Retail Personalization in 2024: The Ultimate Guide22 min readReading Time: 15 minutes
It’s 2024 and the word “personalization” has been thrown around for 5 years with businesses across segments implementing it to varying degrees of effectiveness. Retail has had a complicated relationship with personalization. Some retailers have been early movers with companies like Amazon and Stitch Fix making retail personalization a part of almost every part of the shopper journey. However, many retailers have been slow to adapt, offering bad experiences to shoppers on their websites. Even those who have implemented personalization in retail haven’t done it right.
74% of customers feel frustrated when website content is not personalized –Instapage
Shoppers are moving from retailer to retailer with the knowledge that there will always be a better deal out there. But giving discounts is not the only way to make shoppers come back to the site. It is neither sustainable nor profitable. But what else are shoppers looking for?
Shoppers are willing to pay up to 16% for personalized shopping experiences –PWC
Shoppers expect personalized experiences rather than be flooded with a slew of products. They want their favorite brands and websites to know them as well as they know the brand. Brands are now expected to tailor experiences taking into account each person’s individual style preferences.
But do brands really understand retail personalization?
What is Retail Personalization?
Retail personalization is explained beautifully in this quote by Jeff Bezos.
“If we have 4.5 million customers, we shouldn’t have one store. We should have 4.5 million stores.” –PC World, June 28, 2000
Retail personalization is the process of providing every shopper with a unique journey across every single touchpoint and channel, based on historical data and real-time shopper intent, powered by customer and product Intelligence. The ultimate goal of personalization in retail is to make shoppers feel unique, special, and emotionally connected, to improve their shopping experience.
While almost every retailer today claims to “personalize” their shopper’s experience, very few do it right.
“63% of consumers will stop buying from brands that use poor personalization tactics.” – Smart Insights
What is NOT Personalization?
While the term ‘personalization’ is thrown loosely around, not every retailer has managed to utilize its full potential. In most instances, ‘segmentation’ and ‘personalization’ are used interchangeably, despite both being two independent techniques.
Traditional shopper segmentation treats people as a cohort – a group of people with similar tastes and interests. It focuses on categorizing, labeling, and boxing people, instead of treating them as individuals with specific style preferences and needs. It offers a rather myopic view of categorizing based on limited data points that usually are just broad fragmented categories.
When the two terms are interchanged, we hear cases like a shopper being recommended mosquito nets after purchasing one a week back. Or a person waking up to a dozen of promotional emails about baking trays after buying an oven. These examples don’t imply that they aren’t ‘personalization’. It implies that they are ‘irrelevant personalization’, which would be regarded as a failure.
For example, out of the 100% oven buyers, there might be a percentage that needs trays. However, sending all of them a prompt to buy ‘tray’ reflects that the brand has put everyone from that group under one category, instead of mapping their individual needs. This illustrates the outcome of confusing segmentation with Retail personalization.
Real personalization requires brands to invest in AI & automation and prioritize it (not just on paper) to individualize their strategy, understand shopper intent, and provide relevant recommendations that add value in real-time.
Brendan Witcher, VP & Principal Analyst at Forrester Research disclosed that 90% of organizations are going to invest in retail personalization but only 40% of consumers say that the information they get from a brand is relevant to them. He added that one of the reasons for this gap between the brand’s effort and actual results is ‘segmentation’ in the garb of ‘personalization’ as it creates a wrong experience for consumers
68% of shoppers are unlikely to return to a website or store that doesn’t provide a satisfactory customer experience. –Forrester
In a nutshell:
Segmentation is NOT personalization
With Forrester advocating the move to 1:1 Personalization, retailers need to start looking at new ways to connect with their shoppers, keep them engaged, and, eventually, turn them loyal.
What is 1:1 Personalization?
Every eCommerce site has 3 components: shoppers, products, and actions the shoppers take.
Let us take the case of 10 shoppers shopping from the same eCommerce site which sells 30 dresses. For every product, there are around 10 actions each shopper can take, which are click to view (to view the product description page), buy now, add to cart, add to wishlist, change the size, change color, click on a similar item (from the items similar to this widget), click on a frequently bought with item, back (the back button on the browser), close (closing the tab), etc.
10 shoppers can each take any of 10 actions on each of the 30 items. This means 3000 possibilities can happen.
Typically, eCommerce sites attempt to bucket shoppers who take similar actions into a segment, For example, if 3 out of the 10 shoppers added dress #8 to the cart, they will be put in one segment. Further, the eCommerce retailer will retarget these shoppers with ads across all channels with images of dress #8 and dresses visually similar to it. Sounds logical, right?
