Beyond the Cart
Beyond the Cart
On-site Search: Your Tool to 10X User Experience and Product Discovery
Loading
/

On-site search has come a long way—from simple keyword matching to powering personalized, AI-driven product discovery. In this conversation, Dane Dickerson (Human Element) and Matt Eisnor from Algolia dive deep into the evolving role of on-site search in eCommerce.

Watch this episode on YouTube.

Full Episode Transcript

Dane Dickerson: Hello, hello, and thanks for listening to the Human Element podcast, Beyond the Cart, eCommerce insights from Human Element. This conversation is with me and Matt Eisnor over at Algolia. Algolia is one of our partners for onsite search and Matt brings a ton of experience about how onsite search has changed, what tools are available. We talk about AI, we talk about optimizing for different customer profiles, and really get into the potential of onsite search tools like Algolia to provide better customer experiences and build brand relationships. I really enjoyed this conversation and I hope you do too. And as always, thanks for listening.

Okay, well, Matt, thanks so much for being here to talk through a lot about onsite search. You’re able to come at this with technical and a lot of strategic experience informed by work at Algolia. I’m coming in here with what I work with with clients and what it takes for them to cultivate search. Before we get going too much into some questions, can you just tell me a bit about what you do with Algolia what your role is?

Matt Eisnor: Sure. Dane thanks to you and the Human Element team for having me here today. Excited to talk to your audience. My role here at Algolia is I lead our global partnerships team. So what does that include? That includes our partnerships with folks like Human Element, all of our consulting partners. I also lead the majority of our technical partners, so those platforms that Algolia integrates into. And then our hyperscale our partnerships with the two primary cloud providers we work with, AWS and Azure.

Been here at Algolia for about four years. Prior to this, was in retail consulting for 13 years for one of the big four.

Dane Dickerson: Well, that’s a big job. That’s plenty to do. What are some of the ways that you’ve seen search change over this period of time and what we’re trying to accomplish with an onsite search?

Matt Eisnor: Let me take a quick look backwards into history and then a quick look forwards here as I answer this question. You know, we know 2020 changed a lot for all eCommerce teams. There was a greater focus on eCommerce. There were more digital dollars available and everybody was refocusing their efforts on an improved eCommerce experience.

Dane Dickerson: Mm-hmm.

Matt Eisnor: Back then, from a search perspective, we were really handling keyword and semantic search. Search was good. It wasn’t great. And 2020, we’re only talking about five years ago here, but search has really come a long way since then. Back in 2020, our data shows that the average search query length was about two words.

And then, well, what’s the next big thing that happened to us? 2022, chat GPT comes around, and all of a sudden we see search queries go from two to three to four words, because people were starting to get used to having a more conversational experience with the search application. We continue to see that change today, and now we’re seeing many of our customers with complex product catalogs, these search queries are going from three to four to five words.

And every time we add a word to that search query, everything from the processing of the search to the ranking of the results to the display of the results, all of this gets exponentially more difficult for not only the platform but the merchants themselves. I’ll give you a real example. Last night, I was looking for something very specific. And my search query was organic fertilizer for a shade garden in Maine.

Seven words and four intent signals. And so this is where we start to see the evolution of search from just keyword and semantic to AI search today. In that search query, that search platform needed to understand, well, I was looking for something organic. So natural as opposed to highly processed. It had to understand that I was looking for a particular product, which was fertilizer in this case, the unique attributes of that product, which were: shade garden and then in Maine signaling. I was looking for fertilizer that would work well in a cooler climate. In the past a query like that would look at the keywords. It would try to make some associations between words with semantic search. But now with AI search, we can fully understand that query and actually deliver the very specific results that a customer is looking for.

And as AI search continues to evolve, now we start to layer in that understanding with advanced personalization, customer behavior from the past, and all of a sudden we have an experience that really helps a customer discover.

Dane Dickerson: Technology is everything and the tech impacts even down to an input level of how people imagine utilizing a search function. This is something I’ve been tracking for a bit. I’ve been doing some form of SEO or site optimization for about 12 years.

And so when I started doing this, the big shift at the time was from desktop to mobile. This is where we first got 3G networks that worked well. The change that showed up the most in my work life was when swipe texting became popular and available on most devices. Because on our Google Analytics on Search Console, that was the thing that made the number of words jump up drastically the first time.

