We live in an era where we welcome, and even expect, brands to recommend additional products we might like when shopping online. With artificial intelligence, personalized recommendations have the potential to go supersonic as AI systems use tons of data about your habits to learn your preferences.
We interviewed Klevu’s North America business development manager Miles Tinsley (LinkedIn) to better understand how Klevu works.
In a single sentence or statement, describe Klevu?
Klevu is an AI-powered ecommerce search solution, designed to drive more revenue for mid-level and enterprise-level online retailers through enhanced accuracy and merchandising capabilities.
Natural language processing (NLP) is one of the core components of Klevu, helping us to extract more meaning and context from search queries and match terms to results that other technologies are unable to match. You can see an example of how NLP can support search on the Zimmermann store (a Klevu client) below. As you can see, the product names don’t contain the word “shoe,” but the system is able to understand which products are shoes.
Klevu also uses machine learning to promote products based on how users are interacting with them, providing more real-time promotion of items based on their popularity and effectiveness. The key interactions that Klevu uses are completed purchases, add-to-carts, and clicks. These actions give us an understanding of popularity. This gives us a picture of short-term and long-term popularity and allows us to add a layer of rankings on top of just boosting specific categories or items.
I will be specific here to online retail and highlight three points:
First, AI is very deep and wide in terms of possibilities of optimizations in retail. It is evident that online retailers want to bring natural shopping experiences to shoppers, but AI tools (for example, chatbots) lack a very important piece at the moment, which is the connection to the catalog and queries.
Second, a limitation in my personal view is around productization. There are some fantastic R&D level innovations out there in AI for retail, but a proper productization is lacking.
Third, there is lot to do when it come to understanding unstructured data (examples include search queries, catalog descriptions, or product reviews) and extracting relevance from this data.
The future of AI is geared towards personalisation at scale. Every shopper has several touch points, which could range from Facebook to email. The generic marketing or sales techniques that target multiple touch points without proper context and personalisation will not be around for long.
From a software-as-a-service point of the view, there is immense scope for technologies that help decision makers find what they’re looking for. Technology alone is just one piece of the puzzle, a solid productization will be the key for long-term sustainability in AI-led marketing and sales. Access to new information and exploiting of information from ecommerce platforms will also open huge opportunities for AI in marketing and sales.
Generally, it’s the two things we’ve already mentioned: the NLP and self-learning capabilities. Our use of NLP in particular allows us to go a lot deeper in matching products to queries, which has helped us win a lot of RFPs and split tests.
Other key aspects that differentiate us include:
Klevu is suitable for ecommerce stores of all sizes, however the area where we achieve the biggest uplifts tends to be with B2B merchants and large catalog merchants, purely because this is where the NLP side of things adds the most value. For retailers with lots of more complex and long tail search queries, the NLP side of things helps to understand more and, as a result, serve more accurate products / information.
Examples of large catalog stores we work with include:
From a marketing perspective, we have a very strong reporting offering, which helps to provide more detail around queries that are driving more sales, which regions are driving the most search-led revenue, the products that are selling most, etc. Marketers are then able to take this data and boost specific products / categories accordingly to maximise sales. We also provide various other features that add value for marketers, such as the ability to serve banners for different queries (promote new ranges / products etc), the ability to quickly boost specific lines and the ability to serve different types of content (such as buying guides or blog posts).
AI can provide a better shopping experience, as long as the context is properly applied. It is also important to give time to AI and these kinds of technologies. You can’t expect miracles from the start. Finally, as with any software you adopt, it’s worth looking at its technical architecture in terms of adaptability, flexibility, and scalability.