But what if I told you that dress #8 was a knee-length, turquoise blue dress with a scoop neck and shopper #1 had clicked on every knee-length dress on the site, shopper #2 had clicked on dresses in different shades of blue, and shopper #3 had clicked on scoop neck dresses? This means that 3 different shoppers liked the same dress for 3 completely unrelated reasons.
If the eCommerce retailer captured the affinities of each of the shoppers, they would have sent shopper #1 knee-length dresses, shopper #2 blue dresses, and shopper #3 scoop neck dresses. This affinity could have been captured if every click of the shopper was tracked.
Every shopper is a unique individual. Every shopper is their segment.
Shoppers, products, product attributes, and actions have to be mapped at an individual level to create a unique Shopper Profile for every shopper. This can be used across channels to create a unique journey for every shopper.
This is 1:1 personalization.
But how is it possible to achieve 1:1 personalization? Even in this limited example with 10 shoppers, 30 dresses with 3 attributes, and 10 actions, there were 3000 possibilities. On an average eCommerce site with thousands of products with numerous attributes and thousands of shoppers, the number of possibilities balloons up to the millions! Now think of the giants with millions of shoppers. Seems like an impossible task right?
This is where AI comes into the picture.
AI-powered Retail Personalization
Retail personalization, powered by AI delivers a unique experience to every shopper optimized for greater engagement and higher conversion. Every shopper sees a distinct version of the eCommerce site which is dynamically personalized right from the first click. This experience is then delivered at scale to every single shopper across all touchpoints and channels.
But how is this experience built?
Good data: the prerequisite to Personalization in retail
Data is well on its way to becoming its currency, and that is why retailers need to be front and center in its creation, its management, and its deployment. Data also forms the foundation of inventory management. From the moment inventory lands at the warehouse to when buyers place orders for the season to come, data is the crucial link that connects the inventory management cycle and leads the decision-making process. Data accuracy is necessary from the start of this process to ensure that subsequent decisions are taken in an appropriate and timely manner.
2.5 million quintillion bytes of data are generated every single day.
Data needs to be both precise and accurate to aid in every step of the retail process. But a team creating and managing this data makes it susceptible to human error and inconsistencies. This where AI comes into the picture.
AI-powered Retail Automation ensures that this data creation, in the form of automated product tagging as well as collating individual customer’s preferences. Using AI in Retail additionally ensures the creation of an accurate database that is unique to the retailer’s business and customers.
How is this data used?
This database, when distributed across the retailer’s value chain, is a proverbial goldmine. It can help retailers create shopper profiles for every shopper.
Shopper profiles capture visual affinities (Eg-color, pattern, shape), and non-visual affinities (Eg- brands, categories, price) to build a comprehensive understanding of every individual shopper. These are what go into the Shopper profile of each shopper.
- Behavioral cues: what the shopper is clicking on, what products they’re adding to the cart, the pages they’re looking at and the pages they’re bouncing off of
- Transactional data: what the shopper has bought in the past, what they’ve returned
- Demographic data
Not only does it provide cues for visual merchandising, but it also facilitates styling decisions and impacts buyers’ decisions for buying inventory for the season to come.
But is a shopper profile alone enough?
Take the case of Maria who is a new mother. Spring was also just setting in and she was excited to finally browse through cotton knee-length non-maternity dresses. Before her pregnancy, short knee-length dresses always used to be her thing, regardless of which season it was. Not just dresses that were knee-length, but ones with sleeves that covered her elbows.
But this time around, not only did she not find even a single cotton knee-length dress, but only maternity dresses were recommended to her when she logged on to her most frequented eCommerce sites leading her to look elsewhere for a dress.
Was this an algorithmic glitch? No, the personalization engine worked perfectly. Maria was shown maternity dress options based on what she historically browsed through and bought over the last nine months. Less importance was given to the shorter style of dresses she was recently looking at since there was simply not enough user data for the algorithms to pick up on.
Maria represents a large proportion of disgruntled online shoppers who expect online stores to provide omnichannel personalization that is tailored, dynamic and specific to their shopping needs in real-time. Maria also represents the precarious conundrum that AI (artificial intelligence) based omnichannel retail personalization engines face, which is finding the balance between data-driven algorithms and real-time contextual events that occur in a shopper’s immediate past, like Maria’s pregnancy, and long-term history that is not easily captured.
Personalization engine need to be sensitive to fluctuations in browsing and spending patterns due to life events. This boils down to understanding the true intent of a given shopper.
Contextual Shopper Intelligence
Computer vision-based dynamic retail personalization uncovers shopper intent using a variety of signals. Each click on a product allows the algorithms to clarify a shopper’s specific attribute preferences as well. When this is combined with historical behavioral patterns logged by the engine, it becomes possible for the engine to accurately map to shopper intent, highlighting the most relevant products using real-time personalization.
