A great sales clerk in a physical store will guide your customers around, making sure that they find the right products for them. In the same way, great AI is important for guiding your customers in their online customer journey through your store, and all the way to checkout. This is where Clerk.io comes in. With the most sophisticated AI on the market, Clerk.io provides webshops with relevant and personalized search results, recommendations and targeted marketing solutions for their customers from day one of implementation. ClerkCore, the technology behind Clerk.io, surpasses earlier recommendation methods such as collaborative filtering and neural networks by leaps and bounds – especially on these 3 key points: First, most AIs are helpless when they don’t have any historic data for products and will usually rely on fallbacks that are not relevant. ClerkCore can predict future sales even if they have not happened yet, because the AI understands the product’s placement in the current shopping trends just like a human would and can immediately make the right predictions. This is important both when new products are added, but also for displaying great results for the longtail of products that have not been sold yet. Second, ClerkCore understands the direction of recommendations – after all, a bike lock is a great accessory for a bike, but a bike is not usually a good match as an accessory when looking at a bike lock. Third, most AI’s can only project past actions from customers into the future, but are not able to predict trends that have not happened yet. ClerkCore understands shopping patterns, which enables the API to predict future sales that have yet to happen. So how does all of this work in real life? Let’s walk through a test case with Unisport from November where we take a look
at how these technologies stack up when showing recommendations for a pair of football boots with a history of 36 orders. First, counting is where you suggest items based solely on what has been sold most with a certain product. The counting-based suggestions given included a bootbag that was sold 35 times with the boots, a neck scarf that was sold 4 times, as well as adult keeper gloves, child keeper gloves, and child football boots – all sold once with the product. But were all of the suggestions relevant to future customers? The boot bag can be used by everyone who purchased the football boots, and the neck scarf was a timely, seasonal recommendation due to the time of the year. Yet, the keeper gloves are position specific, and the football items for kids might just be one off purchases. Next is collaborative filtering, where customer profiles are created based on previous customer behavior. In this case, the cross-sell products suggested were only bought once along with the boots in its 36 order history. In this case, only the shin guard that matched the football boots was a relevant recommendation and the only real potential cross-sell product shown, as the position specific keeper gloves and other shoe orders were also one off purchases, that would only be relevant to a very small subset of customers. And the great cross-sell product of the boot bag was not even shown. With ClerkCore, the cross-sell products suggested included the best-selling boot bag and the matching shin guards, but also a relevant shin guard alternative, which was the top selling Nike shin guard and a good match for any Nike football boot. The remaining suggestions of the socks and gloves matched the product in color or came from the same product line, and were great seasonal cross-sell purchases. Most AIs today cannot predict future sales of a new product or customer, they cannot determine the directionality of what makes a great cross-sell product and what doesn’t, and they cannot predict new connections and correlations when they haven’t happened yet. But Clerk.io can. And all without needing to supply massive amounts of data. Grow your sales. Not your workload.