Gilt’s Steve Jacobs on Surviving and Growing the “Noon Frenzy”
January 27, 2016 • Blog Post • By Todd Wasserman, HPE Matter Contributor
IN THIS ARTICLE
- Steve Jacobs, chief product officer and head of growth marketing at Gilt, discusses adapting the company’s computing to account for sudden daily surges
- Jacobs also discusses how Gilt uses Big Data to acquire new customers and target existing customers
How the luxury flash sale site prepares its infrastructure for daily traffic surges
Every day, some nine million people receive an email or push notification from Gilt just before noon offering enticing deals, a happenstance known internally as the “noon frenzy.” Gilt’s IT infrastructure absorbs the brunt of that frenzy time and again with a system designed to use data to boost response rates.
Steve Jacobs is charged with keeping this virtuous cycle going. The former director of high performance computing, solutions and engineering at Merrill Lynch, Jacobs joined Gilt in 2009 and has since become its chief product officer and head of growth marketing, adapting the company’s computing to account for sudden daily surges.
At Merrill Lynch, Jacobs was running a computing system that employed 20,000 CPUs used by traders and analysts to run scenarios and modeling. “That has really prepared me well for this noon spike that Gilt gets,” he says. Gilt’s own system is based on a 13 terabyte database that ingests roughly 10 gigabytes per day.
A pioneer in flash sales
Kevin Ryan, the former CEO of DoubleClick, created Gilt with Alexandra Wilkis Wilson, a retail veteran at fashion brands Bulgari and Louis Vuitton, Alexis Maybank, with a background in eCommerce at eBay and AOL, Mike Bryzek and Phong Nguyen in 2007. Ryan learned that Vente-Privee, a European site, had built a business by buying old inventory from designers and then selling it at a discount online.
Gilt, which launched before other flash sales sites like Groupon and Fab, made a name for itself by partnering with designer Zac Posen. Initially, Gilt offered one sale a week, but quickly scaled up to three a week and then four a week.
Some nine years later, many sales happen on Gilt at once, each targeted to an audience of one. The idea is that Gilt learns so much about its customers that it can anticipate the types of products they might want. This means parsing many data points and plugging them into a predictive software platform.
Surviving daily jolts
Longtime Gilt employees fondly remember a sale for Christian Louboutin shoes in 2008 in which 45,000 women tried to buy the same pair of shoes at once. That brought the site as close to crashing as possible.
Since then, Jacobs says, Gilt’s IT infrastructure has ably handled whatever has come its way, including the recent holiday season’s rushes around Black Friday and Cyber Monday. “It all scales horizontally,” he says, adding that the company does load testing ahead of time to make sure the infrastructure can handle the strain.
Gilt’s personalization algorithm
As Jacobs explains, fashion is primarily about brands. If you’re a Gilt customer and you’ve bought a brand’s products or searched it in the past on Gilt, then you will likely see offers related to that brand. That’s not the end of it, though. “There’s a big element of discovery with Gilt,” he says, adding that Gilt also uses that data to figure out which brands you might be interested in.
Gilt’s personalization algorithm determines which message you see. “We use all that data to personalize, if possible, every email and push notification we send to you,” he says.
Full disclosure on targeting
Gilt members don’t have to wonder why they’re receiving specific deals. The company makes a practice of disclosing its methodology for targeting. If you see a pair of Jimmy Choo shoes, for instance, Gilt will explain that you’re seeing that offer because you’ve bought that brand’s shoes before.
In addition to sharing its decision-making process behind targeting, Gilt also asks for feedback on how members feel about such targeting. Jacobs says close to 90 percent of people who respond say knowing how Gilt personalizes its offers is helpful. “Making it more explicit was really powerful,” he says.
Looking ahead, Jacobs says that he plans to focus more on consumers’ post-purchase experiences and attempt to glean some insights there in 2016. “We have some data there, but we’re going to be investing more,” he says. “We want to really get a better sense of how someone feels that moment that they open the box.”
Big Data changes everything. HPE Software gives you the power to transform it into actionable intelligence so you can capitalize on new opportunities and solve real problems in the moments that matter.