All customer success stories Content Cloud customers Commerce Cloud customers Financial services B2B manufacturing B2B distribution B2C retail Commerce Cloud Content Cloud Content Recommendations Content Intelligence AI Personalization Cloud services All capabilities Financial services B2B manufacturing B2B distribution B2C retail Resource library Customer Hub Trust Center Content management Ecommerce Personalization Email marketing CCPA Articles Guides Reports & Studies Events Webinars Customer success stories Partner Network App Marketplace Training & Certification Support Expert Services Managed Services Customer Hub User Guide Quick start for developers Developer documentation Developer community

Intelligent ecommerce personalization for retailers

Product recommendations are just the start of what retailers and wholesale distributors can do today with personalization. This guide presents the latest personalization techniques that can quickly increase your conversions by 5-7 percent.

Introduction: Why personalize?

It’s widely accepted that visitors who engage with personalized product recommendations on a website convert more, buy more and have a higher average order value than those who don’t. Click through rates of 8-25% are not unusual.

At its simplest level, a good recommendation engine will select and display personalized product recommendations that are the most relevant to each individual whilst they are browsing.

This enhances the findability of products by automatically showing products that the user didn’t realize they were interested in until they were actually shown them. This “discovery” aspect is particularly useful for sites with large product catalogues and for sites such as apparel sites where the user may have only a vague idea of what they are looking for.

Increasing incremental sales

For the user, recommendation engines offer a richer customer journey which will ultimately improve loyalty and customer lifetime value. For the retailer, it offers measurable, incremental sales.

However, extensive research shows that retailers are only using personalization tools in approximately 40% of all their onsite shopping sessions. For those retailers who have invested more heavily, and now feature some element of personalization in more than 80% of all shopping sessions, incremental sales have increased on average by over 10% and up to 20% in some cases.

Shortcomings of segmentation

Some companies create behavior profiles by placing individuals into simple segments. Product recommendations and personalization tools are then organized according to each segment. But a segmented model is not likely to deliver the best results because targeting still remains imprecise.

Multivariate testing suffers under the same constraint, showing products with multiple variations to eventually work out what the best product to serve is at any given point. However, it is not able to deliver a personalized experience for each visitor.

Individualizing the experience

To be most effective, it’s essential that individual profiles are created for each individual visitor who visits a website. Each profile should be unique to each visitor.

In order to refine the relevancy of products recommended there also needs to be a mechanism for capturing and learning visitor actions and responses to products recommended to them. Only in this way can a precise 1-to-1 targeting of the right product in front of the right person at the right time be enabled.

It’s all about generating extra sales

Personalization is an important navigation tool. After all, it helps customers to find more quickly the products they want to buy which makes for a good customer experience which in turn supports conversion and retention objectives.

However, the focus has to be squarely on its ability to increase sales. Key performance indicators should be applied to monitor the ongoing sales results achieved by each page and by each recommendation widget on the page if more than one are used.

Behavior data is constantly evolving so it’s essential to keep validating how well each page type is performing in generating sales.

Keeping an eye on engagement

While monitoring the sales achieved, it is also important to monitor the level of engagement each visitor is demonstrating with respect to the products that are being served. A high click through rate of products recommended will give a clear indication of the relevancy of the products served to each visitor.

The click to purchase (CTP) metric indicates how well a product sells. If this is less than 10% it may indicate that there is something else such as price that is inhibiting the sale.

True 1-to-1 personalization

Personalization has been around for awhile, but how it works and what it can achieve varies enormously. Often the difference in performance is due to the technology itself.

Many vendors claim personalization capabilities, but few are actually able to deliver truly 1-to-1 personalization. This means they never achieve the best return on investment possible.

Precise 1-to-1 targeting, for example, displays the right product to the right person at the right time. In other examples it may be that the technology is not sufficiently well deployed nor is it configured and customized to accommodate the browsing behavior of individual sites.

New personalization strategies

These days websites should not be offering a one-size-fits-all experience, where every visitor sees the same thing. 70% of visitors, in fact, expect some form of personalization when they visit a site.

Busy customers increasingly demand content that is already filtered to their preferences and interests. They have a growing expectation that sites should know about them and engage them in a relevant way.

