SCHEDULE A CALL WITH A PERSONALIZATION EXPERT!

All About Personalization in Retail: KPIs, Challenges, and Best Practices

Gabriella Vas Gabriella Vas
13 min read | December 17, 2021

In the first part of a miniseries examining recommendation systems in each main industry of application, we’ll be looking at personalization in retail – ecommerce as well as brick-and-mortar stores. We’ll identify the metrics used for understanding and measuring the added value of recommendation systems in retail. We’ll list some personalization challenges unique to the field of retail, as well as a handful of proven tactics to address them. 

Interested in joining our staff?

Check out our open positions here.

Why Is Personalization So Crucial in Retail?

The vital role of personalization in retail (and particularly in ecommerce) is basically rooted in abundance. 

On one level, there’s an oversupply of e-retail brands which means competition is tough and users are fickle. Being able to hook them with a tailored experience through personalization is a powerful competitive advantage. Research attributes a 20-percent leap in customer satisfaction and up to 15 percent more sales conversions to positive user experiences. 

On another level, most web stores carry a vast array of products – too many for a user to take in on their own. The resulting paradox of choice creates the need to pare down options to a reasonable range, with speed and scale only AI can manage. 

Retail, and especially e-retail is very fragmented in terms of business models, markups, and the resulting challenges, but personalization is indispensable in every area.

Some sectors – electronics, for instance – compete primarily in price, and profit margins are relatively low. Here, personalization can take the edge off shoppers’ price sensitivity by providing an experience so delightful that they’ll be willing to pay slightly more for it.

Other sectors, like fashion, have significantly higher markups, but on the flip side, no tangible differentiating factors like price. In this case, again, it’s users’ experience – not just of the purchase process but of associated services like delivery or returns – that can make or break a deal. (To increase the challenge, in retail fields like fashion, shoppers’ taste adds a subjective layer to purchase decisions, and specialized algorithms are needed to cater to them. This explains why vendors of personalization in retail can be further segmented based on the fields they serve.)

From the perspective of marketers looking at the conversion funnel as a whole, the benefits of personalization in retail are even more obvious. It’s not enough to generate leads to a web store – it’s just as important to convert visitors once they’re there, in large part by showing them the right recommendations.

The purchasing process itself has several potential breaking points, “sinkholes” where customers can vanish if their user experience isn’t just so. The main page, search and browsing tools, product detail pages, and especially checkout are all rife with the risk of losing a customer and all the investment that went into guiding them so far along the conversion funnel. So personalization plays an important role in making sure all these hurdles are cleared through a consistent, tailored user experience. 

Nowadays, e-retailers are all too aware of these challenges. For them, the question is no longer whether or not to use a personalization platform – the case for better UX through relevant recommendations is clear. The question is rather what algorithms to use; what data to leverage; which vendor to choose; and which points of the user journey to focus on. 

The KPIs of Personalization in Retail

Retail personalization can create added value in a variety of ways: by boosting a web store’s traffic; by increasing basket size or value; and by generating more revenue, not just from core retail functions but from complementary services like product insurance or delivery. 

The key performance indicators (KPIs) of personalization in retail are metrics used to assess the added value of recommendations, or to compare two recommendation systems in an A/B test We can also measure the conversion rates of recommendations at various points of the shopper journey – from main page to category and product detail pages to cart page to post-purchase follow-up emails – and compare the results to see where recommendations work best. 

The “Holy Trinity” of personalization KPIs in any industry are:

  • Number of recommendations –  the number of times the recommendation engine generates a recommendation for the users. (To be precise, this metric is used for measuring a personalization solution’s traffic, not its performance.)
  • Number of clicks – a natural number that indicates how many times the users click on items recommended by the recommendation engine. Clicks convey primary feedback from users on the relevance of the recommended products.
  • Clickthrough (CTR) – the number of clicks on the recommended products divided by the number of recommendations displayed, expressed in the form of a percentage. CTR indicates the relevance of recommendations and, indirectly, the quality of the user experience. 

In order to get the full picture, consider these KPIs specific to personalization in retail:

  • Number of transactions generated by recommendations. A transaction (in case of retail, a purchase) is considered to be generated by a recommendation if it was made within a certain timeframe (generally, 24 or 72 hours) after the user clicked on the recommendation. 
  • As opposed to the absolute metric described above, Transactions generated by 1000 recommendations is a relative metric indicating efficiency. As such, it can form the base of a comparison to the performance of other recommendation systems. 
  • Revenue generated by recommendations and, for the same reason as above: Revenue generated by 1000 recommendations
  • Recommendation share – the percentage of the total revenue that is generated by recommendations
  • Buy counts and buy values – the total number of purchases and the total revenue made in a web store over a certain period of time. These two metrics are used in A/B tests comparing a personalized setup with a non-personalized one.
  • Average cart size and average cart value – these measure the efficiency of cross-sell and upsell algorithms, key features of personalization in retail. 
  • As mentioned earlier, it’s possible to measure conversion rates at various stages of the purchasing process: the number of item page views, add to cart or buy events per total visitor number. In a converse way, abandonment rates at these stages also describe a recommendation system’s performance.

