But Barkha Saxena, VP of Data Science & Analytics, recognized that Poshmark’s existing Amazon Web Services (AWS) Redshift analytics data warehouse lacked the capacity to ingest all of this high- frequency information. So in 2014, she turned to Qubole to provide the company’s big data platform.
Founded in 2011, Poshmark is the leading social commerce platform for the next generation of retailers and shoppers. One in every 30 women in America now sells on Poshmark, which showcases around 75 million items across 5,000+ brands and has $100 million worth of inventory uploaded into the marketplace every week.
Qubole has enabled Poshmark to build data extraction and preparation pipelines that take big data from AWS S3 and make it available to data scientists and business users via the Qubole user console, Redshift, and the Looker business insights platform. As a result, Poshmark is able to use its clickstream user activity data to empower all parts of the business and enable the best experience for its community, while keeping pace with rapid growth. Poshmark’s community of five million Seller Stylists is now serving 40 million shoppers.
Using data accessible through Qubole, Poshmark’s team of over 20 data scientists and engineers develop data pipelines for Redshift and deliver critical business insights through a variety of data science initiatives. The team is also responsible for building data products that create value for Poshmark’s community across all functions — from growth and marketing to product development and business operations.
“Qubole connected us with some of their existing clients, and talking to those customers made us feel that this product and this company will be the right partner on our journey,” according to Saxena. She continues, “Our users give us so much information through data about what they like and don’t like on the Poshmark platform — we believe that it is our responsibility to turn that data into greater value for our community. Qubole has certainly been a great partner in helping us achieve this goal.”
“We believe that it is our responsibility to turn our data into greater value for our community. Qubole has certainly been a great partner in helping us achieve this goal.” -Barkha Saxena, VP, Data Science & Analytics, Poshmark
Before Qubole, Poshmark had built data pipelines that provided limited user activity data and aggregated metrics through data extraction, transformation, and loading (ETL) processes from the production database into Redshift. But these pipelines could only process a very small set of user activity data due to Redshift’s limitations, instead of all of the high-frequency social marketplace activities.
Qubole enabled Poshmark to collect, store, and use this critical data at a granular level. “If it wasn’t for Qubole, we would have probably been delayed months to a year in embarking on our big data journey,” Saxena says. “We would have missed all the insights from the data — insights that have been a strong driver of so many of our growth strategies.”
“If it wasn't for Qubole, we would have probably been delayed months to a year in embarking on our big data journey. We would have missed all the insights from the data — insights that have been a strong driver of so many of our growth strategies.” -Barkha Saxena, VP, Data Science & Analytics, Poshmark
Poshmark is growing rapidly, but that in turn creates its own data challenges. Being a social commerce platform, Poshmark has a network effect that means when new users join, they create social connections with other users by liking, commenting, sharing, making offers, and purchasing. “Hence with every new user that joins, data grows exponentially,” Saxena said.
Qubole has enabled Poshmark to not only keep pace with this goldmine of data, but also to continue developing new data products. One key innovation has been Poshmark’s ‘people-matching’ algorithm, which nourishes its online community by connecting users to ‘style mates’ — other people who closely share their fashion preferences. Through such innovations, Poshmark looks to further increase engagement of its users who already have Facebook levels of engagement.
“That is our main goal for the near future — investing a lot more time and energy and resources in building data products that are driven by machine learning.” -Barkha Saxena, VP, Data Science & Analytics, Poshmark
Qubole continues to play a key role in Poshmark’s data strategy. In the near future, Poshmark wants to better utilize two types of metadata it collects — images of items listed on the platform, as well as seller descriptions of these items and user comments on these items. Poshmark is exploring text mining and image analysis algorithms to potentially use this data to help its consumers through a number of initiatives, from listing new items more quickly to improving sales.
The biggest value of Qubole, Saxena indicates, has been partnering with Poshmark from the very beginning of their data journey. And today the company is at the stage where they are ready to take the next big step. “We are looking to increase our investment in machine learning-based data products multifold, and due to our early partnership with Qubole, we have the data and infrastructure ready to enable that.”
Big data galvanizes corporate growth and high customer engagement
Machine learning algorithms use Qubole to fuel continued innovation