Personalizing behind the scenes

Explicit personalization, where the visitor is aware that an element of the page is personalized, is already widespread. This includes, for example, product recommendations that appear on product pages under a “you might also like” title.

You can now also use implicit personalization, where the visitor is not aware of what’s being personalized for them. This could be done for any content, such as images, banners, messages and promotions, based on what is already known about the visitor.

You might want to show a new visitor an offer that you don’t want an existing customer to see. Or a site selling male and female apparel could show returning women a homepage with female images displayed.

A building supplies site serving trade and consumer visitors could ensure that each customer type only received offers relevant them. These are just a few examples of personalizing the customer experience without the visitor being aware of it.

Using relevant landing pages

It’s also important to consider how visitors arrived at a site, whether through search, a pay-per-click ad or bookmarked favorites. You can then direct traffic to landing pages with supporting products to help maximize conversions.

A good example of great personalization would be a shirt retailer who knows which customers like buttoned-down collar shirts, then shows them those shirts on a landing page supporting a seasonal sale.

Customizing merchandising rules

But for all this ability to personalize, there will be overriding objectives that a site will want to achieve and that personalization activities need to support. This involves creating rules that typically sit on top of the personalization output and change what the personalization technology originally prescribed.

Examples would include restricting product recommendations to show full price products, giving increased visibility to sale stock by suppressing a personalized recommendation, or avoiding brand conflict on a page.

For example, Armani would want to be assured that a Gucci product never appeared on a luxury retailer’s product page showing an Armani product.

Understanding the visitor’s intent

The mix of personalization data available will also vary from visitor to visitor. Contrast a visitor who has visited a site ten times in the last week with a returning visitor who last visited the site over a year ago. Both instances call for different strategies to be employed in surfacing not only relevant product recommendations, but content that is most likely to lead to a sale.

And while previous browsing history is useful, understanding the intent of the visitor in the current session is key to steering the visitor to the products they are most likely to purchase.

Creating personalized, omni-channel experiences

Multichannel experiences involve many joined up channels through which visitors engage with brands. But historically, each channel represented a silo of data and behavior that were not connected to each other. In-store data was not shared with online data, for example, and vice versa.

Omni-channel takes visitor experiences one step further. It represents the development of a coherent strategy where data and learning in each channel are shared and consolidated. It presents a single view of the business to each visitor, regardless of the channels they decide to use.

While retailers think in channels, customers don’t. They pursue whatever they are interested in according to their time, convenience and availability. They can buy everywhere, and in many ways and that brings both challenges and opportunities.

Engagement across channels

Retailers can now capitalize on personalization in all the channels that customers engage with, including onsite, in-store, mobile, call centers, emails, marketplaces and print.

This omni-channel environment is complex, but understanding customers’ buying preferences and their ways of multiple engagement is critical to sustaining customer satisfaction, loyalty and long term value.

The process of shopping is changing

For years marketers have talked about a single view of the customer as being the holy grail of merchandising. You would be able to fully understand all the ways each customer has engaged with your business, then use the data gathered to inform and enrich the relationship.

Now that a single view is an emerging reality, there has never been a better time to personalize and reap the rewards. Here are a few of the ways personalization is crossing channels:

  1. Personalized digital receipts
    Digital receipts issued in store are becoming more commonplace, and when sent to an email address instantly tie offline purchasing data to online browsing activity. This means you can incorporate personalized product recommendations on the digital receipt itself, and introduce more products that can be purchased on or offline.
  2. Personalized delivery notes
    When a product is dispatched, the physical delivery note can now be enhanced with personalized product recommendations that are relevant to the recipient. This means the sales cycle can begin again through whatever channel suits the customer.
  3. Personalized in-store experiences
    Highly contextual and personalized messages can be sent to a smartphone when a customer is in your store.
  4. Digital mirrors
    Digital mirrors are being introduced into changing rooms. With digital mirrors, customers are able to scan the barcode of an item they wish to try on, then on the digital mirror a set of recommended products will appear that are relevant to the product being considered. These can be alternative cross sells or complimentary upsells.

Consistent customer ID tracking

To deliver the potential of omni-channel, however, requires consistent customer ID tracking across all devices and channels. A customer might browse a site on their mobile on their way to work, or they might look again on their PC at work, then ultimately use their laptop at home when they decide to purchase.