Because swipe texting is word by word search. It’s so much faster to get multiple word input and all the attention is shifting that way. On-site search, as long as I’ve been tracking it, tends to lag the off-site search habits. And there’s a seriousness of the job aspect to it of, you know.

Matt Eisnor: Yes.

Dane Dickerson: It takes time and there’s generational splits between do I make purchases on the big computer? Do I make purchases on the phone? Where do I do research? But that behavior is trickled down to everywhere. And like what you’re observing here, the computing habit of I talk to chatbots regularly, I’m engaging in sentence or multi-sentence long exchanges.

You’re right, the task to be done has increased in complexity drastically, but we’re also getting so much more information from the user about what they are trying to perform. And prepositions mean something. Adjectives mean something, and we can use this data together. Something that I try to think about with clients and wanna see what you…where you come from is thinking about all of the jobs that search is trying to do on any commerce implementation. And there’s the obvious one, like Search is supposed to find a product that I’m looking for. But it’s not the only thing. I want search to be there to help me understand what’s in the catalog overall. If I search for some of these industry-type terms, I know if I get two results back that that’s pretty thin within this product catalog. If I get 50 results back, then there’s more robust coverage. It’s an education platform too in understanding what’s in the catalog. So what are the ways you think of the jobs to be done by an onsite search system?

Matt Eisnor: I love what you just said that search is an education platform because this is one of the changes in my storytelling with customers and prospective clients is that we used to talk about search as search and discovery. Search is the easy part. Keyword search, semantic part, that search, that’s the easy part. It has been the easy part for many years now. Helping a customer discover things that they didn’t know in the catalog, that is the hard part now and will continue to be the hard part for some period of time. 

So you ask the question, what is the job of search? What are those jobs to be done by a search platform? And there’s really three parts in our mind here at Algolia. The first is we need to understand your intent and understand the query. That’s one.

Two, we need to go and retrieve those results from your product catalog, from one of our indexes, from one of your large language models that we integrate with, from your inventory management system, perhaps. So first, understand the query. Second, retrieve the results. And then third, deliver those results to that customer experience, whether that’s mobile, desktop, kiosk, clienteling application, whatever it might be.

Those are the three key jobs. And one of the things that’s unique about our product is that we bring AI in across all three of those levels in that process that Search has to do. So back to Search and Algolia as a tool to help customers discover, this is where the power of the tool really supports eCommerce teams.

And the merchants that are responsible for P & L grow basket size, grow average order value, and improve conversion rate. Because if we’re not surfacing the right complementary product or adjacent product or product that goes into a collection, then the platform’s not doing its job. That customer is not going to discover the product, and it’s never going to end up in the basket. One of the things that you talked about is how customer experience has changed over time.

I personally believe, and it’s been a while since I’ve been in a lot of customers’ web analytics platforms, but I believe the amount of time that we have with customers now, today, is shorter than it’s ever been because attention spans are getting shorter.

So if we can’t deliver the right products or the right content to that customer and that limited amount of time that we have, this is why we start seeing increased abandoned cart rates, I believe. Because I think that window of time that we have with them is just getting smaller and smaller, which puts greater and greater demands on any eCommerce platform and then specifically on the search and discovery platform.

Dane Dickerson: Right, right. And on the, on the customer side, the attention they’re giving any one site individually is shorter than ever. Part of what happens there is that there are so many more places to put that attention in the research and discovery process. And so I always try to think of the trajectory of human history in general.

People do tend to get smarter over time. And part of this is in product discovery. I actually have the option as a customer to open 20 tabs and visit 20 catalogs and very quickly know and judge if something is going to be a good option for me. And if I don’t have all of the information I need for evaluating that decision from one provider, I probably have it from enough of them to move forward and save my own time and sanity. But this is really different than at one point you could really stand out as a retailer by having a site at all. And in an information vacuum, that’s an advantage. I don’t know that you could call any aspect of modern life an information vacuum at this point. We have other problems.

So the importance there being, and I think you point this out clearly, is that that search needs to provide sufficient information quickly, accurately, to hold that attention because we’ve got to be valuable enough to make the eyeball stick.

Matt Eisnor: A phrase that we use here all the time is exactly what you just said, the right content to the right person at the right time. And I like to add a fourth aspect to that: as quickly as we possibly can so we can hold their attention. Speed really does matter in this current world that we’re in. And you started to touch on the fact that many consumers are giving up more information to us than they ever have before. Some realize it, most don’t. And I think it’s our obligation on the consulting side, on the technology side, to use that information in a way that benefits the customer. 

