant avenue for boosting profits that is often overlooked: optimizing client-side performance. In this article, we explore the untapped potential of client-side optimization and its surprising impact on a company's revenue. We then introduce the PS Tool and explain how easily it optimizes mobile performance. Finally, we examine the Amazon Prime Video mobile apps for potential performance optimization opportunities.

The Client-Side vs. Server-Side Performance Dilemma

Before discussing the significance and ease of client-side optimization, let's acknowledge the fundamental differences between backend (server-side) and frontend (client-side) systems. 

Backend performance improvements directly impact operating costs and clearly correlate with a company's bottom line. Since backend systems can be tightly controlled, optimizing them often yields predictable outcomes in terms of costs and performance. Amazon Prime Video's case study provides evidence of significant cost reductions achieved through backend optimization.

On the other hand, frontend systems, particularly mobile apps, run on various hardware and operating systems, catering to diverse user devices. These systems demand immediate responsiveness to user interactions and frequent display updates (once every 17ms for 60fps), which necessitate complex multi-threading, making optimizing them inherently challenging without the appropriate tools. Moreover, while frontend performance improvements don't directly impact operating costs, they contribute to revenue opportunities by enhancing user experiences, leading to longer sessions and improved conversion rates. Unfortunately, obtaining concrete data on the expected impact of such improvements can be challenging. Coupled with the perceived difficulty of making such improvements, frontend performance optimization is often overlooked as a priority. However, this article challenges this notion head-on.

First, we’ll highlight a fraction of the available evidence demonstrating front-end performance's huge impact on revenue. Then we’ll explore the ease of frontend optimization, showcasing our new PS Tool.

The Revenue Potential of Client-Side Performance

In today's digital landscape, a company's success heavily relies on its ability to engage users effectively, understand their behaviors, and deliver optimal mobile performance. The interplay between these factors directly impacts a company's revenue potential and overall business success.

Numerous studies and research have shed light on the correlation between user engagement, user behavior, mobile performance, and overall business success.

1) One study conducted by Google emphasizes the importance of page speed in relation to user engagement and revenue on mobile devices. The study highlights how slow loading times can significantly impact user experience, leading to decreased engagement and potentially reduced revenue. 

2) Deloitte's report focuses on web performance and provides compelling insights. The report reveals that even a minor improvement of 0.1 seconds in load time can lead to an 8% increase in conversions and a 10% boost in expenditure for retail sites. 

3) Research conducted by Akamai, a leading content delivery network, further supports the connection between mobile performance and revenue. Notable examples from the research:

  • Fanatics nearly doubled mobile conversions by reducing its median page load time by two seconds.
  • Staples witnessed a 10% increase in conversions by improving page load times. 

The impact of mobile performance on user behavior is equally vital. A mobile app user survey highlighted in the research revealed that when faced with a slow-performing app, 48% of users completely uninstalled it from their devices, 33% ceased using it, and 32% actively sought alternatives. In today's competitive market, users have an abundance of choices and expect seamless digital experiences. A single instance of poor performance can have detrimental consequences, with 57% of people stating that brands have only one opportunity to impress them. Failure to deliver a satisfactory digital service can result in users abandoning a company's offerings and potentially choosing a competitor.

Frontend Performance Optimization – The Present Situation

Frontend clients, particularly mobile apps, are characterized by their dual requirements of continuous user interactivity and frequent screen updates. To meet these demands, they often employ multi-threading, resulting in inherent complexity. This complexity leads to unwanted delays in computation and user interactivity while one thread waits for a resource locked by another. Such idle time represents a waste of valuable CPU resources. Fortunately, these delays present an opportunity for significant improvements in mobile performance optimization. Often, minor system design changes can completely eliminate these bottlenecks.

However, existing developer tools, such as profilers and trace tools, currently fall short of effectively identifying the causes of these latencies. They both overwhelm an Engineer with data, yet at the same time, fail to show crucial connections between program threads. These delays go undiagnosed, leading to much Engineering time spent trying out multiple solutions and potentially even rewriting the app from scratch – often with disappointing results.

