Happier Customers, Higher Profits
Somewhere in the goals of any for-profit company is a healthy stream of profits. Profits are what keep the corporate world spinning. As younger people become more and more careful about what they spend their money on, the need for consistent profits is increasing exponentially.
Luckily for many companies, the rise of AI and Machine Learning might be exactly what you need to keep traffic coming to your websites. Young people love ordering things online--even more so if they feel it’s within their price range and easily accessible. Give them all of these, and you just might have some new loyal customers.
The Importance of Search
We all use search engines. Any time we have a specific question or an open-ended query, sites like Google and Bing are our best friends. Most people never really stop to think about how these work, but for companies, understanding that these are driven by AI and Machine Learning--and more importantly, how AI and Machine Learning drive them--is key to creating an effective website that fits the needs of its customers and increases its profits.
While search is undoubtedly the most well-known and widely used innovation that has sprouted from the development of AI and Machine Learning, it still continues to pose tough, unsolved questions. Why do some search results yield the exact information we were looking for, while others are way off the mark. In most cases, it all comes down to the algorithm. But we’ll get back to that later.
How the Search Works
At the end of the day, customers are using search for practical reasons--therefore, its functionality needs to be practical as well. By nature, search has the capability of accomplishing the tasks that are asked of it because it exists everywhere and is able to unlock information from anywhere that information exists. That being said, there are still certain steps that a company can take to make their searches a better experience for their customers.
Mapping User Intent to Products
In the simplest terms, user intent is what a customer puts into the search bar with the hopes of yielding a specific set of results. In order to map these, Prasad Joshi, who has spoken about his experience with the AI program UNBXD, says that there are two main steps: mapping the searches and ranking the results.
Step One: Mapping Searches
Also known as Query Enrichment, the first step of mapping AI is all about keywords. Keywords are entered into a search and are then combined and filtered to find the most relevant products in the search. So, if a customer is searching the online Target catalogue for lined winter boots, “boots” is filtered within the AI as a keyword, but “winter” changes the meaning to narrow down the search.
The addition of “lined” adds another layer to the prospective results. Not only is the product categorized under “boots” and, specifically, “winter boots,” but it the “lined” filter also means that the search results should yield only those boots that have some kind of fleece or fur lining for insulation.
Another thing that needs to be considered when implementing AI--especially when you have a product or service you’re trying to sell--including an algorithm for what’s actually popular within searches at that moment. This means that all of the relevant keywords of the search must also be included within the results of popular products--including seasonal ones--to ensure that they also appear in the search.
Step Two: Ranking the Results
In every web search, there is a specific problem that’s trying to be solved. Whether that’s simply a reminder of the definition of a word you don’t often use (i.e. “ubiquitous,” which I had to search as a part of my research for this article) or the question “what is the meaning of life,” the asker usually has a problem, and they’re trying to solve it by means of their search.
If you’re searching for a definition, your problem might be that you want to use a word, but you want to make sure that it’s relevant to whatever you’re writing. If you’re searching “what is the meaning of life,” you might be having an existential crisis. It happens.
The primary task of AI and machine learning, then, is to rank the results to your question in order of relevance. This means taking the results that you generated in step one and ordering them in terms of how likely they are to meet the customer’s needs.
That is to say, if a customer searches for “lined winter boots” again, the first step will call up the products matching those keywords, but the second step will be for the AI to rank those products according to how well they match the customer’s request. Usually this involves taking stock of what other customers who have searched for the same items have wound up purchasing and showing those first.
Most searches actually have a fairly low repeat frequency--meaning the task for the AI to sort through different products doesn’t occur all that often. Unfortunately, however, with the boom in the creation of new products that are added to the ecommerce catalogue every day, it’s beginning to become more and more difficult for many online searches to continue functioning in this way.
Machine Learning and AI require models to learn about data, but when the models aren’t updated in a timely manner, we can’t learn about the data as quickly as we need to if we want to show newer products in the results.