And if that benefit is as simple as personalization, using their clickstream data. So for example, when a customer lands at the website, Google Analytics is tracking their clickstream as they move through the site. If we use that clickstream data as a signal to indicate intent and we can improve search results, I think that’s a responsible use of that data that the customer is giving up that they probably don’t know they’re giving up to us. Or another example, purchase history.

We all know that purchase history sits in the CDP, along with my profile, the segment, the type of customer I am, how often I purchase, and my customer lifetime value. That purchase information, they’ve given that information up to the merchant. And again, I think we’re obligated to use it in a responsible way. And if we can surface products that match their buying history in the past and alter our search rankings based on that, I think that’s great. That gives the personalization that the customer now I think really expects.

And recent data I’ve seen from Gartner or Forrester, forget which one, it’s something like 86% of all B2C customers expect a personalized experience at this point, which is crazy.

Dane Dickerson: Right. Well, there’s this little bit of magic that happens where personalization done well, it feels invisible in some ways. It’s very subtle. Personalization done poorly feels obvious and intrusive. And I can rag on Amazon for this because for all of their resources, they can’t figure it out. If you’ve ever made a purchase from Amazon, and then for the following two months receive ads across the whole platform and other sites for the thing you purchased already. You’ve highlighted, like they’ve highlighted, we know what you bought and highlighted, we didn’t think at all about what that means for you. It just means that we’ll throw you in a product set. The intelligent versions of this involve thinking about, okay, what products go with this purchase? What other interests does this align with? Might you need support to be coming in? What does this mean for your life broadly? 

There’s so many directions you can go and maybe you can talk a bit about the search technology side of what it looks like to personalize and to use these data points for an onsite search to create that kind of useful experience for a customer.

Matt Eisnor: Yeah, you just brought up the idea of which products go with one another and how do we know? So this concept of collections, you know, let me talk about collections in a few different flavors of retail. So when we think about luxury retail, these merchants who are very focused on their brand, the image of the brand, the products that go with one another, the idea of collections there has been there for a long time. When we’re talking about an Algolia customer like LVMH.

Dane Dickerson: Mm-hmm. Yeah. Yeah.

Matt Eisnor: Their merchants, yes, they want the AI to support them to help associate one product with another. They want our AI to help them develop collections. But ultimately, they want the ultimate say in what happens with those collections, which products are shown with which. So just unleashing the AI, the tool on a product set to deliver collections, that’s not always what a merchant wants. And we see this with luxury retail especially.

And so they use our dashboards to control the results of the AI and make sure that all the product associations are accurate. Let’s go to the opposite extreme. A general merchant like a Walmart or an Amazon or a Target with many types of products across their catalog with much of their product data coming directly from the manufacturer in different formats with different levels of detail.

Sure, Targets and Walmarts of the world try to put everybody into a template. They try to get all the same data from their suppliers. It doesn’t always happen. So this is where AI, and specifically generative AI within search, can really help. So a tool like Algolia, we can pull in signal from some of these product attribute enrichment tools, put that data into our index associated with the individual products in the catalog, and then support a general merchant’s use of our AI to create collections automatically. And there’s quite a bit of value there, especially when we’re talking about the scale and velocity of the addition of those products. A single merchant or even a category merchant, they don’t have the time to do that.

Dane Dickerson: All of the areas that it takes to look at, to piece together the story of search, there’s some very basic levels of looking at search trends themselves and getting back reporting of, what is returning results? What’s not returning results? Does that line up? You have click throughs. That is one level, and that is one level that already is a bar too clear for a lot of site managers. And then the second is more of what you’re talking about here of how to understand collections going together, what is AI going to interpret on the site? And something to think about as well is if we unleash some AI tools across a product set and are not satisfied necessarily with what that’s returning back,

This may not be the tool’s fault. It may be that we have some information disparity here. And there’s on-site work to be done and on-site content to be created that will piece it together. If one version of AI is it’s your dumb, confident friend. And if my dumb, confident friend can’t piece together what on earth I do, that’s the thing I’m going to address with them. And what the search tools are doing with the data we can collect is give the tools that are needed to piece that together. It’s something that we do. We have a search optimization service and product. And what that looks like for us is to provide for our clients engaging with that.

