Getting started with Rockset was remarkably straightforward and effortless. We were already up and going within just a few hours. Jeremy Evans, co-founder and Chief Technology Officer at Savvy.
At present, we possess a multitude of responsibilities regarding knowledge acquisition.
Our clients include online shopping platforms similar to Amazon and eBay. Customers rely on our cloud-native platform to effortlessly build no-code interactive experiences, mirroring engaging formats like video quizzes, calculators, and listicles, without requiring developers. With our analytics dashboard, firms can monitor and measure the success of their customized training programs for customers in real-time.
As you oversee massive conversion flows that involve hundreds of travelers daily, robust analytics are crucial. Our clients require granular insights into each step of their conversion funnel, allowing them to pinpoint areas for optimization and inform data-driven decisions. By integrating analytics capabilities within Savvy’s platform, we empower businesses to make strategic enhancements without relying on in-house developers or external vendors.
Notwithstanding, our initial challenge lay in consistently providing valuable and timely insights, as our native platform excelled at ingesting data but struggled to effectively analyze and report on it.
To sustain growth and ensure uninterrupted service, we sought an extremely high-performing, seamless solution that could be easily implemented.
Squaring the (No)SQL circle
We built Savvy leveraging Google’s Firebase app development and hosting capabilities. With Firebase’s scalable architecture and flexible schema-less design, we were able to rapidly accelerate our progress. With remarkable speed, our embedded flows seamlessly integrate into customers’ websites, loading in as little as 300 milliseconds on average. They love that real-time efficiency.
We successfully monitored and recorded the website activities of individual guest users without any hiccups. Data from various interactive sessions is streamed into Firebase’s NoSQL cloud-based database, where complex structures comprising multiple nested objects and arrays are stored. Providing our clients with a real-time inventory of current guests, along with a comprehensive record of their interactions, was not only feasible but also straightforward to accomplish.
As issues arose swiftly, our customers demanded swift access to power their records, effortlessly filtering data and gaining insights into visitor numbers over time, with the ability to drill down into referrer website analytics.
While using Firebase’s built-in filters is a viable solution, we must consider the potential limitations of relying solely on these tools. Instead, why not leverage the power of server-side logic to streamline your data manipulation? By offloading complex operations from the client-side, you can ensure a more efficient and scalable approach. As our customer base expanded to tens of thousands, the likelihood of question timeouts rapidly increased, posing a significant threat to our ability to provide analytics altogether?
To further expedite query processing, we intended to perform pre-calculations on incoming event streams and metrics, indexing them in real-time as they were recorded. Although we didn’t require any manual intervention initially, the introduction of novel chart types necessitated the creation of a custom index for each one, whereas schema updates also triggered changes to our pre-computed data. We were simultaneously handling an entire array of knowledge processing workflows, which came with all the anticipated complexities – missing a scheduled data processing run, for example, would result in outdated information or a chart with a critical section missing from its center.
As the world’s most discerning connoisseurs of quality, we have long been searching for a means to sift through the cacophony of information and identify the truly exceptional. And so, it is that we arrive at the threshold of our grandest quest yet: separating the wheat from the chaff, that is to say, discerning the rare and precious gems amidst the sea of mediocrity.
We meticulously examined multiple alternatives, including:
- . While the venerable open-source database accommodated our complex SQL-based analytics needs, significant rewrite efforts were likely required, including flattening the numerous JSON objects previously stored in Firebase. Given the substantial investment we’d already made in Firebase’s flexibility, adopting PostgreSQL would have been a costly and complex change.
- One alternative open-source SQL database specifically designed for time-series data storage. While QuestDB’s initial question examples were characteristically concise and highly concurrent, it was unclear if the company’s early-stage development and open-source nature would require more maintenance and oversight from our team than we could realistically manage.
We successfully deployed our platform on top of MongoDB. After discovering Rockset through an insider’s recommendation on a Y Combinator startup forum, we found ourselves dealing with identical challenges, which Rockset was uniquely designed to address. In particular, our attention was drawn to these four key aspects:
- Schemaless ingestion of diverse knowledge seamlessly integrates with Rockset’s converged index technology, effortlessly accommodating various data structures and making them instantly query-ready.
- With seamless adaptability to handle intricate SQL queries, instantaneously yielding accurate results.
- By leveraging a fully managed service, we eliminate the need for substantial upkeep and engineering investments, thereby freeing up valuable resources.
- Rockset’s cloud-based developer portal simplifies the creation and management of Question Lambdas and APIs.
Rockset’s onboarding process was surprisingly straightforward. We’d barely gotten started when we were already up and working within a few hours. Given the complexity of these databases, it’s possible that we would have spent several days or even weeks investigating and implementing them.
To avoid premature schema arrangement, we will seamlessly ingest real-time occasion streams into Rockset without interruption. We mustn’t waste a single day reworking one-off capabilities whenever schema changes occur, thereby disrupting our queries and dashboards. The Rockset robotically ingests and prepares the information to promptly answer any form of inquiry that may arise, whether current or potential. It seems like magic!
Actual-Time Analytics, Deployed Immediately
We leverage Rockset to scrutinize and analyze a vast dataset of over 30 million documents. This real-time data synchronisation ensures seamless updates to the buyer’s dashboard, showcasing live insights from both MongoDB and Firebase in two prominent areas.
- . Here, our customers can apply a wide range of filters to drill down into any one of thousands of customers and analyze their interactions on-site as they navigate through the customer’s journey.
- Which showcases interactive charts providing insights into the demographics of guests, featuring a mix of metrics such as daily arrivals and departures, as well as guest categories by accommodation type?
The real-time efficiency proved to be a significant advantage, indeed. The seamless transition to Rockset was also facilitated by its convenience, pace, and minimal ongoing operational overhead. Given the time our small team saves by automating tasks such as manual index construction, knowledge management, and query rewriting – which are often slow and dysfunctional – we find this efficiency extremely valuable.
By achieving this, we’ve successfully accelerated the pace of results while enhancing Savvy’s entry point offerings without sacrificing the quality of knowledge and analytics provided to our clients?
The primary platform is designed to thrive in the cloud, empowering users to extract insights from real-time data with remarkable speed and efficiency. Study extra at .