All these devices need to be accurately identified as being related to an individual customer. This means that however and wherever a customer chooses to engage there is a single repository of data about how they like to shop. This ensures that every personalization initiative is precisely targeted.

Personalizing product and basket pages

In this section we look at where personalization is best deployed in the customer journey and some tips about how best to present products and content. Firstly, it’s important to understand how a visitor arrives at a site. The use of search, pay-per-click ads and social media means that typically customers are entering websites on category, product and editorial pages.

The homepage, however, is the place where loyal shoppers are likely to enter. Using this space to insert personalized product recommendations and personalized messages has proved highly effective, increasing click-through rates by over 15% when compared with click-through rates for non-personalized homepages.

Getting above the fold

Secondly, product recommendations need to appear above the fold since 60% of customers won’t scroll down. This is less so when accessing a site using a smartphone, as users are more likely to scroll there, but the best product recommendations still need to be visible as soon as possible.

Presenting products in a skyscraper widget rather than landscape makes this potential problem disappear.

Optimize every page

Thirdly, consider all the page types for some element of personalization. Retailers tend to consider the product and basket pages but these account for only about 45% of all shopping sessions. That leaves a massive 55% not exposed to some form of personalization and therefore not receiving some extra encouragement to purchase.

Customers may enter the site through many points, representing a clear opportunity to optimize every page. Research shows that where over 80% of all shopping sessions are personalized the incremental sales achieved are 10%-20% greater than on sites personalizing only 45% of shopping sessions.

Personalizing page types

Experience tells us that different configurations of personalizing and merchandising strategies are effective at different customer points. At any given moment, the depth of behavioral data available for each visitor will vary considerably, and this needs to be reflected in the strategies selected at each point.

Here we look at some of the different strategies used on generic homepages, while recognizing that there can be many variations depending on the behavior data profile of each site.

To drive the strategies that deliver the most relevant product recommendations, for example, Episerver Perform uses over 100 algorithms that help determine what a visitor will be most likely to purchase.

The homepage

Visitors to this page can be divided into two categories: new and returning.

Visitors who have not visited the site before, and for whom there is no behavior data as yet, might be served product recommendations based on a number of merchandising strategies.

However, Episerver’s personalization engine begins working immediately to understand the intent in the current shopping session and serve products accordingly.

New visitors

  • Popular items from the most popular categories based on:
    • Conversion
    • Units
    • Revenue
    • Click-through rate
    • Product view
  • New products added to the catalogue
  • Emerging items (products trending in last seven days)
  • Most purchased and viewed categories from popular categories

Returning customers and visitors

Returning customers who have previously browsed or purchased present clear opportunities to personalize the next shopping experience.

Personalization strategies might include

  • Abandoned basket content and alternatives
  • Targeted discounts on recently viewed products
  • Recently viewed products
  • Recently bought and cross-sells
  • Products based on categories the user has recently viewed or purchased from
  • Cross-sell products based on recent and previous purchases
  • Popular products from previously browsed categories
  • New products based on categories the user has recently browsed or purchased

In all cases, it is the combination and configuration of all the strategies working together that produces the best results. Episerver’s personalization engine, for example, may identify ten products that meet the strategies selected in any one instance.

If the retailer can only serve four products on a given page, it will choose those products with the highest propensity for purchase. If the customer doesn’t click on what has been served, products with a higher propensity score will be shown the next time the page is visited.

The ability of Episerver’s technology to automatically self learn is crucial, as it constantly adapts to finding the best products to present to each visitor without any merchandiser intervention.

Case: Hawes & Curtis

  1. The homepage presents current promotions and new, exclusive products.
  2. A widget shows recommendations based on each visitor’s behavior and preferences.

Category pages and product listing pages

After the homepage, the next stage in the shopping journey is probably the category page. Again, consideration should be given to treating new and returning visitors differently.

But firstly, at the category level, the page can be further optimized by reordering the display of thumbnail images, based on what is known about the individual visitor. This means that the most relevant products appear first and at the top of the page, in the best possible and most visible position.

