AI-powered, real-time personalization for each of your online shoppers: this post gives you detailed insights into the new FACT-Finder Next Generation Blixt.
“Once in a while lightning strikes,” announced Emile Bloemen, CEO, six months ago upon news of our acquisition of Loop54, the Nordic market leader in 1:1 personalization. Ever since that moment, our product teams have worked diligently to incorporate Loop54’s revolutionary AI technology into FACT-Finder Next Generation. Today, we are thrilled to introduce you to the union of our two market-leading technologies, bringing together the best of search, personalization and recommendations. Following Emile’s quote, we christened this technology with the Swedish word for lightning: Blixt. In this post, you’ll find out what it’s all about.
What will not change in the next 10 years?
Hardly any industry is as fast-paced as eCommerce, with businesses working endlessly to stay ahead of the competition. Many people wonder what will change over the next ten years, but very rarely do we stop to consider what will stay constant. And yet these constants are the ones on which you can build a sustainable, impactful business strategy that will give you a competitive edge.
Whatever the next 10 years will look like, one thing is certain: personalization is a megatrend of the 21st century, and the possibilities are near limitless. While brick-and-mortar stores invest millions in “the one” perfect store layout, eCommerce can achieve the perfect shopping experience for every customer by providing the right product at the right time.
But not all personalization is the same. Many eCommerce stores still try to provide tailored content based on customer groups, but no segmentation can reflect the individual context of your customers. “Personalization by segmentation is dead,” explains Robin Mellstrand, Chief Strategy Officer. “Today, we need to be able to deliver an ultra-relevant experience to our customers in real-time.” And that’s exactly what FACT-Finder Next Generation Blixt makes possible: real-time, AI-based 1:1 personalization.
“The future is ultra-relevant and each customer deserves their own customer experience. That’s why we continue to innovate.“Robin Mellstrand, CSO at FACT-Finder + Loop54
The Sweet Spot Between Human and Artificial Intelligence
Next Generation Blixt marks a new milestone in search and personalization for online stores. This product combines the patented search algorithms of FACT-Finder with GOLEM, Loop54’s revolutionary artificial intelligence algorithm. FACT-Finder’s market-leading search, navigation and merchandising functionality is thus enhanced with Loop54’s real-time, 1:1 personalization and recommendations.
Where FACT-Finder empowers the business user with an intuitive UI, Loop54 provides a neural network that automatically maps relationships between products and optimizes in the background. Together, we remove the age-old “black box problem” and allow you to have as much – or as little – control as you’d like. We call this the sweet spot between artificial and human intelligence.
Below, we describe the innovations of FACT-Finder Next Generation Blixt in detail. Among them:
- Wisdom of the Crowd
- 1:1 personalization in real time
- Recommendations, powered by AI
Wisdom of the Crowd: Human interpretation of search queries
What does a great online shopping experience look like? At FACT-Finder and Loop54, we have over 30 years of combined experience in eCommerce. Several hundred integration projects have clearly shown us that the shopping experience stands and falls with the quality of search results and product recommendations. The more relevant the products displayed, the higher the conversion rate, average shopping cart value and customer loyalty.
Until now, it has been a real challenge to show the most relevant products to each customer at any given time. A big reason for this is that the context of human search queries is not easy for machines to interpret.
For example, someone searching for a “PS4 game” probably wants to buy a game for the PlayStation 4. But someone who searches only for “PS4” most likely doesn’t want to buy a game, but the actual console (and maybe later a game). Humans intuitively understand this distinction, but machines don’t because they find “PS4” in the product data in both cases. And that’s exactly what leads to a frustrating shopping experience. As a customer, you don’t feel understood and might get the impression that the store doesn’t offer the product you’re looking for. The logical consequence? You leave the store and try your luck somewhere else.
This isn’t a new problem – it is widely known and can occur in any industry. For example, in a grocery store that first displays all orange colored or flavored products for the search term “orange”. How have such misinterpreted search results been improved traditionally?
- through manual (and time-consuming) optimizations such as synonyms, antonyms and ranking rules.
- through automated (but slow-learning) result optimization with machine learning.
The advantages and disadvantages of both options are obvious: manual optimizations are effective and targeted but require time and effort. Moreover, it’s impossible to cover all search queries, as the amount of long-tail combinations in online stores is infinite: an analysis by Loop54 shows that 50% of monthly search queries are completely unique.