Collaborative and content-based filtering often provide irrelevant recommendations as they fail to look at shopper behavior in context. Dynamic personalization uses real-time and historical data, to deliver the right product, to the right shopper at the right time, even when a shopper is looking to make an ‘unusual’ purchase, for a gift or a special occasion.
While a good retail personalization engine can understand visual and non-visual cues to build shopper profiles for every shopper, and then combine this with contextual shopper intelligence to deliver the right product at the right time for the right shopper, this experience needs to be delivered to the shopper. This is delivered through recommendations.
Retailers have large inventories, making it difficult for shoppers to sift through the find exactly what they are looking for. Recommendations guide the right shoppers to the right products, helping them find what they want more easily.
Recommendations are used by almost all eCommerce retailers today. Amazon pioneered this almost 2 decades ago. Before Amazon became the juggernaut that made Jeff Bezos a household name, its recommender system was already helping customers find what book to read next.
Today, eCommerce retailers use a variety of product recommendations through different parts of the shopper journey, across channels.
They broadly fall under 3 types.
- Global Recommendations
These recommendations are based on global trends and insights. This is especially useful for new visitors landing on the site for the first time. It also helps shoppers find new products and categories which they haven’t explored before. As this strategy usually isn’t based on personalized data, it’s especially useful for new shoppers or for low intent shoppers who are just browsing around the site.
The recommendations used in this type are:
- Most popular:
This recommendation is used to showcase products that are the most popular on the site. This is found by assigning a weight to every interaction on a product, like buys, clicks, add to cart, etc. The products are then ranked and the top products are displayed. Another variant of this is the Most popular in category.
- Trending now:
This recommendation is used to push products that are currently popular rather than based on historical data. More weight is given to purchases made in the recent past rather than those made a long time ago.
- New products:
This recommendation displays the products that have been launched most recently. New products tend to get lost in the noise as they don’t get recommended due to a lack of historical data. This widget promotes the newest products allowing shoppers (both old and new) to discover them.
- Contextual Recommendations
Contextual recommendations are based on product affinities. These strategies rely on visual product attributes, such as color, style, the category it falls under to recommend products to shoppers.
The recommendations used are:
- Visually Similar Products:
Recommendations of products that are visually similar to the product that the user is looking at. This is based on colors, patterns, and other attributes like sleeve lengths, neck types, etc.
- Complete the look:
Products are described based on various visual and non-visual attributes like style, color, pattern, material, category, etc. Categories are then mapped to each other based on rules. For example, shirts are mapped with pants. Attributes are then mapped with each other. For example, the color black is paired with the color grey, or a busy pattern is mapped with a solid color.
Based on these affinities, and the shopper’s profile, the most relevant products that …“complete the look” are displayed to the shopper, increasing the AOV.
- Viewed together:
Based on browsing history and overall website trends, products that are viewed together are displayed as recommendations. This allows shoppers to explore products outside their usual style, opening up new possibilities.
- Personalized recommendations
These recommendations are based on the shopper and their affinities. They take the shopper data and apply product context to surface relevant recommendations for each shopper individually.
The recommendations used are:
- Inspired by browsing history:
The last 10 viewed products are taken into consideration. Based on these products and their attributes, product recommendations are obtained. These recommendations are then ranked based on a sorting logic and displayed.
- Top picks for you:
The entire user history is taken into consideration including all buys / adds to cart/page views etc. A user profile is created that has user affinity data in multiple dimensions. These dimensions include category affinity, brand affinity, affinity towards visual attributes (color/patterns/sleeve lengths/neck types, etc.). Based on this, products from the catalog are surfaced that are most relevant to that particular shopper.
- Dynamic personalization:
Taking into account every page view and click that is made, the weighting of the parameters used for computing similarity (color and pattern) will be adjusted dynamically in real-time to reflect the user’s preferences and show the most relevant products.
While it sounds like personalized recommendations seem to be the best type of recommendations, they don’t work on every page of the website. The eCommerce retailer needs to use the right combination of recommendations on the right pages to maximize engagement and conversions. But what recommendations work where?
Retail Personalization across the entire shopper journey
Different recommendation strategies need to be chosen and optimized for each page type, maximizing potential revenue.
- Homepage Personalization
Most shoppers enter a site from the homepage. Recommendations on the homepage often tap into global recommendation strategies, which showcase the best-selling products to new shoppers. For returning shoppers, personalized recommendations can help guide the shopper towards products they have an affinity for.
The 3 recommendations we use that have helped retailers achieve 4.5x better product discovery are:
- Top picks for you / Top picks (for new shoppers)
- Personalized trending / Trending now (for new shoppers)
- Inspired by browsing history
2. Category Page Personalization
Category pages are for shoppers who are looking for a specific type of product but don’t know which exact one they are planning to buy. They are exploring the options available and trying to find the right product for them. For a new shopper, global recommendations like “Most popular in category” can be used. For return shoppers, personalized sorting will help them quickly find products they have an affinity for.
