How to get real-time results from high-value purchases

By Team Session AI on March 6, 2018

Securing the Sale of Infrequently-Purchased Items through Machine Learning Models

Did you know that it costs five times more to attract a new customer than to retain an existing one? With this in mind, many e-commerce companies today have shifted their customer engagement strategies to center more heavily on building long-lasting loyalty amongst their current customer base, rather than focusing solely on new prospects.

But for online retailers of more infrequently purchased items, like furniture, cars, or big-ticket technology products, it can be much more difficult to ‘know’ their customers and use their past behaviors to inform future action. Prospects visiting these types of e-commerce sites are more likely to be inexperienced with the brand, and it would not be uncommon for the enterprise to have little or no historic purchase data to draw upon for use in tailored messages and offers. In these cases, securing the sale comes down to an enterprise’s ability to analyze the customer’s in-the-moment behavior to identify if they are likely to buy, and if so, determine what further action is needed to convert that interest to a sale—all in-session.

Differentiating Prospects from Perusers

The reality of e-commerce is that not every individual who visits a website is looking to make a purchase; rather many know they are simply browsing even before they hit the page. For e-commerce companies selling infrequently purchased items, it is therefore critical to identify true prospects and allocate resources to those most likely to buy.

While it is often easy to separate prospects from perusers in person, doing so online requires intelligent analysis of customer data. In fact, McKinsey reports that enterprises acquire 23x more customers when leveraging advanced customer analytics than when they do not.

The most effective way to maximize this customer data is through the use of machine learning (ML) models. These models are able to analyze a customer’s on-site interactions, along with additional factors such as search history and environmental conditions, to determine the legitimacy of the prospect. True ML models then continue to learn and improve with every customer interaction. This proves integral for infrequent purchases by enabling e-commerce companies to analyze with high levels of accuracy patterns of behavior that are indicative of customers looking to buy versus customers who are simply looking.

Converting Interest to Sales

After the ML model identifies the true prospect, enterprises must connect these individuals to the right product or offering to secure the sale—even if the prospect is a new, unfamiliar customer. Doing so effectively for first-time shoppers comes down to analyzing historic data from previous interactions with like customers and using intelligent decision-making to determine when personalized intervention is needed to influence behavior. Once again, this can be achieved through the use of an advanced ML model. For example:

Valerie is a new visitor to an online department store. When she logs onto the site, she is first attracted to a homepage ad that features children’s clothing on sale—suggesting she has a large family with young children. After clicking on this ad, she then specifically searches for stackable washers/dryers. Attuned to this information, the e-commerce site auto-populates a “You may also be interested in” feed that features high capacity washer/dryer units optimized for large loads and tough stains. Valerie places one of the products from the feed into her cart but does not proceed to checkout. As she is about to navigate off the page, the retailer strategically intervenes with an in-app message, “28 people have purchased this product in the past 24 hours” and directs Valerie toward reviews of the product. After reading the reviews, and knowing how many people have recently purchased the product, she gains the confidence needed to make this purchase for her household and completes the order.

With an advanced ML model in use, the e-commerce site interprets real-time browsing patterns, intervening when appropriate with a situational offer that generates the purchase confidence needed to complete the sale.

ML models can evaluate customer behaviors further still to enable personalized intervention that varies based on the model’s analysis of the scenario and need, including: 

  • Product Confidence: Brian has spent a considerable amount of time reading the product specifications on a number of televisions across an e-commerce website. An ML model could recognize that he looks for third-party validation before committing to a purchase, and so sends him a confidence-based notification: “Check out our top reviewed televisions!”
  • Price Confidence: Jenna browses for the right couch and pillows online, adding several possible pillow options to her cart along the way. While shopping, she avoids clicking on the highest priced options. As her click-thru rate on the website slows—indicating to the ML model shopper fatigue—it sends her an offer that says, “Get two complimentary pillows with the purchase of any couch,” and specifically features pillows from Jenna’s cart.
  • Sense of Urgency: Savanna is looking for a prom dress and filters her search to only show dresses available in size 6. She adds a couple to her cart and is greeted with a message that reads: “Hurry! Only one left in your size at your nearest location. Head there now to try it on!” 
  • Convenience: James has been searching for a new outdoor furniture set all winter. He comes across an e-commerce site for the first time, and immediately navigates to the delivery and removal page. At this point, he receives a pop-up notification: “It’s time to start fresh! Get free delivery and removal of your old furniture when you spend over $1,500.

Through 1:1 personalization efforts—championed by true ML models—enterprises are able to optimize the first-time or infrequent shopper’s experience to maximize their likelihood to purchase.

At ZineOne, ML is at the heart of our AI-powered Intelligent Customer Engagement (ICE) platform, extracting historical, real-time, and environmental data to predict customer intent and deliver 1:1 personalization to all customers, whether previously known or first-time visitors to the website. Contact ZineOne to discover more about how our powerful ML model can drive results in your enterprise.

Featured posts

How to prevent consumers from gaming promotions

A consumer shares your promo, and then your budget is spent qithout reaching your goal. This blog explains 4 ways in-session marketing reduce the risk of improper promo usage.

Read Blog
Strategic guide to selling your overstock products this holiday season

A strategic guide to selling your overstock products this holiday season

Selling overstock products during the holiday season can be challenging, but it’s not impossible. By understanding the challenges, you can develop a strategy that helps you move excess inventory and make room for new products.

Read Blog
5 ways to unlock higher AOV

5 ways to unlock higher AOV and revenue with AI for ecommerce

AOV plays a crucial role in profitability as it directly impacts your bottom line. By increasing the AOV, you can generate more revenue without incurring additional marketing or acquisition costs.

Read Blog

Request a demo

Apply now