Site   Web

August 18, 2015

How to Personalize an eCommerce Customer Journey

Helping customers find items they’re interested in is the primary task of any online shop. Intuitive site navigation and relevant site search are the most common ways to connect customers with the products they want to buy.

However, even the best site navigation and search can fall short when e-shops have large inventories spanning many product categories. Search engines on ecommerce sites have improved greatly over the years, but are still usually not as powerful as Google or Bing. As a result, shoppers must search using accurate terms to get meaningful results. Looking up “chinos,” for example, will probably return so many results that it’s impossible to sift through everything (on this search returns 474 items).

A well-designed eCommerce site should be simple and intuitive for anyone to navigate. Is it actually possible to make one site that fits everyone’s needs though? (Answer: NO). Online shops tend to design their experience for the ‘standard’ shopper. Unfortunately, this ignores the needs of many customers who don’t fit the ‘standard’ mold.

If customers can’t efficiently find what they’re looking for, then they will just navigate to a different site or go back to their list of Google results and try to find the product somewhere else.

Even the most successful online shops have conversion rates of only around two percent and bounce rates as high as 80 percent.

The average eshop has an inventory of around five thousand products. According to Softcube’s customer data, however, the average number of pageviews per eCommerce site visit is only five to six. That means that if customers are shown 25 items per page (a realistic figure), then they will have 200 pages worth of inventory to sort through.

With today’s Big Data, it is easier than ever to help customers find exactly what they want with eCommerce website personalization.

Personalization of online shops is accomplished with computer algorithms that make intelligent predictions about what customers are interested in buying. The data that these algorithms rely on is based on customer’s real-time interactions with an eCommerce site. Thanks to real-time data processing, even two or three clicks on a given site can be sufficient to start generating personalized product offers for a brand-new site visitor.

Product recommendations are at the core of modern eCommerce personalization. Recommendations are generally displayed in blocks, which can be thought of as widgets that are placed on any page of an eshop to showcase items from a shop’s inventory. Some examples of recommendation blocks include: “Customers Who Bought This Also Bought,” “Products Recommended For You,” and “Style Match” (using image analysis).

Personalized product offers are generated by product recommendation engines (also known as recommender systems) that make intelligent predictions based on clicks, items added to shopping carts, purchase history, and other data. Adding personalized recommendation blocks to a product page has been shown to boost sales by five to 10 percent. And site-wide personalization, like Amazon employs, can generate up to 30 percent of total e-shop revenue.

Two Paths to Personalization: In-house Solution or SaaS (Software as a Service)

For huge retailers like Amazon, Asos, Target and Tesco, it is possible to develop proprietary personalization services in-house. While this solution is very expensive, there are some advantages, including:

  • Local data storage;
  • Ability to develop and implement solutions that surpass what competitors are doing;
  • And ability to deeply integrate personalization (for example, taking into account data from point-of-sale systems in stores as well as online sales).

Developing in-house recommender systems requires data scientists and machine learning experts — and such talent isn’t cheap. In addition, processing recommendations requires buying and maintaining (or at least leasing) large numbers of computing and data storage systems.

SaaS solutions like Softcube allow even small and medium-sized eShops to offer personalized shopping experiences without these overhead costs. Pricing for SaaS recommender systems is generally a monthly rate based on either site traffic or sales volume, making recommendations profitable and scalable for practically any online shop.

Personalized Product Recommendation Systems are Powered by Three Types of Information

Product recommendation systems take into account three different categories of information: information about customers, information about products, and business information about a particular online shop.

1. Customer Information

Customer clicks, product views, shopping cart and purchase data, shoe size, favorite brands and colors… all of this is part of customer information. In addition, we might consider shipping methods and the location of the customer: for example, to show items that are eligible for next-day or even same-day shipping to their location.

2. Product Information

Product information includes product names, text descriptions and product images. For in-house recommender systems this information might come from an internal database. For SaaS-based systems it is generally gleaned from product information that is publicly available on eCommerce sites.

3. Store Information

Store information might be thought of as “business logic.” It includes such concerns as product margins and conversion rates of products (based on a store’s prior data), and also includes information about shipping terms, pricing for extra services (for example, personalization of products), and other services that an online shop offers their customers.

Customer information, product information, and store information are analyzed by product recommendation services to create a customized shopping experience for each unique site visitor.

Four Ways that Personalization Can Lower Bounce Rates

There are many factors that affect a site’s bounce rate. Irrelevant search results, 404 pages, out-of-stock items, and Thank You pages after purchases are made all tend to have high bounce rates (and that shouldn’t come as a surprise). Here’s how personalization can work in each of these cases to re-engage shoppers.

1. Search Results

Personalizing search results ensures that the most relevant results always make it to the first page. Showing the correct sizes, favorite brands, and preferred styles of clothing on the first page of results makes shopping a breeze.

2. 404 Pages

There are plenty of examples of boring 404 pages on the Web. Don’t let yours be one of them. A well-placed product recommendation block on a 404 page can draw customers back into your store to continue their shopping journey.

3. Out-of-Stock Items

There’s no disappointment like finding just what you’re looking for only to realize it’s out of stock. Sometime out-of-stock products should be removed from a site, but sometimes there is still a lot of organic traffic coming in through those pages. But since the items are out of stock you can’t sell them.

Instead of removing every out-of-stock item from your shop (especially for those that you anticipate getting back in stock), you can instead use the out-of-stock item pages to highlight similar items that are currently available. This product might be similar based on brand, color, or other characteristics.

4. Thank You Pages

Once a customer has completed a purchase you might think that the customer journey is over. But it’s not! Recommend something that they might have missed — a USB cable for their printer, for example — in order to bring customers back even after they’ve processed an order.

This is the age of Big Data. Online shops have the power to predict customer interests, improve the customer journey, and profit while fulfilling each customer’s unique needs and wants. E-shops of any size must take advantage of big data insights. After all, personalized recommendations and tailored shopping experiences are not just about boosting sales, but are also about creating value and building customer loyalty.



Russell Snyder is a community manager at Softcube Russell interprets big data concepts for the non-data scientists of the world. He holds a BA from Lehigh University in Pennsylvania, is currently studying at the Jagiellonian University in Kraków, and would like a dachshund.