Only with automatic result optimization can long-tail search queries also be covered and resources saved. The disadvantage? Until now, it took a lot of time and learning for the improvements to take effect. But with FACT-Finder Next Generation Blixt, this is changing.
Let’s use the same example as before, “PS4.” Thanks to the integrated GOLEM engine, Next Generation Blixt already understands what is meant within the first customer sessions. Where previously countless interactions were necessary (“big data”), today only a minimal amount of data is sufficient to understand the context of each search query. What’s more, Blixt can adapt immediately to changes in search and click behavior, allowing it to respond flexibly to changing demands. Seasonal products are thus reliably and automatically ranked higher or lower in the search results.
What makes the AI algorithm GOLEM so unique?
GOLEM is perfectly tailored to providing relevance and personalization in eCommerce, as the AI combines the strengths of artificial and human intelligence. Based on the individual product catalog, the GOLEM algorithm develops a neural network that maps the relationships between different products. Instead of relying on data about individual products, GOLEM generates product-independent neurons that gather knowledge about a specific product type and its possible attributes. Since information is shared between similar products in this way, very little learning data is required. This creates the following advantages:
- Personalized, top-relevant ranking and sorting:
GOLEM understands the search intent within the current session and delivers a tailored sorting of results. Classic machine learning methods require much more time and input to achieve similar results.
- More flexibility, products can be interchanged at any time:
Because GOLEM develops its understanding independently of individual products, the quality of results remains consistently high even when the product assortment changes. GOLEM understands even complex relationships between products and immediately knows where new products belong – even if no interaction with them has taken place yet.
- Understanding expands exponentially:
Information about personal preferences and search intent spreads within the neural network. Each interaction thus impacts thousands of products, rather than just one.
“With our technology, every eCommerce store can provide ultra-relevant and personal customer experiences. We believe that this is the key component our industry needs to move forward.”Robin Mellstrand, CSO at FACT-Finder + Loop54
AI according to your rules
The FACT-Finder backend is where human intelligence comes into play, allowing you to track exactly how the AI affects individual search queries. For example, The Cockpit shows you where products were ranked originally in the search results or you can directly compare them in the side-by-side comparison, one with and one without AI optimization.
For fine-tuning search results, you can use the tried-and-tested FACT-Finder tools, such as ranking rules and campaigns. Define rules to bury all non-deliverable products, always place sponsored products at the top of the search results, and more. Next Generation Blixt gives you all the options to combine AI with your sales goals, and to meet (or exceed) your business requirements.
You can also be sure that the AI takes your requirements into account without turning a blind eye to relevance. Next Generation Blixt overrides ranking rules if the general learnings in the store or the individual preferences of a customer speak against a rule. Let’s say you had a ranking rule that ranked all Lego products slightly lower. If someone who is clearly interested in Lego searches for “Star Wars,” our AI will override this ranking rule and correctly show Lego products first.
- Less manual effort, more relevant search results
- React quickly: fast adaptation of the sorting to seasonal demand changes
- Transparency: see how the AI is influencing results with our intuitive UI
- Perfect balance between automatic and manual optimization
1:1 personalization in real time
Of course, the context of a purchase journey is not only influenced by general search behavior (aka Wisdom of the Crowd). Individual interests and preferences also determine the intention behind a search query. The same search query can be meant in completely different ways based on the intent. With Next Generation Blixt, search results always match the individual context of your customers.
The best way to explain how this works is with concrete examples. Let’s imagine a toy store where customers are looking for products related to “Star Wars.”
Personalization within a session
Let’s start with the Smiths, who are actually not a single person but rather a family visiting the store for the very first time today to redeem a voucher.
Looking for the Lego set from a TV commercial, the Smiths first browse the homepage and click on a Lego product. Through this event alone, Next Generation Blixt learns that this user is currently interested in a product from the Lego category – at least with a high probability. Just like in a brick-and-mortar store, where they would now be in the Lego section.
The Smiths now enter in the query “Star Wars” into the search bar. Due to the previous interaction with the product on the homepage, the search results now predominantly show Lego products – and hardly any PS4 games or model kits. The AI thus recognizes the individual interests and wishes of the customer within one and the same session.
Furthermore, if the Smiths were to then go back to the homepage or a category page and click on another product type, the personalization would adapt to this product type. In a brick-and-mortar store, this would mean that they would have gone to a different aisle or section, making the online experience nearly as smooth and intuitive as being in a physical store.
