Credit score: Phil Goldstein
Jerry Wang, Peloton’s Director of Knowledge Engineering (left), and Evy Kho, Peloton’s Supervisor of Subscription Analytics, talk about how the corporate has benefited from utilizing Amazon Redshift.
New York-based Peloton, which goals to assist individuals world wide attain their health objectives by means of its linked health gear and subscription-based lessons, noticed booming development within the early stage of the COVID-19 pandemic. In 2020, as gyms shuttered and other people appeared for methods to remain lively from the security of their properties, the corporate’s annual income soared from $915 million in 2019 to $4 billion in 2021. In the meantime, the corporate’s subscribers jumped from round 360,000 in 2019 to 2.76 million on the finish of 2022.
As Peloton’s enterprise continued to evolve amid a altering macroeconomic surroundings, it was important that it might make sensible enterprise selections rapidly, and among the best methods to try this was to harness insights from the massive quantity of knowledge that it had been gathering over current years.
Throughout that very same time, AWS has been centered on serving to prospects handle their ever-growing volumes of knowledge with instruments like Amazon Redshift, the primary absolutely managed, petabyte-scale cloud knowledge warehouse. The service has grown right into a multifaceted service utilized by tens of hundreds of shoppers to course of exabytes of knowledge every day (1 exabyte is equal to 119 billion music downloads). With Amazon Redshift, you get entry to a contemporary knowledge structure that helps you break down inner knowledge silos, share knowledge securely and seamlessly, and assist a number of customers who don’t have specialised knowledge and analytics expertise.
When Jerry Wang, Peloton’s director of knowledge engineering, joined the corporate in 2019, he wanted to verify the service would deal with the corporate’s huge and rising quantities of knowledge. He additionally wanted to make sure Amazon Redshift might assist the corporate effectively handle the big variety of knowledge and the customers who would wish to entry it, and ship insights on that knowledge at excessive velocity—all whereas being cost-effective and safe.
Wang was delighted to see that as Peloton skilled its huge development and alter, AWS continued to launch new Amazon Redshift options and related capabilities that might completely go well with his firm’s wants at simply the best time.
“Over time, I’ve at all times been within the stage the place I hope Redshift can have a brand new, particular function,” Wang says, “after which, in a really restricted period of time, AWS releases that form of function.”
Peloton’s knowledge volumes soar because the enterprise grows
Credit score: Peloton
As Peloton’s enterprise has developed, the quantity of knowledge it’s producing and analyzing has grown exponentially. From 2019 to now, Wang studies the quantity of knowledge the corporate holds has grown by an element of 20. In reality, a full 95% of the full historic knowledge the corporate has generated has come within the final 4 years. This development has been pushed each by surges within the variety of customers on Peloton’s platform and the number of knowledge the corporate is gathering.
Peloton collects reams of knowledge on its gross sales of internet-connected train gear like stationary bikes and treadmills. The corporate additionally collects knowledge on prospects’ exercises, which it then gives again to them in numerous studies equivalent to a month-to-month abstract, giving them insights into how typically they labored out, their finest output, developments of their exercises, the trainer they used probably the most, what number of energy they burned, and extra. All of this knowledge helps Peloton make strategic enterprise selections, refine its operations to turn into extra environment friendly, modify its programming, and drive subscriber engagement and development.
In 2019 and into 2020, as Peloton’s enterprise boomed, the corporate wanted an analytics system that would assist it handle an explosion of knowledge, each from customers and associated to its enterprise. The corporate embraced Amazon Redshift due to the service’s versatility, ease of use, price-performance at scale, steady tempo of innovation, and skill to deal with concurrent queries from dozens of inner knowledge groups.
Wang mentioned that when he joined the corporate, there have been two sorts of customers who have been performing every day knowledge operations in Peloton’s Amazon Redshift knowledge warehouse. One group carried out extract, remodel, and cargo (ETL) operations to take uncooked knowledge and make it out there for evaluation. The opposite was a gaggle of enterprise customers who, every morning, would carry out queries to generate native knowledge visualizations, making a surge of capability on the Amazon Redshift knowledge warehouse. “So, when these two masses ran collectively, the efficiency suffered immediately,” Wang says.
