Superscalar Advisory: How to think about recommendations

I believe that technology shouldn’t be restricted only to the largest players, and everyone should be able to take the fruits of sophisticated methods worth modern digital marketing. e-commerce comes to a website, and the website does not have a blood’n’bone advisor. Thus, obviously, every manager needs to fulfill that gap somehow.

Let’s start from the beginning:

A recommendation is basically a filter that you put on your product list. Simple as that. The role of this filter is to pick ones that fit certain criteria. We can apply a universal answer to it: It depends:

• Requirement: Products best fitting the client’s needs. It guarantees purchase.
• Probability: Most likely, the client is going to buy these. Nobody said that the product the client needs is the one that he/she desires, and desire is a powerful sales driver.
• Revenue: Products that will grant you the highest income, as some of them are more expensive. At the same time, fulfil needs that clients have but are not aware of.
• Discovery: Sometimes, customers discover the existence of products that fulfil their needs but aren’t aware of.
• Extension: Products that extend the usage of other products owned by customers.

You’ve probably already associated a few of them with widely known techniques like, for example, cross-selling and up-selling. Still, generally speaking, everything is designed to help clients in any way possible. As always, AI comes in handy here. I like to call it the Ghost in the Marketing Machine (or Ghost in the Shell if you prefer, it’s also a cool name, though.)

E-commerce is an excellent field for Artificial Intelligence development, as it operates on data, numerous inputs, large scale, and efficient verification methods. While planning marketing moves, usually you encourage users to perform only one action at a time, like a clear call to action. Things that are easy to ‘understand’ and easy to analyze at the same time. Every step users make before the finale can be recorded. Thanks to that, it is easy to analytically determine the connection between the cause and the effect – the essentials of AI development.

Where to recommend?

Usually, sales recommendations are placed in web layers or pop-ups. Speaking of sales recommendations, I mean all sales/marketing activities, including static ones, like banners, coupons, and pieces of information placed on the site. We can divide them into five types:

• Non-Personalized, Non-Real-Time – Promotion banner, new collection info.
• Non-Personalized, Real-Time – Best Sellers, often bought, for example, in this category of products.

Relying on AI:

• Non-Personalized, Real-Time – Recommended for this item (item-based).
• Personalized, Non-Real Time – Recommendation on purchase/cart history (user-based).
• Personalized, Real-time – Purchase based on all activities, including product, viewed.

As you can see, there are a lot of tactics.

It’s good to rely on static non-personalized recommendations. However, the Spirit has to dwell in your eCommerce, not to be it, so you shouldn’t ever limit yourself only to them, for AI power reaches beyond customers’ understanding 😉

…Where users seek

Besides static and dynamic frames, eCommerce wields another tool, which we, as users, get used to so much that we can barely think about it as a tool of recommendation – the old-fashioned search bar.

A search bar is a behavioural Cornucopia as clients leave tons of explicit and valuable pieces of information. They say clearly what they need, and all you have to do is make products that match their expectations pop up before their eyes. But remember that it’s your call what it will be.

Besides that, search engines have another secret power.

Natural Language Search

So far, clients have used search engines in a simplified way, feeding them with various keywords to find desired products. But what if they could just type the query using Natural Language? “I’m searching for a dress good for a wedding as a guest” is a great example.

As I mentioned above, explicit data is also a vital type of information. On the one hand, users sometimes lie or are not entirely honest. On the other hand, if the client finds the information he/she provides vital (from his perspective) – it is in their best interest to provide them. Search queries are that type of information.

Why is that? Why does natural language make the difference? It makes us feel like we are talking to someone alive, then customers are willing to say more, even unconsciously.

• An assistant in a stationery store is going to receive more preferences data.
• In eCommerce, the role of the assistant is performed mainly by search engines. Customers there willingly give away information, unfortunately usually with the help of keywords.

We want search engines to be as good as possible to use even more “legal information” that we can use to personalize these search results. Customers want to ask NL questions; they just need to be allowed to do that. The customer needs to know that they can describe their needs, and the search engine will help them.

Then, if you know what the user is searching for? Feel free to use this knowledge when you suggest items in search results! How exactly? It’s your call!

What is interesting, on the very beginning of search engines… they worked exactly like this. It’s just customers that have already adapted to a particular “search convention,” in which they use keywords without particularly bothering to formulate them into a meaningful sentence: it’s simpler to develop, but that does not mean it works better.

…Where users ask for

Last but not least, chat/voice chat/call. It’s basically the extension of the previously mentioned Natural Language Search. The smart, virtual sales assistant is the ultimate advisory and recommendation tool for eCommerce, personifying the idea of the Ghost in the Marketing Machine; thus, it’s literally an Artificial Being that helps the customers to shop. To develop such an assistant is a bit more challenging; however, we have an idea of how to do that and are working hard on it!

Virtual Sales Assistant

The first step is the development of a smart search engine (mentioned earlier). But why are Natural Language Search and Smart Conversational assistant such a huge advantage for the competitive eCommerce market?

Because every type of customer knows how to use language flexibly and efficiently, even in case of a misunderstanding, they can answer additional questions asked by the customer advisor. The language is the most basic interface we developed over millennia of evolution. Language itself is the ultimate source of information.

Information we badly need to recommend items properly.