Via Algolia, what search data we are collecting, what basic improvements we need, the basic things, things like synonym replacements, misspellings, these like the basic fail states, but that if that failure happens and it’s not detected by a customer, that the attention gets lost, we wanna fix those. And the reporting and suggestions based on bigger, trends we see in the data. What does this mean for the collections? What does it mean for product data? And something that we enjoy having with this being in our slate with our development services and with our marketing services is when the right action to take is we need to add content, we need to restructure content, or we need to change something about data sources.

The team is there to put those changes in place and see some iterative improvement each month on what it looks like for people to use search on the site. Because what we want to get out of is the fail state of an onsite search system is where customers go to a third party search just to find things on a specific website. There will always be some amount of that, but search for retailer website products in Google. This is always a huge set of searches and that search experience on Google is necessarily limited by so many factors. It can’t even provide consistent descriptions of what is on a page because those are auto-generated and I’ve been fighting with them for decades, decade singular.

Matt Eisnor: Mm-hmm.

Dane Dickerson: It can’t provide good filtering based on the site. So if that has become the best option for customers, we have serious work to do. 

Matt Eisnor: Yeah, that concept you just mentioned there, retailer product description, doing that search on the web via Google or Yahoo or Bing or whatever your favorite engine is, we recently got invited into an RFP with a major retailer. This was their number one use case that they were trying to solve.” I don’t want my customers going back to Google off my site to find the product that we already know exists in our catalog.”

Dane Dickerson: Mm-hmm.

Matt Eisnor: “Help us solve that.” That was the top line statement. We see this quite a bit, especially when retailers depend on the native search that might be in an eCommerce platform.

So whether we’re talking about Salesforce or Magento or Shopify or Commerce Tools, pick your favorite eCommerce platform. All of them have good, but not great search. And in many cases, there will be specific search queries where somebody is going offsite to find the product to come back on site. And that’s a real problem from a customer experience perspective. I want to go back to something you said. I really like the term you used, our overly confident friend, AI.

Dane Dickerson: Yeah.

Matt Eisnor: A real problem, especially right now, is we all get used to using AI tools. We are all learning together and AI is learning right along with us, albeit at a faster rate than us as humans. And when we think about overly confident AI, we start to talk about these hallucinations or misinformation that AI brings back to us.

And any AI tool, one of the biggest limitations of most AI tools is that when you try to figure out how the AI in the particular software you’re using comes up with its results, that’s usually very difficult to do because many of the algorithms and the code underneath, it’s not transparent to the user, to the merchant. This is one of the things that we do very differently here at Algolia.

If a customer wants to understand why a certain set of search results was ranked in the way it was, they can click through and see all of that data. And then if they want to train the AI and perhaps add a rule that the AI can react to the next time, they can do that right in our dashboard. So all of our customers can see exactly why every set of search results was found and then specifically how those search results were ranked and displayed to the customer on the page.

Dane Dickerson: Right. And re-weight how we are going to do results, which is maybe one of the biggest pieces, especially for, you know, our client slate includes quite a few B2B, includes some manufacturing. It includes these industries that do not use, they’re going to use terminology in industry specific ways. And the weights that are supposed to happen there are going to be different. Terms mean different things than they will outside of that ecosystem. Acronyms may be the most infuriatingly so. But this is where a third-party search is not going to account for that, or it’s at least not going to account for that in a universal way.

Matt Eisnor: Mm-hmm.

Dane Dickerson: If Google gets enough context to know that I’m searching specifically within the world of electrical manufacturing components, this is something I was working on a couple of weeks ago, if it has enough context in the search to understand that’s what I’m doing, then it interprets the acronyms right most of the time. And the times it doesn’t, I get results for master’s programs online that happen to share the same three letters that I’m looking for in the product selection. Or I get text support, or I get all these other areas that with an on-site search, we can weigh against it. I mean, not just get relevant results. We hope everything here is gonna be relevant, but to actually get at what these types of searches mean.

Matt Eisnor: And you’ve mentioned the word weight a few times. And weighting is so important to search and discovery for a variety of reasons. You just brought up the use case of industrial manufacturing. Well, oftentimes in that specific use case, you need a specific product with a specific product number, keyword search. You don’t need a product that looks like this or a product that is one digit off on the serial number. You need the exact product itself.