Understanding User Intent
The best way for companies to effectively cater to what their customers are looking for is to understand user intent. We need to put on our pseudo-psychologist hats for a minute to understand who the user is and what exactly it is that they’ve asked for.
A lot of times, users will use a variety of names to search for the same product. With the “lined winter boots” example, customers may type exactly that into the search bar, but they may also type something like “snow boots” or “pac boots” with the same exact intention.
When we can map each name for the same products with the specific keywords that are associated with them, we are better equipped to give the customers what they want. Mapping products to a specific model helps Machine Learning and AI with recall and precision for the future searches of others looking for the same products.
Making Search Work for You
It’s one thing to understand how AI and Machine Learning work at the base level, but in order for it to be the most effective it can be, you need to know how to make it work for you. Joe Layton, a retail executive and former Vice President of apparel retail at The Children’s Place is a bit of an expert in this area.
AI is the Tool, You are the Builder
With the help of AI, you can dig into the different attributes that make up your products to find out how to better cater them to your customers, but keep in mind that AI is there to help. While it is statistically more efficient and more accurate than any human would be, this doesn’t mean that it will never make mistakes. You will always need people on standby to go in to make corrections to the AI as needed. AI is only as good as the people who create it.
Make it Flexible
If you want your search function to be as successful as possible, it needs to be able to adapt--with and without your help. Each item on your website should have a short description that includes keywords the AI and Machine Learning can use for filtering. Once your system is able to recognize the keywords for filtering, a rule is created, which allows the processes to be carried out over and over again for repeat queries.
At any point in time, the rules of your AI can be adjusted and bent to procure better results. And this is, ultimately, the most important thing to remember when you’re implementing a search function to your company’s website. Better results mean happier customers, and happier customers mean a greater profit for your company.
This is a handy thing to keep in mind if you have to make an argument for AI with the leadership of your company. With the help of this article, you should be able to explain how AI and Machine Learning can benefit your company.
Nurture Your AI
Chances are, your system isn’t going to work exactly the way you want it to from the get-go. There will likely be some kinks to work out, but as Layton points out, nurturing your results is the best way to help the AI to become more and more accurate.
Like most things, nurturing your results is going to take time. A sickly baby bird with a broken wing isn’t going to heal overnight. But if you nurture it--provide it with the right medicine and give it time for the wing to heal--it will get better.
The most relevant example of when you’ll need to nurture your results is when seasonal products start to come out. Most seasonal products appear in cycles. Beach towels, for instance, don’t typically start appearing in most temperate weather stores until mid-to-late June.
Say you kickstart your system in mid-November. When you add beach towels to your model the following June, the algorithm may not be as accurate in picking them up in searches. As more years pass, however, and the system becomes more familiar with the new products, it will become more successful when people search for them.
Continue to play with the rule set of your system for the most effective nurturing. Whether or not you’ve added season products the system doesn’t yet recognize, it wouldn’t be strange for you to see some funky results the first time you test it out. Some results may not seem relevant to the relevant searches, but it probably just means that there’s a bug in the rule set. Sift through it to find out.
If you’re still finding strange results after you’ve played with the rule set, take a look at how the data itself is getting into various systems. A lot of the time, AI can get tripped up if the system layout creates issues for how the data can get into different spheres. Think about which inputs will be necessary to allow the algorithms to feed and learn properly, and you should be able to adjust the system to input the data correctly.
Your first priority is probably always going to be maximizing your profits. High profits give your company what it needs to keep operating. Advanced AI and Machine learning can help you get there, but this isn’t even the extent of their benefits for your company.
Because AI can primarily function on its own, this frees up a lot of time for the people who keep it working. It increases morale when your workers are spending less time stressing over how to make sure customers’ needs are being met by their own hands and more time to focus their attention to other projects. It sparks creativity--enough to help your company grow in ways you wouldn’t have expected. It might be exactly what your company needs to grow.