2022 will be a year of many changes for the product discovery space.
As more and more consumers have moved to online shopping for the long haul, many retailers are now looking to strengthen their search, browse, and recommendations capabilities to meet the demand.
But what will those optimizations entail?
What are the biggest retailers in the world working on to ensure their product discovery experiences drive the right KPIs, make users happy, and improve the lives of their team members and merchandisers?
This article will cover exactly that. We’ll start with one of the more unexpected optimizations:
Stop Optimizing Site Search for Relevance: Why Relevance Isn’t Enough (And What Should Take Its Place)
Traditionally, when we think about relevant search results, we’re talking about search results that in some way relate to the search term used to find them.
The challenge with optimizing your search for relevance is that it’s subjective depending on each person, not an objective goal the company can easily identify and optimize for. The products that appear for a query optimized only for “relevance” will almost certainly not be the products both you and your users will benefit from seeing.
Here’s an example of this concept:
Style Gallery is a leading apparel company with a decent size product catalog. 100 customers come onto their site and search for “t-shirt”. The company has over 500 different types of t-shirts that it sells. If you were working at Style Gallery, how would you determine which t-shirts show up at the very top of search? How do you determine if the t-shirts you choose to appear at the top are “relevant” to each customer that searches for a t-shirt?
A more extreme example is a search for “chips” on a grocery website:
If a user searches for chips on a grocery website, should the grocer show Cheetos? It’s not strictly relevant to the search because Cheetos are not chips, but what if users who search for chips are highly likely to buy Cheetos? Should the retailer pedantically disagree and only show potato chips, or should it show the user Cheetos, even if they aren’t strictly relevant?
So, that leads us to the big question:
What do you optimize site search for if it isn’t relevancy?
“The answer is to focus on the business metrics the company cares about: revenue, purchases, profit, or whatever else most needs to be increased. Then both evaluate the search, and drive up the probability that each customer has a successful buying journey using the clickstream data that happens after the search. In short, show each customer products for each search in the order that will most likely lead to an increase in the business metric.
To achieve this, retailers should use three key attributes balanced together: relevance, attractiveness, and personalized attractiveness. Relevance should prevent something like shoes showing up for that search for laptop. Attractiveness should ensure that the most attractive laptops (the ones most likely to be purchased from that search) should show up at the top, and personalized attractiveness should use what we know about each user, their individual clickstream data, to show the particular results most attractive to them.”
Leading search companies like Google and Amazon optimize their search results to be attractive to each user and tie it to a particular business KPI to make it objective. This KPI can be any number of things (ex. conversions, margins, click to conversions, add-to-carts, etc.). If we take our t-shirt example from before and optimize it for conversions, the answer to the question becomes simple: we want to show each user t-shirts that they are most likely to convert on.
By implementing algorithms that optimize for a particular KPI, you are able to generate search results that benefit both you and your users.
Want more info on how to determine your search KPIs? Reach our full article on the topic here.
But AI-driven search systems can do more than just optimize your KPIs. They can also make your (and your merchandisers’) lives much easier.
Merchandising in 2021: Manage by Exception
A majority of the merchandising work that goes on today is educated guesswork.
It’s rare that merchandisers will know exactly what products to place in specific positions for any given query to guarantee an increase in business KPIs. As a result, they’ll turn to things like “relevance” or “popularity” to guide their work (which isn’t nearly the best indicator of a product’s success, like we previously mentioned).
But intelligent search systems today do know how to rank products for any query to increase business KPIs.
We define clickstream as data on the actions & pathways your users take on your website — whether it’s searching, browsing a category page, clicking, or adding to cart/purchasing.
Rather than guessing what products each and every user wants to see and purchase for any given query, clickstream data shines light into what your users are telling you they want to see.
Like we mentioned before: the products that should be ranked the highest are the products that users interact with and purchase most (i.e. the products most likely to lead to an increase in your business KPIs).
With the addition of clickstream-backed machine learning, merchandisers no longer have to worry about tasks like manually re-ranking products and creating synonyms and redirects. The AI does that for them.
“Won’t this approach make the merchandiser’s job irrelevant?”
We get this question often at Constructor — and the answer is always a solid no.
Machine learning doesn’t eliminate the job of a merchandiser. In fact, it enables them to be smarter and more effective.
Merchandisers can now focus on being creative and finding ways to drive real business value without relying on educated guesswork.
Maybe you’ve just secured a partnership with a brand and would like to create an awesome collections page to boost their products. Maybe you’d like to curate themed collections around a sale or a holiday and promote them on your home page. Or maybe you’d like to manually merchandise queries with a high lost-sessions rate (see below for example):
When your merchandisers can trust your systems and your data, all of this becomes possible.
And in case you aren’t sold on the idea of AI-driven search, here are some other use cases:
At its simplest, autocomplete helps users search for products faster by estimating their queries and recommending completed queries. But at Constructor, we believe autocomplete can be used for much more than basic query estimation.
Great autocomplete corrects phonetic misspellings, keyboard proximity typos, punctuation nuances, and omitted character typos. Is it “blue ray”, “blue-ray” or “bluray?” It’s officially spelled “blu-ray”, but you shouldn’t require users to know this to get good results.
Great autocomplete can also detect synonyms within the query to serve autocomplete results. For instance, if a user types “pop” instead of “soda,” the autocomplete system should detect this.
As a hypothetical, if a user searches for “sleek red dresses for women,” the system should know that “sleek,” “red,” “dresses,” and “women” are the most important keywords in the query, and should serve recommendations based on those words.
We briefly discussed the idea of the benefits of ranking items based on business KPIs rather than relevance, but there’s still one point that needs to be addressed:
Users have preferences.
While a majority of users may convert on “Lays” chips for the “chips” query, that doesn’t mean everyone will. And that’s where personalized product re-ranking comes in.
Great product discovery systems take the idea of “relevance v.s. attractiveness” and apply it on a user-by-user basis. Once you have enough data about any specific user’s preferences, instead of continually re-ranking products on a broad basis, you can begin ranking products specifically for that user that will give you the best chance of increasing your business KPIs (like conversions).