Why Enterprise Site Search is Hard for IT Teams Let's face it, managing a large,…
If a moat is an uncrossable chasm, then data may not be a moat. No amount of data can make it impossible for the competition to catch up. But it could be an inconvenience to the competition who is trying to climb your walls.
The struggle is that there is very little data a company can capture that other companies can’t capture. Everyone has access to third party data, but maybe first party data can slow the enemy down a bit. There is also an interesting counter-argument: that massive databases, real-time response and hyper-personalized experiences do actually make that difference.
In this episode Steve and I explore how data is like oil: it makes the engine run. But data as a differentiator is not the game. The game is, what do you do with the data?
We also explore how Word is now incorporating AI-based features to improve writing within Word. As I like to say — all data is training data. Never to be totally one-upped on the AI game, Google also dropped an interesting release: CallJoy, which allows small businesses to answer calls using AI.
This is a big deal. The more that we can make this technology visible in practical ways, the more trust there will be in the technology in more sophisticated ways.
2:00 The empty promise of data moats
13:30 Counter-argument: data as differentiator
20:05 Word is using AI to Improve your writing
25:05 Google launches CallJoy
SearchChat is available on
Search Chat is SoloSegment’s podcast dedicated to all things search AI and content marketing related. Who is SoloSegment? We’re a technology company focused on site search analytics and AI driven content discovery to improve search results, increase customer satisfaction and unlock revenue for your company. If you think we might have the answer to your conversion problems, feel free to connect with us.
Tim Peter: Hi, I’m Tim Peter and welcome to SearchChat. SoloSegment’s podcast dedicated to all things search, AI and content marketing related. Who is SoloSegment? Well, we’re a technology company focused on site search analytics and AI driven content discovery to improve search results, increase customer satisfaction and unlock revenue for your company. SoloSegment, make your search smarter. You can learn more at solosegment.com. In this episode of SearchChat, SoloSegment’s, CEO Steve Zakur and I talk about Andreessen Horowitz and the empty promise of data moats. We also look at the counter arguments that explains why data may in fact be a moat and how you make data work for your brand and business. Finally, we give some great examples of how people are doing it in the real world every single day. All that and more coming at you on SearchChat right now.
Tim Peter: Well, hey Steve, how are you doing today?
Steve Zakur: I am doing really well. I’m very happy not to be sitting in a car today. I had one of those days yesterday where I was driving to Boston, took a couple of calls on the way to Boston, had meetings in Boston, had dinner in Hartford and then back home. So it was a busy day. Lots of good conversations and some good business being done, but at the same time, time in the car is, well, I guess better than time in a plane, maybe I’ll leave it at that,
Tim Peter: Sure, but not just not just road warrior in this case, but the I 95 warrior.
Steve Zakur: Yeah, exactly, exactly. Good Times.
Tim Peter: Well that’s fantastic. Well, uh, great talking with you as ever. I think we’ve got some really cool stuff, this week and I want to start with this really interesting post that was written, over on A16Z.
Tim Peter: So this is Andreessen Horowitz’s blog specifically, and they’re talking about — people have long referred to data as a moat. So this is Martin Casado and Peter Lauten writing at Andreessen Horowitz, A16Z.com. And they really make the case that data is not a moat, they call it the empty promise of data moats. And, I’m gonna read just a little tiny piece of this where they say, you know, data has long been lauded as a competitive moat for companies and they say, but for enterprise startups, which is where we focus, we now wonder if there’s practical evidence od data network effects at all. Moreover, we suspect that even the more straight forward data scale effect has limited value as a defensive strategy for many companies. And this is a thing that we have talked about. More than once, right? Both here on the show and you know, when we’re actually just talking about work.
Steve Zakur: Yes.
Tim Peter: And I’d love to get your take on it. You know, first and foremost is data of moat or not? And in either case, why or why not?
Steve Zakur: Right. Well, first of all, let’s define what a moat is. Yeah. If a moat is an uncrossable chasm, then data is not a moat. If we think about data as an inconvenience to climbing the walls of competition, maybe. And I think it’s usually shorter term defensibility, but I tend to agree with what they’re saying in this article that there is very little data today that a company can capture, that other companies aren’t capturing. And by the way, you can, who doesn’t buy third party data to augment their first party data, right? So it is a very large industry that allows you to do that on almost, you know, pick your industry, pick your topic, and you can buy third party data about it. And if you can’t, you can do primary research to gather that third party data. So, you know, I think the days when data really was defensible, and we’ll talk about D&B I think in a little bit, but we’re really was a defense where you had kind of just in having the data and being able to bring the data into a single repository so that it can be useful.