The recommendations we have used to help retailers 4x the page views are
- Personalized sorting
- Most popular in the category
- Trending in category
3. Product Detail Page (PDP) Personalization
A shopper lands on a product detail page because they like a specific product or at least the attributes of that product. If the shopper likes that specific product and there is an opportunity to increase the AOV by recommending products that it can be styled with (for fashion sites) / bought together with (personalized based on the shopper profile of that shopper for return visitors). Or if the shopper likes a few attributes of that product, visually similar products can be recommended (for fashion sites) / similar products (personalized based on the shopper profile of the shopper for return visitors).
The recommendations we have used to help retailers achieve 3x conversions are
- Visually similar / Similar products
- Personalized styling and outfitting
- Frequently bought together
4. Cart Page
When shoppers add items to their cart, they have shown clear intent to purchase the item which means they are more likely to buy it than on the PDP. Here they can again be shown items to Style it with (for fashion retailers) / Frequently bought together to increase the AOV. This can be personalized based on the shopper’s profile for return visitors. Similar products can also be recommended to the shopper to increase the chances of conversion.
The recommendations we power on 100s of global retailer’s sites are
- Visually similar / Similar products
- Personalized styling and outfitting
- Frequently bought together
5. Post checkout
After the shopper has bought items from the site, the retailer can show other products that can entice the shopper to come back and buy more. Global recommendations like Trending products can help shoppers discover more products on the website.
The recommendations we have used to help retailers achieve 3x products viewed per session are
- Personalized styling and outfitting
- Frequently bought together
- Trending Now / Trending for you
Now we know what recommendations work where, how to identify shopper intent, and how to personalize based on the shopper’s profile. Combining all of this information to practically implement it on your website sounds hard right? There are so many metrics to constantly keep a watch on, so many different parameters to tweak, so many business rules to set and constantly optimize. eCommerce teams are increasingly expected to function at superhuman levels and be experts in marketing, merchandising, and technology. They end up using a dozen tools to manage customer experiences and these tools don’t ‘talk’ to each other. This adds to their workload and leads to more broken data.
A solution to one retailer’s problem is not a solution to another’s. Just as shoppers are all different and retailers need to understand their individual preferences and intent, retailers too came in various shades. A single, one-size-fits-all solution simply won’t cut it. The inventory a retailer carries, whether they have an in-house label, their brand strategy- all have a big impact on how a retail personalization solution should be designed.
Retail Personalization is no child’s play and it would be almost impossible to do it right without one platform that offers all of these levers to control at your fingertips. Without one single platform that controls all of this, eCommerce retailers would need a huge team of people whose sole job is to take care of personalizing shopper journeys.
Vue.ai’s Personalization solution is an AI-powered personalization platform
Vue.ai’s Personalization engine is an AI-powered retail personalization platform designed to help eCommerce teams manage the mammoth task of creating personalized shopper journeys that result in growth.
From a single, easy to use the platform, retailers can bring their strategies to life, deciding what journey different segments of shoppers should be taken on, ensuring every shopper sees content that is personalized to them, A/B testing the journeys to identify winning ones to scale, and finally to measure business impact.
In just a few clicks, retailers can pick what product recommendation engine strategies they want to use on each page, for each customer segment. They can pick from a wide range of ready-to-use, proven recommendations from our library. Our customer success teams work closely with retailers to identify what recommendation works best on which page, based on industry best practices and the retailer’s historic data.
They can then customize the AI algorithm based on their business. Have an in-house label? Increase the weightage on brands. Have more block colors than patterned dresses, increase the weightage on color.
Retailers have complete control over how the recommendation will appear on the website (template). They can add business rules, decide how many products should appear, labels, CTAs, etc. Retailers can have more than one recommendation strategy appear in a template.
They can even have different recommendations for different categories. For example, they may decide that if a shopper is looking at tops, visually similar recommendations should appear on the PDP. However, if they are looking at bottom wear, outfitting recommendations would appear instead. That’s the level of flexibility build into the product.
It’s now ready to be previewed!
Making data-led decisions becomes second nature with our easy-to-use design. Retailers can A/B test every journey before they scale it if they choose to. Tests are designed to measure impact on specific KPIs and will run till they reach statistical significance.
This takes the guesswork out! In our experience of having worked with 100+ retailers, A/B testing is amongst the top 3 factors that contribute to sustained growth.
Finally, Vue.ai’s Personalization engine has an easy-to-understand metrics dashboard that captures a wide range of KPIs including directly attributable revenue. This helps the entire organization look at the data and align on strategies and goals.