One of many options Peloton adopted was Amazon Redshift Concurrency Scaling, which gives constant and quick question efficiency even throughout hundreds of concurrent customers and concurrent queries. This helped remedy the issue by mechanically including question processing energy in seconds and processing queries with out delays. When the workload demand subsided, the additional processing energy was mechanically eliminated, so Peloton solely needed to pay for the time when Concurrency Scaling knowledge warehouses have been in use. Wang says Peloton was working about 10 hours of Concurrency Scaling on a constant every day foundation to cope with the congestion, which, he says, “solved my downside at that second.”
In 2020, because the pandemic impressed hoards to hop on bikes of their dwelling rooms, Wang additionally upgraded Amazon Redshift with the newly launched Amazon Redshift RA3 situations with managed storage (RMS). These represented a brand new technology of compute situations with managed, analytics-optimized storage designed for high-transaction, quick question efficiency and decrease prices.
“The brand new occasion … was a fantastic function for us,” Wang says. “It solved our concern about transferring from terabyte scale to petabyte scale.”
Peloton’s enterprise is pushed by quite a lot of knowledge for a variety of customers
Credit score: Peloton
Peloton’s enterprise mannequin is pushed by all kinds of huge volumes of knowledge. Along with promoting bikes, treadmills, and indoor rowing machines, and increasing its subscription platform to incorporate non-equipment-based exercises, the corporate has dozens of instructors in 5 nations, and it licenses music from three main music licensors. In 2022, it started renting bikes in addition to promoting them. Internally, Peloton workers working in finance, accounting, advertising, provide chain operations, music and content material, and extra are utilizing knowledge to trace subscriber development, content material engagement, and which gross sales channels are resulting in probably the most web new subscriptions.
“There was a time once we have been only a bike firm, and now we’re a lot greater than that,” says Evy Kho, supervisor of subscription analytics at Peloton.
There may be additionally a a lot wider vary of gross sales channels for Peloton gear than only a few years in the past. Up to now, Peloton prospects might solely buy bikes by means of the Peloton web site or secondhand. Now, prospects should purchase {hardware} from third-party websites like Amazon. That launched “a extremely attention-grabbing knowledge downside” for Peloton, says Kho, because it strives to find out easy methods to hyperlink subscription signups again to train gear gross sales.
Within the face of this variability, complexity, and wish for immediate entry to knowledge to tell enterprise decision-makers, Peloton embraced Amazon Redshift Serverless as an early adopter after AWS launched the function in late 2021. Redshift Serverless permits firms to rapidly run and scale analytics capability with out database managers and knowledge engineers needing to handle knowledge warehouse infrastructure.
Redshift Serverless additionally has the flexibility to rapidly spin up analytics capability for various customers, or personas, inside a corporation. This permits completely different groups throughout Peloton to carry out analytics on the identical datasets on the similar time to generate insights on their particular person components of the enterprise. It’s “extremely necessary when it comes to assessing what’s been good for our enterprise,” Kho says.
Wang additionally says Peloton is contemplating supporting particular personas for individuals who want analytics round monetary data ruled by securities rules, and one other for customers who must carry out analytics on knowledge ruled by rules round personally identifiable data (PII).
Wang factors out that Redshift Serverless additionally permits him to spin up Amazon Redshift knowledge warehouses to deal with particular utilization patterns. For instance, ETL masses are sometimes excessive I/O however require low CPU sources, and are very predictable as a result of Peloton controls the method. Nonetheless, when inner customers wish to carry out knowledge analytics or machine studying, the corporate doesn’t have management over the demand for these queries, and the load on Amazon Redshift knowledge warehouses will be variable, with some queries extra CPU-intensive than others. Beforehand, any provisioned knowledge warehouse would have a set price, and it must be provisioned to deal with the very best doable workloads even when the utilization charges turned out to be low. Now, for these completely different situations, Wang creates completely different Amazon Redshift situations to deal with that variability with out these heavy, fastened prices.
As Peloton’s use of Amazon Redshift has developed and matured, its prices have gone down, in line with Wang. “In the event you have a look at Serverless, the quantity … that we spend on the Serverless is definitely a lot smaller than we did beforehand, in comparison with the Concurrency Scaling price.”