Personalization with User ID
David has no interest in Lego; he is a fan of building miniature models. He regularly orders from his favorite hobby store, where he also has his personal user account. After logging in, he now enters the same search query as the Smith family – but with a completely different intention.
Through the user ID of the login, Next Generation Blixt can use the previous purchase patterns to identify David’s current intent. The purchasing behavior clearly reflects that David prefers to order model kits, so FACT-Finder interprets the search query “Star Wars” not with Lego results, but with model building products from relevant brands.
Personalization according to individual preferences
Anna, a passionate gamer, wants to buy the latest console game in the series.
A few days ago, Anna sporadically browsed the store’s video game category, bought a controller, and checked the current console offerings. Next Generation Blixt can associate these historical transactions via cookie with Anna’s current session. In the results for the search query “Star Wars,” FACT-Finder now predominantly shows video games, as these best match the customer’s previous behaviors.
- Higher relevance leads to higher conversion
- Personalization takes effect immediately, within one session, thanks to a unique machine learning process
- Significantly faster personalization than comparable solutions
Recommendations that exceed customer expectations
When it comes to product recommendations, online shoppers expect the same quality and relevance they are used to from Netflix, Amazon, Spotify and Co. To meet this high standard, many retailers invest a lot of time and money into manually configuring their recommendations. What may be somewhat manageable for assortments with 1K products, however, becomes an impossible task with 100K products or more.
As such, retailers with large assortments have relied primarily on data about products that are frequently bought together to have recommendations generated automatically. Unfortunately for them, in most cases the data is not sufficient to provide a truly relevant experience. In addition, recommendations are limited to popular products – ignoring new products and niche products with higher relevance.
With Next Generation Blixt, you can take your recommendations to a whole new level – without manual effort and even if you only have minimal data to work with. Just as with the personalization of search results, GOLEM understands how products relate to each other. Whether on the product detail page, in the shopping cart or elsewhere in your store, the AI shows your customers exactly the recommendations that best fit the context and the purchase intention of your customers. Three different recommendation types are available:
Personally tailored product recommendations:
Based on your customers’ individual purchase patterns, FACT-Finder generates personalized recommendations that match the product they are currently viewing, the context, and purchase intent. This is a reliable way to increase your customers’ shopping carts and thus your average order value.
Similar products as purchase alternatives:
With its extensive understanding of the product range, our AI algorithm can also recommend alternative products. This type of recommendation is ideal for inspiring customers and keeping them in your store – even if they are still undecided or haven’t found their desired item right away.
Manual recommendations as a complement:
FACT-Finder Next Generation Blixt can provide 100% of your product recommendations and automatically increase order value. In some cases, however, you may want to have a say in what is recommended – for example, in the case of very frequently clicked products. Manual optimizations can still easily be implemented with just a few clicks, giving your team full convenience and flexibility. Your manually created recommendations are always prioritized by FACT-Finder and all remaining recommendation slots are then filled with the most relevant recommendations from the AI.
- Personalized and highly relevant product recommendations with minimal data
- AI with unique understanding of product relationships in the assortment
- Perfect balance of automation and control
Blixt: Your competitive advantage for the next 10 years
Nowadays artificial intelligence can interpret the context of users with enormous reliability. More and more retailers want to offer this almost human understanding of search queries in their online stores as well. FACT-Finder Next Generation Blixt makes this possible and creates a shopping experience with which retailers can compete away from the price war:
- Top-relevant hits. Thanks to “Wisdom of the Crowd”, your store interprets search queries like a human being and thus increases the relevance of results (and consequently your conversion).
- 1:1 personalization and recommendations. Search results and recommendations adapt in real time to the individual interests and context of each customer. Higher conversion, preprogrammed.
- Boost & Bury. Ranking rules, pushed products, synonyms, and much more: In the FACT-Finder UI, you can also influence search results yourself – and let AI follow your direction.
We hope you were able to gain some exciting impressions of FACT-Finder Next Generation Blixt. It makes us incredibly proud to see the proprietary algorithms of FACT-Finder and Loop54 combined into one powerful eCommerce solution. The first of our customers are already live with Blixt and the initial results are very promising. We’ll be posting case studies over the next few weeks – be sure to follow us on social media and join our newsletter to stay in the loop.