Streamlining predictive analytics in retail marketing
Online flash-sale retailer Zulily uses BigQuery and Tableau to help power its predictive analytics, which, in turn, boosts its marketing efforts and ability to manage incoming data.
With the right tools and strategies in place, companies can find big payoffs by using predictive analytics in retail marketing.
Predictive analytics can add fuel to an ad campaign, help build new consumer bases and produce a higher consumer retention rate.
Zulily, an online flash-sale retailer of women's and children's products, said following best practices for using predictive analytics in marketing has played a key role in the company's growth.
A driving factor in that expansion "is leveraging data science to optimize marketing return on investment," Sasha Bartashnik, marketing analytics manager at Zulily, said in a recent webinar produced with Tableau.
A new system
With millions of social media followers and dozens of ads that launch each day, Zulily has "a lot of data coming in," Bartashnik said.
"I have data on customer feedback, what they're saying about us, what they're buying and how they're making their purchase," she said. "On a typical day, we launch hundreds of ads, serve millions of impressions and convert tens of thousands of customers."
Even amid that swarm of data, with the help of some best practices of predictive analytics in retail marketing, Zulily is able quickly collect and process the data and enable marketers to call upon insights, Bartashnik said.
Sasha Bartashnik
Since 2014, the Seattle-based company has used a self-service analytics platform that integrates the visual analytics tools of Tableau and the data warehouse and analytics capabilities of Google's BigQuery.
The system, according to Bartashnik, gives users a single view of data from across a number of marketing and social media platforms, and it blends that data in with Zulily's own sources, which include customer order transactions and product catalogs.
The product of collaboration between Zulily's marketing analytics and tech teams, the system replaced an existing platform that was essentially a combination of a SQL Server database and a Hadoop cluster.
The old platform was limiting in terms of scope and size, Bartashnik noted, and it slowed down the analytics team by forcing it to rely on IT for everyday tasks and forcing users to rely on analysts for even basic insights.
With the adoption of the new system, "the tech team pretty much put all the data in our hands," Bartashnik said. In essence, "data was democratized," she said.
Open data
In order for marketing to be more self-service, we had to open up the data on the back end.
Sasha Bartashnikmarketing analytics manager at Zulily
Also among best practices in predictive analytics in retail marketing is the importance of data being democratized -- that is, accessible to the analytics team, Bartashnik said in the webinar.
"In order for marketing to be more self-service, we had to open up the data on the back end," she said.
The Zulily teams put in place special data sets and dashboards to enable marketers to get to the data and easily use it themselves.
The dashboards are "really flat for our marketers to use any metrics they wanted, instead of having to ask for an analysis whenever they wanted a slice of something," Bartashnik said.
With all of the data stored on BigQuery, the system is highly sliceable and scalable, Bartashnik noted. Tableau's data visualization system then helps the analytics team turn the data into reports that are easy to read and dig into -- an important piece of predictive analytics in retail marketing.
Predictive analytics in retail marketing
The Zulily predictive analytics system ties collected data to each customer, allowing the analytics team to use targeted marketing tools to show individual customers products they might be more likely to buy. Because the system is scalable, allowing analysts to view data of any size at any level of detail, data can also be combined to give analysts insights into more general trends, Bartashnik said.
The company can use collected data to identify "high-value customers" or repeat customers, and then zoom out to identify what distinguishing behaviors the customers display.
With those behaviors in mind, Zulily is then able use predictive analytics in retail marketing tools to identify which new customers could be high-value and potentially drive growth and profits.
The webinar, "Remove Data Analytics Bottlenecks," was presented in conjunction with Tableau Software and hosted by Adweek in June.
In related news, Amazon recently announced that Zulily will be moving its infrastructure to Amazon Web Services and upgrading many of its databases to Amazon Aurora.
Sasha Bartashnik