In a serverless surroundings, there is no such thing as a upfront price to Peloton. “I can set it up as rapidly as I can and we pay as we want it,” Wang says. “It scales up when the load goes up. So, it’s an ideal match.”
Peloton makes use of Amazon Redshift to get to insights sooner
Credit score: Peloton
Peloton’s deal with effectivity and sustainable development has meant that it must act extra rapidly than ever to make sound, data-informed enterprise selections. Peloton, Wang notes, is long gone the stage the place all it cared about was development. “We’re a mature firm now, so operational effectivity is essential; it’s key to the enterprise,” he says.
When Peloton launches new merchandise, for instance, two issues sometimes occur, Wang says. One is that there’s a spike in knowledge volumes, each in site visitors to its web site and the variety of gross sales transactions it’s processing. The second is that the corporate’s administration crew will need real-time updates and evaluation of how gross sales are performing.
Redshift Serverless and knowledge sharing lets customers rapidly begin performing real-time analytics and construct reporting and dashboard functions with none extra engineering required. Wang confirms this profit, particularly within the instance of a brand new product launch, saying it “will scale up by itself with out me having to intervene. I don’t must allocate a funds. I don’t want to vary any configurations.”
Up to now, when Peloton solely supplied its health gear by means of its personal web site, it was simple to affiliate success knowledge on orders with subscriptions. Nonetheless, as these channels grew and have become extra complicated, Peloton turned to the info sharing capabilities of Amazon Redshift to share knowledge rapidly and simply throughout groups. Peloton’s groups for subscriber analytics, provide chain, accounting, and extra want quick entry to success knowledge to make sure they’ll observe it precisely, reply if modifications are wanted, and decide how success knowledge aligns with subscriptions and income.
“Getting them these outcomes even sooner has been extremely useful, and is barely changing into extra necessary as we have now turn into way more data-driven than I believe you might argue we have been earlier than,” Kho says.
Amazon Redshift marries knowledge safety, governance, and compliance with innovation
Like all prospects, Peloton is worried about knowledge safety, governance, and compliance. With security measures like dynamic knowledge masking, role-based entry management, and row-level safety, Amazon Redshift protects prospects’ knowledge with granular authorization options and complete id administration.
Prospects are also capable of simply present authorizations for the best customers or teams. These options can be found out of the field, inside the usual pricing mannequin.
Wang notes that Amazon Redshift’s safety mannequin is predicated on a conventional database mannequin, which is a well-understood and strong mannequin. “So for us, to provision entry on that mannequin is kind of easy,” Wang says.
At each stage of Peloton’s evolution during the last 4 years, the corporate has been capable of flip to AWS and Amazon Redshift to assist it successfully handle that development and complexity.
“After I began,” Wang says, “I mentioned, OK, I would like a short lived enhance in capability. Then got here Concurrency Scaling. After which I mentioned, I would like cheaper storage, and [RA3] comes alongside. After which the last word problem [was], I’m not happy with a monolithic Redshift occasion. Serverless solved that difficulty.”
Be part of AWS Knowledge Insights Day 2023
If you wish to find out how your organization can use Amazon Redshift to investigate massive volumes of knowledge in an easy-to-use, scalable, cost-effective, and safe method, don’t miss AWS Knowledge Insights Day on Might 24, 2023. Throughout the day-long digital occasion, study from AWS leaders, consultants, companions, and prospects—together with Peloton, Gilead, McDonald’s, World Foundries, Schneider Electrical, and Flutter Leisure—how Amazon Redshift and options like Amazon Redshift ML are serving to drive enterprise innovation, optimization, and price financial savings, particularly in at the moment’s unsure financial instances.
To study extra about Amazon Redshift, see Amazon Redshift and Amazon Redshift: Ten years of steady reinvention.
Concerning the creator
Phil Goldstein is a copywriter and editor with AWS product advertising. He has 15 years of expertise writing expertise, and previous to becoming a member of AWS was a senior editor at a content material advertising company and a enterprise journalist protecting the wi-fi trade.