Conversion rates for online retailers are generally considered pretty dismal by most common measures. Imagine if only 1–3 percent of shoppers entering your store ended up making a purchase. Maybe you’d think of trying a new strategy. The new strategy employed by many online retailers is called retargeting — the use of search and display campaigns to target the 97 percent of visitors who came to your e-commerce site but didn’t convert, meaning they did not make a purchase, fill out a form, or request a demo or call. Retargeting works by keeping track of people who visit your site and displaying your retargeting ads to them as they visit other sites online.
There are many ways to approach retargeting, including the use of vendors who specialize in this strategy as a service and offer significant technology that helps drive the process.
Retargeting in a Nutshell
Retargeting is designed to address the situation where an online shopper goes to an e-commerce website, browses the catalog, places one or more items in the shopping cart, and then leaves the site without making a purchase. Using a retargeting service like TellApart, the retailer can subsequently present a tailored ad including items you previously showed interest in but ultimately did not purchase, as well as other items that you viewed several months back and also didn’t purchase. Retargeting is able to combine past browsing behavior and current intent with real-time incentives, e.g. the items may be offered at a discount in the retarget attempt, to deliver a highly personalized advertisement. Retailers don’t have up-front costs for retargeting services but instead are charged a fee when a shopper clicks on a retargeted ad and goes on to make a purchase.
The technology behind retargeting has gotten more sophisticated in the past few years. Now, using services like AdRoll, retargeting occurs on Twitter and across devices, so if you visited an e-commerce site via a browser on a desktop computer, you could see retargeting ads on your smart phone when using Twitter.
Not All Retailers Need to Retarget
Of course, not all retailers choose to retarget customers, and in fact many companies think long and hard about availing themselves to the process. The clear concern is the threshold to which retargeting becomes bothersome to the potential customer. Further, some companies and products do not match well with the premise of retargeting. For instance, it is doubtful that you’ll be retargeted for browsing the Range Rover Website since high-end retailers don’t see their brands requiring such extra efforts. Volkswagen, on the other hand, may pick up your buying intent along with your physical location for intelligent geotargeting (displaying an ad for a Boston dealership for someone living in Cambridge). Clicking on such an ad may result in you seeing a very similar ad to provide a consistent experience. If you continue to click deeper, you may find that the retailer increases the frequency of the ads. “Frequency capping,” however, is a best practice to not send quite so many ads (only show this ad three times per day, or five times per week).
Retargeting is not without problems. There are situations in which a shopper is retargeted for products or services already purchased, and in some cases a promotion is only good for new customers. So already converted customers are ineligible for the promotion. Incorrect geotargeting happens to top the list of retargeting mishaps.
You'll have the best results if you segment your visitors and customize the ads for each segment, or not retarget them at all. This is where data science and unsupervised machine learning come in handy. Clustering, for example, is naturally applied to this segmentation effort. By intelligently grouping visitors into clusters, it is possible to more accurately do retargeting. Clustering algorithms are typically used for discovering effective customer segments. Some of the leading vendors in this space use distributed computing architectures such as Hadoop and HBase to apply sophisticated predictive analytics and machine-learning techniques to the massive amounts of rich online shopping data, off-line CRM data, and other behavioral information to which retailers have access.
The retargeting process provides even more value to the retailers by relaying analytics on how consumers interact with the ad. This insight is used to automatically optimize the ad campaign — and to adjust messaging where necessary to bring the consumer back to the site to purchase. The retargeting technology can continually optimize in real-time for conversion using machine-learning techniques in order to do the following:
- Determine which shoppers matter most to retailers.
- Identify shoppers across channels and devices.
- Develop a deep understanding of SKU-level purchase intent for each shopper.
Exactly how this is achieved requires scalable technology such as a Big Data platform that manages multiple petabytes of data, grows by at least 1TB/day (serving > 100K transactions per second), and tracks the online shopping behaviors of hundreds of millions of users in real time. TellApart boasts a Hadoop cluster with more than 600 nodes on average and well over 1,000 nodes at peak. As it continues to scale its system, the company is employing the Lambda Architecture as well as building a SQL layer on top of HBase.
To build a fast, scalable, redundant, and cost-efficient infrastructure, TellApart uses Amazon Web Services (AWS). It uses EC2 Spot Instances for less urgent Hadoop-based data processing jobs, plus an architecture based on EC2 Reserved Instances to reduce costs for more predictable front-end ad serving. To do this in real time, the response to bid requests from the Google/DoubleClick Ad Exchange must be under 120ms. The current rate is more than 10,000 of these requests per second making these capabilities require Big Data and serious computing power.
Although not for all retailers, retargeting represents a technology-based strategy that many businesses can use to improve their online sales performance. It is an evolving technology that will benefit from increased use of data science principles to provide more precision for the manner in which the repurposed ads are utilized.
Daniel D. Gutierrez is a Los Angeles–based data scientist working for a broad range of clients through his consultancy AMULET Analytics. He’s been involved with data science and Big Data since long before it came in vogue, so imagine his delight when the Harvard Business Review deemed “data scientist” as the sexiest profession for the 21st century. He is also a recognized Big Data journalist and is working on a new machine-learning book due out in later this year.