Or sometimes you might be looking for complementary products, which is where we need to understand intent and the relationship between products, which becomes semantic. One of the things that we allow with our platform is the waiting for particular types of retailers to wait semantic versus keyword search versus AI search. That’s one set of ways. How we rank the results is also important.

So something that we just released, we allow our customers to now bring in multiple additional signals into Algolia’s index, and we can rank our search results based on multiple signals at the same time. So what’s a use case like that look like? We might want to look at margin. We might want to look at products that are in inventory in the store closest to me.

We might also want to layer on top those products that are most popular in a specific geography. We can take margin, inventory, product popularity in and re-rank our search results by one of those, two of those, or all three of those. And the merchant has the ability to say, for margin, I want that to influence the rank by 20%. I want inventory to influence the rank by 70%.

And then the last 10% goes down to popularity, perhaps. So this weighting and control that a merchant can have, it’s really important. And it’s important for, specifically for the display of those results to the customer. And how does that merchant want those results to appear?

Dane Dickerson: Right, right. And something at a customer experience fundamental level is the idea that, you know, these, we talked about three jobs, search is performing, understanding, intent and query, retrieving results and delivering results. Something that is not on the list is jam the customer down the funnel on the search results page.

You know, it sounds simple when you say that out loud to say, well, of course that makes sense, but it’s, it’s tempting in digital marketing in general, where we talk so much about conversion rate, so much about steps in a funnel, to look at search results and say, well, I want to singularly optimize everything for converting as fast as possible.

Converting as fast as possible is the shortest of short-term strategies because you can pick up everybody who happens to when you make that change, be ready to do a purchase, and maybe they get there one step faster. What search is trying to do for a customer is all of that work that’s involved in making a confident purchase and building a relationship with that brand that you’re engaging with. 

And so when it comes to things like weights, there’s a balance to be struck between weights that are retailer focused, where inventory is a big one. We know we can deliver more easily. We have inventory concerns. That’s something that’s very minded toward me as a site manager. But we also have weights that are going to be the most customer minded and get them connected to related products, get them related to what they have searched for previously, that this is relationship building work. They are building the experience of “I engage with the search product for this retailer and I consistently find things that are valuable to me.” Building the search habit, that is how services become popular. That is the long-term growth of Google, that it’s consistently provided useful results enough times. And the key metric isn’t how many times the search got used and how quickly someone jumped to a result from the search, it’s how valuable that is. It’s always a little more difficult to quantify, but thinking hard about this over time pays off.

Matt Eisnor: Mm-hmm.

When I think about discovery, sure, you need a technology platform that’s capable, but once you’ve found one, the implementation of that technology is relatively easy, relatively risk-free. I think on your side, on the Human Element side of this equation, with your user experience strategy teams, I think that’s where the hard work starts. Because what you just said is important.

There are many different types of customers. There are those customers who are coming in mission driven. I want to find my product. I want to check out and I want to get out. Reducing friction, reducing steps for them so that the eCommerce experience stays out of the way. Really, really important. Then there are those customers who are coming for almost like that in-store experience. I want to browse. I want to learn. I want to be inspired by a product. I want to see how two products come together. This is where digital still has a tough time. But we’re starting to see this improve, especially in more difficult to shop categories, with assisted selling. And search can power any assisted selling module, because really, what assisted selling is, is just a series of queries that happen, one question after another. And each one of those questions in an assisted selling module is really just a search query. So your part’s the hard part.

Dane Dickerson: Mm-hmm.

Matt Eisnor: How do I allow a customer to find both of those experiences on a single pane of glass, on a single page, allow them to opt into their journey and then allow the journey to take them? And then if they realize that they don’t want assisted selling, you also have to give them an opportunity to get back to, well, I just want to transact and get out. That’s the hard part. I’ve always believed that’s the hard part.

Dane Dickerson: Right, allowing for flexibility is crucial and allowing for users to change the way that they want to engage with the site, with the brand. Because this is what you can ruin with personalization when not done thoughtfully, is to say, okay, we picked up enough signals that we put you in the category, of you’re a direct purchaser and we’re going to focus everything on making fast purchases happen. Why have I changed my mind? Because it’s not working for me. Is there enough flexibility in the site experience to go another way? Is there enough flexibility in search that I can engage in multiple ways and know I can get back to what the experience was before if I figure out what I need to?