Needless to say, there is no “one-size-fits-all” solution in retail personalization. Every recommendation system has to be precisely tailored to the client’s domain, brand message, site structure, and a host of other defining aspects. Many of these factors are beyond the control of the personalization vendor, so measuring the performance of recommendations is essential for fine-tuning the platform. This is why setting up a SaaS recommendation system takes a few weeks of coordinated effort from vendor and client.

The Playbook of Personalization in Retail: Challenges and Best Practices 

Recreating the Ideal In-Store Experience Online

In a brick-and-mortar store, service plays a major role in shaping user experience. Well-trained sales assistants can understand a shopper’s needs and preferences based on a brief welcome exchange, and subtly alter their approach and level of service accordingly. In the online sphere, recommendation systems are meant to stand in for competent sales clerks, but in many ways, their task is different, and possibly more difficult. 

In a virtual encounter between AI and human, it’s harder to gauge 

  • the shopper’s intent; whether they know what they’re looking for, or need advice to get started; 
  • their level of knowledge about the desired product or category; 
  • the driving force behind their purchase – whether it’s emotions or a rational consideration; 
  • how they deal with the paradox of choice – whether they’re maximizers, painstakingly looking at every single option before choosing the very best; or satisficers happy to settle for a good enough choice and cut the shopping short;
  • how much time and money they can spend; 

and a slew of other factors that determine the best way to assist the given shopper in their endeavor.

In order to overcome these hurdles, personalization in retail can fall back on two tricks of the trade: 

  • Advanced search. For shoppers who have at least a vague idea of what they’re after, search is an indispensable tool. Today, recommendation systems offer sophisticated search functions, going well beyond the capability of any sales clerk to retrieve just the right selection of relevant products. For instance, an autofill feature completes search entries as they’re being typed. Dynamic search lists personalized results instantly. Suggested filters are personalized, and so is the order of search results. 
  • Site personalization. To visitors who start out by “just browsing”, a personalization engine can generate recommendations on several levels: entire product categories, products within a category, or various SKUs of the same product. If, let’s say, there are ten categories in the navigation bar on the main page, even those can be listed in a personalized way, placing more relevant ones on top. To loop back to the sales clerk comparison: the offline equivalent of such level of service would be reorganizing the entire store for each and every visitor.

The “Omnichannel” Buzzword

Providers of personalization in retail often refer to omnichannel recommendations. However, interpretations of the term vary, and the ensuing confusion only helps to elevate “omnichannel” to buzzword status: no one is quite sure what it means, but everybody wants some. 

From users’ point of view, omnichannel personalization means being able to continue a purchasing process from one device or platform to another – to browse, select, compare on mobile and buy on desktop, or the other way around. For the retailer, it means the ability to track users across digital channels, in order to create a consistent user experience on web, on mobile and in-app. In yet another interpretation, omnichannel includes physical stores as well as digital platforms. 

Technically, omnichannel simply means that (1) one main backend handles all the customer data whether on web, on mobile, or a in a brick-and-mortar store, and that (2) websites, email offers, social media posts and physical stores all display the same messages, offers, and products. 

Herein lies a challenge unique to personalization in retail: how to sync user data from physical and virtual stores and create 360-degree user profiles to provide a consistent UX offline and online, on screens large and small. Two proven best practices:

  • Loyalty program. A reward scheme for registered users is a win-win: not only does it encourage loyalty, it also helps to connect the data representing a user’s online and offline behavior. As such, it’s a huge opportunity for personalization in retail. 

Without a loyalty program, the link between online and offline user activity cannot be established. The data collected separately in-store and online can suggest global trends, attributable to locations, but not to specific users. This information can also be used to generate recommendations, but its personalization potential is definitely inferior to that of a loyalty program. 

As for online-only retailers, a loyalty program is less relevant from a personalization perspective; its role is primarily attracting and retaining high-value customers.   