This is particularly effective where the visitor is using a mobile phone. With limited screen real estate, it’s essential to show the most relevant products first. Although mobile browsers are prepared to scroll more than desktop browsers their attention span is still limited.

Strategies adopted might include:

  • Variant colors prioritized based on user preference
  • Abandoned items
  • Recently viewed
  • Popular items from users’ favorite subcategories
  • Last week’s popular items (by units)
  • Category variety emphasized for new visitors
  • Best sellers

Case: Cooksongold.com

Cooksongold displays eight widgets on each category page. Each widget represents a different personalization strategy, such as sale items, new products and popular items.

Product pages

Product pages are perhaps the most important pages on a site, given that visitors will spend the most time browsing and choosing between products to add to their basket.

Important considerations here include how many recommendation containers (widgets) to include, and how many products to display. Widgets displayed as right-sided skyscrapers offer the best visibility and outperform widgets placed at the bottom of the page, particularly where they appear below the fold. No amount of creative input or intelligent personalization is going to work if the recommendations are invisible to the visitor.

It’s also important to have a clear strategy here: whether the goal is to offer complimentary items to increase the average order value or whether to offer alternatives to help improve conversions.

Most retailers wanting to achieve the best of both worlds will use a hybrid of strategies that can be A/B tested to establish the best mix.

Strategies adopted might include:

  • Alternative products
    • Recently viewed
    • Most recently viewed/purchased products from the same category
    • Products linked to past browsing and purchases
    • Most strongly associated products which encourage conversion
  • Cross-sells
    • Complementary products from other categories
    • Outfit build products
    • Frequently bought products from a different category based on the product viewed
    • Products linked to past browsing and purchases
    • Recently viewed products form a different category

Case: MissSelfridge.com

  1. The widget on the left presents items that match the style of the item viewed.
  2. The widget on the right presents items that are alternative choices for the visitor.

Basket pages

With basket pages, it is essential to help the customer to complete their purchase successfully and not distract them with more products that have no relevance to what’s already in their basket.

With the right care here, it’s possible to increase the units per order and consequently the average order value. Tests measuring recommendation revenue derived from the basket page reveal that it can account for 5-10% of total recommendation revenue achieved.

Strategies to surface the appropriate products might include:

  • Products from a different category based on items in the basket
  • Subcategories of products considered as add-ons, accessories, gift vouchers
  • Cross-sells frequently co-purchased with the basketed product
  • Display a highly merchandised set of products only according to trading requirements.

Case: Toa.st

  1. A duplicate of what would be seen in the product page is presented underneath the basket recommendations. This provides users with the necessary information and allows them to choose a size and add it to the basket without leaving the page.

Zero results page

These pages are a missed opportunity if the page only returns “No Search Results”. Click-through rates can be as high as 22% using simple strategies such as:

  • Products linked to search keyword or phrase
  • Recently viewed with variety
  • Popular items from users favorite categories
  • Popular items from popular categories, variety promoted
  • Site wide best sellers

Case: Cooksongold.com

  1. Cooksongold keeps it simple by offering a selection of products based on user behavior. The widget appears above the fold, which allows users to click through and continue browsing or to search again by rephrasing their search.

Personalizing landing pages and emails

Personalizing landing pages are a proven method for generating the best returns for online marketing campaigns.

There are many different types of landing pages. Essentially, a landing page is a web page a visitor arrives at after expressing interest by clicking through from one of several forms of advertising:

  • An organic search engine result
  • PPC campaigns such as Google AdWords
  • Google product listing adverts
  • Banner advertising
  • Links in emails
  • Print advertising
  • TV commercials

The purpose of the landing page is usually to convince the visitor to convert to a customer by taking a specific action. Personalizing landing pages increases the likelihood of conversion in two distinct ways:

  1. Where personalization data is available for a visitor, the landing page will not only include the message or product that encouraged them to visit but also additional products that are relevant to their preferences.
  2. If a visitor is not known, products can be displayed that are associated with or have an affinity to the product that caused their initial interest. At its simplest level, an example would be “here are some products that were also viewed by people who viewed the product that brought you to the landing page”.

A good example of personalized landing pages in action is reflected in our work with Hawes & Curtis, supporting product listings adverts (PLAs) in their Google Shopping campaigns.