PS Tool – The Future of Performance Optimization

Our PS Tool addresses these challenges by offering two key solutions. First, it streamlines the information provided, delivering relevant and insightful data while filtering out unnecessary details. Secondly, it establishes connections between threads, revealing the interplay between functions in different threads.

Here’s how it operates: A user records a user flow in their app using PS Tool on their mobile device. The tool then uploads the data into PS service, which can be accessed via a web portal. The web portal presents synchronized visualizations of the thread call stacks and app screen recordings.

By selecting the UI rendering thread and scrubbing to the start of the user flow, the tool's intelligent technology highlights the execution path, showcasing the sequence of functions involved. The user flow is comprehensively linked across all methods and threads involved in the execution, with idle time clearly distinguished.

The PS Tool's ability to provide visual insights for the immediate identification of performance issues in multi-threaded environments sets it apart from traditional tools. Its user-friendly interface and intelligent data curation empower not just software developers but also DevOps, Product, and UX Designers to investigate performance issues and identify opportunities for improvement. Consequently, the PS tool emerges as a transformative asset, enabling businesses to pinpoint latency causes in client-side mobile applications swiftly.

Transitioning from Intuition to Data-Driven Decision-Making:

Businesses can embrace data-driven decision-making to optimize client-side performance by leveraging the PS Tool. The tool eliminates the need for time-consuming exploration of problems and reliance on speculative, "best guess" engineering fixes, resulting in significant cost savings. Engineering leads to gaining access to the vital information needed to formulate actionable plans for resolving delays with predictable costs and outcomes. Furthermore, precise data on the root causes of performance issues fosters improved company-wide alignment and support across teams, paving the way for comprehensive solutions. With its ability to expose the intricate nature of multi-threading, the tool also enables more aggressive parallelization and optimal utilization of CPU resources.

An Example – Optimizing the Prime Video App

At Product Science, we have successfully identified performance optimization possibilities in the Prime Video apps, even without direct access to the app source code. We have made noteworthy observations based on our extensive expertise in assisting clients with app optimization and analyzing the Prime Video apps themselves.

To assess the current state of the Amazon Prime Video apps, we evaluated the most recent production versions on comparable iOS and Android devices: the iPhone 12 and Samsung Galaxy A52. 

Under optimal networking conditions, the user interface (UI) exhibited consistency across both platforms. However, we noticed some disparities in the delays encountered during various user flows between Android and iOS. One particular delay caught our attention when switching categories. On iOS, transitioning from the "All" category to "Movies" or "TV Shows" resulted in a notable lag in the top section that displays the featured preview. In contrast, the Android version demonstrated quicker loading time (almost instantly) for the main section of the screen under similar network conditions. 

Drawing on our experience and insights, we hypothesize that this delay on iOS stems from a lack of pre-loaded content for this specific user flow. This could be verified using PS Tool. 

Addressing this issue presents an excellent opportunity for performance improvement, as we estimate it can reduce approximately 800ms of latency, which accounts for 40% of the total rendering time. We can simulate the anticipated improvement in screen rendering as follows:

Simulated gain from pre-loading for Prime Video

Conclusion

While optimizing backend systems is essential for cost reduction, it is equally important for companies to recognize the revenue potential associated with frontend performance. Numerous studies have established a strong correlation between app performance and revenue, underscoring the need for prioritizing mobile performance optimization. 

By leveraging the power of the PS Tool and directing efforts toward front-end optimization, businesses can unlock significant revenue growth by delivering enhanced user experiences, driving improved engagement, and achieving higher conversion rates. Adopting the PS Tool as a business solution empowers organizations to embrace data-driven decision-making, leading to more predictable engineering costs and performance gains.

It's time for companies to recognize the untapped revenue potential of optimizing client-side mobile performance and seize this opportunity to drive their business forward.

It's time for companies to recognize the untapped revenue potential of optimizing client-side mobile performance and seize this opportunity to drive their business forward.

If you’re interested in tackling challenges like this, join our team! Time is humanity’s most valuable non-renewable resource. Our mission is to help all people in the world stop experiencing delays from software inefficiency.

Request a Product Science demo today