Matt Eisnor: And back to your Algolia optimization services for a second. The initial implementation of any search product, Algolia specifically, it’s relatively easy. Algolia or any search product can be implemented in weeks, not months, for simple use cases. But it’s the application of the technology over time that really matters.

It’s understanding how the new technology has impacted user experience, learning from mistakes perhaps, learning from successes, and then getting into a phase two, phase three, phase four implementation of the technology to further extend, to add additional innovation, to bring in new use cases, and overall improve customer experience. Many of our customers here at Algolia, they’re eCommerce teams, they’re very busy.

They implement search, they check a box because somebody told them search is not good. Usually it’s a key executive or a board member or somebody. It’s usually not a customer who’s yelling at them. Some ELT member, a board member says search needs to be improved. Great. Let me go find a search product. They implement search, they check the box, they move on to the next emergency.

Dane Dickerson: And it’s a credit to the value of the tools that almost all the time you can do that and the results do get better right out of the gate. But what you are purchasing, the thing that is more valuable is the set of tools for refinement and for improvement and everything that lets you jump off to. And yeah, it works well for us and for our clients to have that tool set available and those reports in the mix of the whole picture of this site, alongside the development, alongside other marketing to get from the starting positions of good results that have not been tailored to exactly your customers, to here is a search experience that really is unlike any other site that is going to exist in your industry and is so focused on how to get the best results for the people you work with.

Matt Eisnor: Yeah, it’s funny. Oftentimes I’m asked a question, something like this. Who’s your best customer that uses Algolia completely? The answer is we don’t have one, if I’m honest. We have customers who use aspects of the program in a best of breed way, but there is no one customer that uses our platform completely. Walgreens, for example, they use Algolia really well for inventory aware search.

Their implementation is almost perfect there. Hobby Lobby really helps their customers discover adjacent products when they’re searching for something specific. Well, here are the next three products that go with that product you search for the first time. With LVMH I mentioned earlier, their merchants do a really great job of controlling the AI and making sure the AI is delivering the results that they want. But there is no one perfect customer. Nobody ever uses technology to its full extent.

Unfortunately.

Dane Dickerson: And unfortunate on one hand, on another, there are much more interesting things to be than perfect, I think. You know, both as a brand and as a customer. And there’s some humility in the fact that like the way discovery is gonna work in an industry changes over time. And something that I just spent a lot of hours with, especially from the SEO side, is the ways that a business can think about its catalog can start to really diverge from the way that its audience is looking at discovering those same products. And sometimes it’s a legacy thing. And B2B is maybe the biggest example if you came from a product catalog mail order world.

The way these businesses might think of their product slate is literally picturing the print catalog in their mind and where physically that’s located in the catalog. It’s about product numbers. And may be obvious to say, but that is not the world we discover products in anymore in 10 different ways. We don’t use print that way.

When we do use print, it’s because the visual component of print is being brought into that experience. That’s not really focused on product numbers. And having the analytics and the reporting out of an on-site search product is one of your clues into what actually is changing in your industry as well. And I’ll just voice it here from SEO.

For me doing keyword research for offsite, I can get 10 times more value out of that brand’s on-site search intent and results than I can picking through Google Search Console. Because the data available is, it’s higher quality, it’s more complete. And this tells me more about where things are going for that audience.

And I know there’s always gonna be some reflection of what they do in the onsite search and what’s happening offsite. You can line those up.

Matt Eisnor: Speaking of SEO, one of the basic best practices that many customers using a third-party search tool don’t do is taking that off-site search signal from Google, taking that query, piping it into the on-site search tool so that we can create dynamic landing pages when they come. We’re just not doing that often enough.

And it’s a relatively simple implementation, requiring the integration of your analytics tool with something like Algolia. But many of our customers aren’t implementing events because they come back to, they solved a search problem, saw results, and moved on to the next emergency. 

A couple of customers in this space that you’re talking about specifically, they’re joint customers of ours. So Pentair, a common customer between the two of us. Watsco, another common customer between the two of us.

Dane Dickerson: Mm-hmm. Yep.

Matt Eisnor: I think both of them are really, really strong examples of how B2B experience has improved dramatically since 2020. And I think that’s where I started this conversation was, you know, that refocus on eCommerce back in the COVID days in 2020, it was the B2B companies who really realized first that they had some ground to make up.

Because customers were bringing their B2C expectations on experience into the B2B website and they just weren’t getting it at first.

