Speaker: John Murcott, VP Products
Please introduce yourself and what you're hoping viewers to get from this walkthrough.
John Murcott: Hi, I'm John Murcott and I run Products and Strategy for HawkSearch. I'd like to spend some time today to go through a feature called Recommendations, but not only explain the feature but also explain a little bit how it works underneath the covers.
How does HawkSearch improve Recommendations?
John Murcott: So the first thing we should do is just make sure we're on the same page with what is a recommendation? It's essentially a list of products that HawkSearch is recommending to a visitor. It's presented as a carousel. Now, you might ask, why is HawkSearch providing recommendations? Because HawkSearch does search. But that's the critical aspect of our value proposition in showing recommendations. It's the same way that you would ask a friend for a recommendation. If you ask a friend for a recommendation, that friend is gonna wanna know two things. First of all, what is it that you're interested in buying? Let's say it's a car or an article of clothing. The second thing though. And the reason why it's so important to ask a friend is that your friend knows you, they know what you're interested in. They know what kind of products you've bought before. And that's why their recommendations are so important. And it's also the reason why HawkSearch does such a great job with recommendations because we know what you've searched for. We know what you've clicked on, we know what you've added to your shopping cart. What you purchased, etc. And we capture all of that just to help with the recommendation itself based on that activity, we can also base it on activity that other people do and what other people have put in their shopping cart and what other people have purchased and clicked on and search results. So all of this information is analyzed through machine learning and then is used dynamically when recommendations are shown on the site.
What events can you track with HawkSearch?
John Murcott: Any time we click on a link on a website, an event occurs and that event would do something. For example, take me to the next page on the website. HawkSearch has Event Tracking API that listens to specific events that drive those recommendations that we were talking about before the most common one would be what you're searching for. What is the keyword that people are searching for? The second thing we'd want to track is in the results that are returned. What are the actual items that are clicked on? Because that would make a nice relationship between what people are searching for and what they actually click on. We would track, was the item added to a shopping cart? if it's an eCommerce scenario and or was it purchased? All of these events are tracked, not only as individual events, but also as sessions. So we can see what the relationship is between what people are searching for, what they're clicking on, what they're adding to their shopping cart and what they're purchasing.
How can we use this data to improve recommendations?
John Murcott: I mentioned earlier that the best way to get a recommendation is often from a friend because the friend knows you, they know your style, they know your interests, they know what you've bought before. But they also want to know what are you interested in? Are you getting a new car? Are you getting some new clothing? What have you And that's how HawkSearch works. The first part. What are you recommending at a category level is driven by our end users. The HawkSearch merchandiser that says we want to build a recommendation about a certain electrical component or a car or a piece of clothing. The second part though that context that makes the recommendation from a friend so valuable is all that additional data that we were talking about before? So yes, the recommendation might be about a certain category of items but the adjustment, the recommendation, additional magic, so to speak comes from the data. We were saying before. Have you purchase similar products like this before? What are other people searching for? And when they do search on that type of category, what do they click on? What do they purchase? So the data enhances the initial outline if you will that the merchandiser selects the merchandiser says, well, I want a recommendation on this category or this group of products. The AI aspects, the Machine Learning aspects of HawkSearch leverage all the data we were talking about in order to improve those recommendations.
Which web properties are positively impacted by recommendations?
John Murcott: I can't think of too many types of websites that could not benefit from a recommendation. But let's talk about two of the most common scenarios. The first would be a content driven site, a site with a lot of blogs or just regular web content. And at the end of each one of those articles or blogs, what have you, press releases - there could be an option that says here are other articles you would like that recommendation that we were talking about before, what other people have searched on what other people have clicked. And that's a very common use case. The second one and the one that HawkSearch sees more often is on eCommerce. So when I'm on a product detail page or learning about the product, and hopefully considering a purchase, you can have a recommendation for other products to consider. Also as you continue that journey. So let's say I do add the product to my shopping cart. That would be another perfect use case. These are other products that people have purchased when they added that first product inside their shopping cart. And then another option of course is on the actual purchase page. Would you like to buy more? Here are recommendations, you still have their attention. It's a great time to make a recommendation. But one of the things that's also critical to understand how these recommendations work. And I mentioned this in the beginning a little bit underneath the covers. In some cases, you may want to put that recommendation experience, that carousel view of list of products on other parts of your site on the home page, maybe even be on a blog page even though the site itself is for eCommerce because maybe the blog is about a certain category of product or in general, you would like to show which products are trending on the site. So you have a lot of flexibility regardless of what type of site you have.