Directing customers to the right product

Hawes & Curtis recognized that PLA campaigns drove qualified traffic to its website and could include narrower search terms to help direct the customer to the right product at the right time. To increase the relevancy on the PLA landing pages, we worked with Hawes & Curtis to personalize the page and improve conversions with a dynamic widget. This offered up to eight alternative products.

These recommendations were based on agreed configuration strategies for the particular product advert, which is optimized and tailored to an individual’s behavior and preferences. From the customer’s perspective, it meant that the landing page they were shown was more relevant to them.

Personalizing without data

If no personalization data is available for an individual because they haven’t previously visited the site, default merchandising strategies are activated such as “Relevant similar products.”

Personalizing the landing page in this way resulted in a 10% reduction in the bounce rate. Even more significantly, it led to a conversion rate increase of 26% and a revenue per session increase of 32%.

The company’s head of ecommerce commented:

Brand engagement is a critical success factor in all our paid search campaigns. The Episerver PLA widget has enabled us to offer a better customer experience by showcasing additional relevant products on the landing page, which reduces the bounce rate.

In this way, Hawes & Curtis are able to show their customers that they really understand them and their preferences, while demonstrating to the business the benefits personalized recommendations have on their conversion rate.

Personalized search

It’s been mentioned earlier that personalization helps visitors to find products more quickly that they are interested in buying. For large catalogs, personalized search can help visitors to discover products that they didn’t know they were interested in until they saw them. This was one reason why apparel sites were early adopters of personalization as visitors often had only a vague idea of what they were looking for.

With the incredible growth of shopping on mobile devices, retailers have been seeking ways to filter and refine search capabilities that help visitors find products on relatively small screens.

Using behavior data in search

The onsite search capability within the Episerver platform incorporates behavioral data. This means that search results are influenced and ranked according to an individual’s browsing and preferences.

For example, if an individual inputs “red shirt” and has happened to browse buttoned down collars previously, red buttoned down collar shirts will appear first at the top of the search results.

Personalizing the search function saves time, reduces the length of the customer journey, and promotes really satisfied customers.

Navigating through category pages can also be fully optimized by reordering the thumbnail images, based on what is known about the individual browser.

Personalized and trigger emails

Behavior data gathered online, capturing what customers do and don’t do, is very effective in driving increases in revenue through email campaigns.

The wealth of personalization data available to support personalization strategies will vary considerably from customer to customer. One customer may have purchased recently, another may have abandoned a basket or recently viewed some products, while another may not have been on the retailer’s site for some time.

Adapting emails to available data

When mailing a customer database, the system that generates the best products to recommend needs to be able to adjust to the data available for each customer. It needs to manage what personalization data is available and use the best strategy for generating a sale. When no data is available, the system needs to default to merchandising strategy options.

When it comes to personalization strategies, the Episerver Advance dashboard is truly unique in the market. It enables users to create personalization campaigns using a comprehensive selection of up to 25 personalization and merchandising strategies.

How Episerver Advance works

In order to optimize which personalized products to display, users can create a cascade of multiple strategies. If the data is not available to support a particular strategy, Episerver Advance will automatically cascade to the next selected strategy to find a suitable product, and so on.

Ultimately if no personalization data is available relevant to that email address, the user can either create a default manual list of products they want to promote or select some merchandising strategies to ensure that there is no blank space within the body of the email.

This guarantees that the best products with the highest propensity for purchase are presented to each individual in an email.

How to get value out of what customers don’t do

Trigger emails are very effective in getting customers to complete an action that they didn’t complete when they were on a website. Episerver Advance provides behavioral targeting triggers that can be sent as batched emails or streamed to go at a specific interval after the on-site action occurred.

Streamed emails would include abandoned basket or checkout actions, where the customer failed to complete a purchase or stopped browsing. Emails showing what was abandoned and alternatives to what was abandoned see click-through rates as high as 30%, with click-to-purchase rates averaging around 20%.

Batched emails would include targeted discounts, where customers who previously viewed products but didn’t purchase can be sent an email when the product is reduced in price. M&Co saw email conversion increase by 55% using targeted discounts.


Episerver uses cookies and browser technologies for analytics, personalized content and ads.