Dane Dickerson: Mm-hmm. And there’s one way to look at those changes, which is to raise your fist and say, liked the way it was before. It’s how this has shifted and there will be just generational turnover in who is making purchases, personally, for businesses, for all of these things. 

It’s just like the people on the ground. People retire and when somebody new is brought in, it’s 25, has been asked to make purchases, the way they’re going to do it is going to mirror their experience up to that point. And so they don’t think, I’m bringing in B2C search practices in a B2B and I should know that’s wrong. That’s just how they do research. And we can meet them where they are and look at that with curiosity and what data points that brings in to make the product of search perform better than it could have before. Maybe a way to go out here is, we’ve thought a lot about what has changed in the last few years and the way that search is happening now. 

What are some of your thoughts about what is going to change in on-site search in the next few years? And what should we be watching for?

Matt Eisnor: I almost wish that you didn’t ask that question because it’s really difficult to answer. The pace of technology is changing faster than it ever has been before. We haven’t used the word agentic yet. Okay, I guess now is our opportunity to use the term agentic. But really what is agentic commerce? It’s really just the orchestration of multiple AI tools into a single experience.

That is here already. Agentic commerce, AI orchestration, this becomes a big topic over the next 12, 24, 36 months. Beyond that, I don’t know. And I don’t know that we completely understand as Algolia. We have a vision. But one of the things that we’re doing with the product is future-proofing it. We’re trying to keep everything as open, as flexible as possible, because we know that we can’t predict the future.

Nobody would have expected here in 2025 that we would be this far down the AI path, if not for ChatGPT back in 2022. So what comes out in 2025 or 2026 that changes what we do by 2030? Who knows? So we’re continuing to evolve the product to become more flexible, more open, so that there are more options for our customers and merchants with the tool.

Dane Dickerson: It’s a cop-out answer, but I’ll take it. No, it’s fair. The thing that I try to keep in mind when looking at the technology changes, looking where things are going, is how important it is to collect data points on what the actual people are doing and what value did they get out of that experience? And sometimes this is the most old school, which is where you find 10 people that did some reach them purchases and you call them or you email or you reach out, actually ask a person what they did. Because you can look at Google Analytics, say…

Matt Eisnor: Actually asking a customer for feedback. Yeah.

Dane Dickerson: …okay, well these are the services people are using to find us. ChatGPT, that’s a service people are using to find us. And that has literally been true for Human Element as we get increasingly more leads that come out of these sessions. What does that analytics tell me about what they did there? Nothing. The only way that I know is because these are a black box of information is by literally talking to these leads and saying like, what kind of conversation were you having? What information were you hoping to find? And this is something to apply in all of the discovery channels, including how, like how did you navigate the website? And you find these use cases of saying, well I search, did a search on the site. I found a handful of products. I threw them out of the chat bot. I asked it to summarize some information.

I did comparison shopping, came back on my phone, which made a new session, and that’s where I purchased. And this sequence of steps, there’s steps all along this way to optimize for that there’s no possible way to get this experience summarized in an analytics platform. But if I pay attention to that, I’m curious about that now, I can put the tools in place to tie that together. So I don’t know what will happen.

Matt Eisnor: This is where I still think, I still believe you have the toughest job in the industry at this point. Pulling these tools together, pulling the experiences together so that it’s not disjointed and is authentic for a customer. That is the tough user experience strategy that Human Element is great at. The technology part is the easy part. You have the hard part.

Dane Dickerson: Well, thanks. I’ll take all the help I can get.

On that note, we have hard jobs. Matt, I thank you a lot for taking some time to talk through these questions. I’ve learned a lot and I really appreciate everything we get to work on together and everything we can collaborate with where search is going, what we provide for clients. Do you have any parting thoughts to share?

Matt Eisnor: This next chapter in the next 12, 24, 36 months, like I said, the pace of technology is gonna move faster than ever before. I think sharing information in a forum like this or in any forum is really important right now because we are all learning together. There is no one consulting company individual who has all the answers.

We are all learning together. So to the extent that we can share this information openly so that many benefit, I think that’s important right now. But Dane, thank you to you for the time today. And thank you to the Human Element team. I really appreciate it.

Dane Dickerson: Thanks a lot, Matt, we appreciate it too. You have a good one.

Matt Eisnor: See you later.

Leave a Reply

Your email address will not be published. Required fields are marked *