TDWI | Training & Research | Business Intelligence, Analytics, Big Data, Data Warehousing

Q&A: Predictive Analytics Can Shape Customer Behavior

Using data management capabilities from its online address management services, PCA Predict is turning its attention to using predictive analytics to monitor and respond in real time to online customer behavior.

• By Linda L. Briggs
• August 1, 2016

"The holy grail of predictive analytics is providing the individual attention to detail that a customer would receive at a small corner store," according to Jamie Turner, CEO and cofounder of PCA Predict. Turner's British-based company provides cloud-based address management services to thousands of businesses worldwide.

The firm's platform fields over five billion queries a year and is used by businesses including Trip Advisor, Cross Pens, Virgin, Dow Jones, and Disney; it is now turning its attention to predictive analytics for monitoring online user behavior during website visits.

Upside: How has online behavior changed in the past several years? Are companies keeping up with those changes?

Jamie Turner: Advances in technology are constantly changing the way we interact with brands online. It has never been easier or more crucial to understand your customer, identify and determine trends, and react quickly to changes in the norm. Customers today expect a seamless experience across all devices; companies need to provide an experience that will keep customers coming back.

One of the main technology trends we have seen in the last few years is the use of predictive analytics, which can be used to measure behavior to anticipate how engaged customers are in the online process. In theory, if your customers are happy because what they are seeing is of interest, they'll stay with your site and continue the journey. If not, they'll leave.

Predictive analytics can help figure out why a customer abandons a visit to your site. By carefully tracking where a customer goes and when during a site visit, analytics can be used to predict at what point someone becomes less interested, and steps can be taken to retain them, such as coupons and special offers. Also, designers can change the look and feel of content or product pages to keep customers on track.

The holy grail of predictive analytics is providing the individual attention to detail that a customer would receive at a small corner store. There, the owner knows the customers -- their likes and dislikes -- and is able to cater to each of them on a very personal level. The challenge is providing that level of service at the scale of a website.

How far along are companies in doing that?

So far, businesses have tried to create models of typical customer behaviors and then superimpose those models across all customers. However, adopting a one-size-fits-all strategy is not the most efficient way of providing the level of service expected by today's customer. Predictive analytics helps companies look deeply at what the customer is actually doing at the website. That's what really matters, not what characteristics they might share with other customers.

Advances in technology, particularly with predictive analytics, mean that individual behaviors can be measured and responded to based on what the customer is doing or has done. Using these techniques will enable companies to deliver personal, relevant online experiences at the scale required in today's fast-moving multichannel environment.

Going even further, pioneering enterprises are starting to implement analytics that allow them to track customer behavior in real time and respond immediately. However, that's still a minority of early adopters. Many still struggle with the realities of creating the infrastructure needed to turn customer behavior data into valuable, actionable insights.

Most businesses are halfway there -- they know they have a wealth of data at their fingertips, and they realize its importance, but they don't know how to use it. Any business with a website today has access to a digital record of transactions, visitors, and subscribers. Coupled with advances in technology, these businesses can start analyzing their data to improve their customer experience -- and ultimately, their revenue streams.

Your predictive analytics software, Triggar, can be used to monitor online user behavior. What does it monitor specifically? Can you talk about how it works and how it might be used?

We monitor on-page activity and actions to build a profile of behavior. By considering the way people interact with the page, we can infer likely next steps. If we know what's likely to happen next -- for example, maybe the customer is about to abandon the cart -- we can offer targeted incentives to keep them on track.

As we all know because we've done it ourselves, people often leave websites due to frustration -- perhaps they can't find the product they want, or a process is too slow. Our Exit Prevention tool analyzes on-page behavior to predict when a customer is about to leave, providing companies with a range of incentives to encourage customers to stay on the website.

What sorts of incentives can encourage customers to stay?

It depends on the individual situation and behavioral characteristics. It may be an offer of assistance, a coupon, or a limited-time offer. Eventually, the company might make a change to the content in order to explain or present things to the customer in a more appropriate way.

Could companies make website design changes based on feedback from Triggar?

Absolutely. Our Triggar technology can highlight choke points in the flow and offer advice for improving them, then show evidence of whether the improvement actually worked as intended.

How is your "fuzzy matching" technology, which finds matches or eliminates duplicate addresses regardless of typos and misspellings, being used by websites?

Almost $4 trillion worth of revenue is lost each year due to online cart abandonment; 47 percent of customers abandon carts due to long checkout forms. Our flagship address verification tool is already used by thousands of online retailers around the world to capture and verify postal addresses in online forms.

By adding it to website forms, apps, and CRM systems, our customers can significantly cut the number of user keystrokes required to automatically complete an address. This helps improve conversion rates and the user experience, while reducing the chance of failed deliveries. The online retailer ThinkGeek said our tool "definitely intensified site visitor activity" -- they reported a significant increase in revenue from using it.

We recently added "fuzzy matching" to our address capture tool. That means it will recognize and capture an address even when typos and misspellings are entered.

What were the challenges in developing that sort of technology? What new technologies are helping make it possible?

Address capture is superficially very simple but making it fast, reliable, and relevant is a real challenge. We're smaller than Google, but we're indexing billions of addresses. That means we've had to build very specialized search software -- nothing off-the-shelf came even close to meeting our performance requirements. It's very memory intensive, and it has taken time to build. It's only in the past few years that memory has become cheap enough to make it a reality.

Predictive products have required three major leaps. First, the algorithms have been massively refined and optimized to make them work on commodity hardware. Second, hardware is cheaper than ever before. As computer games have become more sophisticated, gaming hardware has become really powerful. It turns out that gaming hardware is also perfect for crunching these algorithms -- at a fraction of the cost of regular hardware. Finally, we have data, lots of data, to build these systems on.

Where is predictive analytics heading? What kinds of things might we see in the next five years?

I think that real-time analytics is the next big step in data analysis: the ability to understand what people want and what is influencing them right now, rather than trying to extrapolate what they wanted last month. We're doing that now, but unfortunately for most businesses real-time analytics is more a distant ambition than a priority today.