  • Mobile app. For users, a retail brand’s mobile app helps to extend personalization into the physical realm – for instance, by suggesting deals available exclusively in-store, based on the user’s browsing or purchase history. For retailers, the mobile app is an invaluable source of offline user behavior data – for example, which in-app discount coupons a shopper redeems at checkout.  

Curious about a real-life example? See our Cora case study to find out how Yusp helped the retail giant bridge the gap between its online and brick-and-mortar stores by combining the benefits of a loyalty program and a mobile app. 

The Session-Based Recommendation Problem

E-retail can hardly be considered the land of steady habits: online shoppers are an unpredictable lot, with ever-changing agendas and preferences. The product they’re after today may be irrelevant to them tomorrow, and that’s bad news for personalization engines that rely heavily on users’ browsing history to generate recommendations.

In addition, user tracking technologies are often unreliable when it comes to identifying users across devices and platforms. Third-party cookies fail to provide a consistent experience to users who decline them; at the same time, they may pose a privacy risk to those who accept them. One user identification tool that is universally acceptable is the first-party cookie, but it only tracks activity within a given browsing session.

The bottom line: in order to stay relevant while complying with privacy regulations, e-retail personalization often has to make do with in-session data.

There is but one, fairly new solution to this problem: deep learning algorithms. As opposed to more conventional collaborative filtering, these algorithms have been developed to provide users with relevant recommendations based exclusively on their in-session behavior. 

This is made possible by deep learning algorithms’ capability to glean information from multiple layers and categories of data. As a result, these algorithms can identify products in rich detail (even so-called cold start items, with no previous interactions or ratings), and recommend them to potentially interested users.

Shopping Cart Abandonment

According to current ecommerce stats, about 70 percent of all items placed in carts are never purchased. In addition, browsing abandonment occurs frequently as shoppers exit web stores even before adding anything to their cart. All this means a massive loss of revenue for e-retailers and a key challenge for personalization in retail. 

The most common reasons for users leaving a web store without completing the purchase are extra costs for shipping, taxes or fees; the checkout process being too long or complicated, including the need to create an account; the website crashing or raising concerns about trustworthiness; or payment issues. Often, cart abandonment is simply a by-product of today’s shopping habits: people add items to their carts just to be able to compare them or to save them for later, without ever intending to buy them in the first place. 

Browsing abandonment occurs typically when a shopper researches options for a high-involvement purchase like a laptop; having made their choice, they’ll close all other tabs featuring the alternatives. But this research and comparison process can be demanding for many users: sometimes, they’ll grow exhausted, confused, or distracted and give up, at least temporarily, before completing the purchase.  

For the e-retailer, the first step is to identify the last point of conversion along the user journey to understand where it was interrupted. Was it right after the search results came up? Was it after viewing a category page? Or did the user get as far as selecting a product, maybe even adding it to cart? 

Next, the possible causes of browsing or cart abandonment need to be investigated. Some of these factors may fall outside the control range of personalization. If a large chunk of cart abandonment cases can be traced back to errors in the payment system or complications in the checkout process, these need to be sorted out first, or personalization efforts won’t make much difference. 

Once these primary reasons have been dealt with, however, there are plenty of personalization tactics that can further reduce churn at checkout or even earlier in the purchase process. Some examples:  

  • Triggered messages. Recommendation systems can detect patterns in user behavior signaling that they’re on the verge of abandoning the purchase process – like when they become inactive on a page, or fail to complete the next step at checkout. At these critical moments, the system can send the user emails or instant notifications, made relevant with personalized recommendations. 
  • Exit intent popups. Similarly, when a user is ready to leave the web store (for instance, their cursor movement suggests that they’re about to close the tab), an overlay with personalized, exclusive deals can persuade them to change their mind.
  • Retargeting banners. Once a user has left the web store empty-handed, it’s still not too late to lure them back, by displaying on third-party websites banner ads of the products they have been considering. 

A Few More Things to Keep in Mind

Remember, personalization only generates added value if it’s visible enough. E-retailers have to adhere to a consistent personalization strategy and display recommendation widgets prominently, near the first fold, in order to see return on investment. 

Furthermore, a recommendation system needs to be constantly optimized to keep up with trends and evolving industry standards. In other words, personalization is not a one-off, static solution – rather, it requires tireless tinkering. Nowadays, most vendors of SaaS personalization platforms include optimization in the package. Still, the quality of research and development that supports this function can be an important differentiating factor when comparing potential providers. 

At Yusp, we have amassed valuable knowledge about personalization in retail over the years. Got any questions on the topic? We can help.

Interested in joining our staff?

Check out our open positions here.

What to read next

Join our newsletter

Get to know the ins and outs of personalization