Steve Zakur: I think those days are over. I think those days are over because first of all, the inventory of data is, you know, again available, you know, at many, many vendors. So, there’s not a lot of scarcity, if you will, around data itself. I also think that the mechanisms by which data can be aggregated and matched and whatnot, a lot of that what used to be the application layer where you really had to go off and create something unique to, to bring together data sets, a lot of that has commoditized and become platform that you can get off of AWS or you know, downloaded from some [unintelligible] or whatever. Right? So it’s, it’s a lot of stuff has become application layer has become kind of platform layer if you will. And so, so I think at least for those two reasons, you know, data is a moat.
Steve Zakur: But that said, if you think about the moat in terms of filled with alligators and water and, and kind of slowing down the attacker, I think in the short term, especially if it’s first party unique data, it might give you some, you know, kind of first mover advantage. But I think as soon as the market figures out that oh they are just using x data or combining x data with y data, and the market’s pretty good at sussing that out because if nothing else, they’ll hire your engineers. The advantage is going to very quickly be eaten away. And so I think that the data of itself as a differentiator is not the game, right? The game is what are you doing with the data? And platform versus application layer, I guess is the way I think of it from a technical perspective.
Tim Peter: Right. And I actually, I want you to expand on that for just a moment, but you and I were talking before the show when I’ve, I’ve kind of repurposed of metaphor other people have come up with, this is not my original idea by any stretch, but, uh, you know, I like this concept of data is like oil. And not in the sense that most people mean it, but in the sense of, you know, it’s something that makes the engine run. Yeah. Right? Data’s not a moat in and of itself, but it can be a moat for you to cross if you don’t have data, right? Like you have to have some data before you can build those application layers and things along those lines. But in and of itself, it doesn’t have tons of value if you don’t do something useful with it. So I’d love you to expand on that just a little bit. When you talk about the application layer is what makes the data interesting or makes it potentially a moat and what does that look like and what should people be doing there? Or how should they be thinking about it?
Tim Peter: Oh, no, go right ahead. Go right ahead.
Steve Zakur: Because where we do differentiate by the way, where everybody differentiates is somewhere beyond the data layer. So it is actually in the application layer. So what are you doing with that data? What is the unique kind of point of view and technology that you bring to the table to make your data more meaningful? And so if you look at what we do, you know, yes, everybody captures “no results” and most people capture clicks, but we capture this other set of behaviors that allow us to identify failed searches and by of course mathematical properties, the inverse being successful searches. But we look at that set of data in a very unique way that allows us to identify those successful behaviors. And that’s how we differentiate. And so that’s, that’s one example of I think of how, you know, companies need to certainly be thinking about the data layer. But again, I don’t think it is a persistent or a durable differentiator in the long run because there’s no patent, there’s no Ip that protects the ability to gather most sorts of data out there on the marketplace.
Tim Peter: Well and the data that, you know, even the data that you get as a proprietary data, which I mean, we would not only argue that you should, we would absolutely argue that you must, you know, collect, you know, your competitors are collecting comparable data. It’s not going to be exactly the same. It’s not going to necessarily tell them the exact same story, but it’s going to tell them a useful story. Just as you collecting your data should tell you a useful story.
Steve Zakur: Last night I was having dinner with somebody, in a large personal lines. I was in Hartford, right where all the personal lines insurance companies are. But I was having dinner with them and it reminded me of some discussions I had previously with a company. And this is one of these companies as well known. They have their ads on TVs. And one of their ads a couple of years ago was “we’re going to show you the quotes for what you would pay us for your insurance and what all of our competitors would charge you for that same insurance. ” Right? Sure. And was it because, um, this company had a unique set of data that allowed them to do that because of course their point of view was we’re really good at underwriting and assessing risk and so therefore if somebody else is going to charge you less, they’re taking more risk than they should.
Steve Zakur: We’re happy for you to go to go do business with because they’re going to go out of business quicker because we’re better at it. They are right. And that was the whole price. It wasn’t that they had better data because everybody has the same underwriting data. We know how many accidents you’ve been in, how old you are, you know all that stuff, what your health condition is. All these insurance companies have that data. But what they had is in the application layer, they had a better technology for evaluating and pricing risk. So that’s just another example of where differentiating yourself on the data that you gather is very, very difficult. Now it could be that this company was, you know, had identified some piece of data somewhere that was relatively unique, but eventually somebody’s gonna figure that out, right? That’s, they probably got it from some consultant and your competitors are going to hire that same consultant.
Steve Zakur: And there you go. Now, now their competitors have the same piece of data. So it really is in what they’re doing with the data that allowed them to differentiate their underwriting process, differentiate their quoting process and with confidence be able to sell. Yes, you should go buy from All State instead of us because they’re going to write a bad policy from their perspective and they’re going to go out of business sooner. So it’s that sorta thing.
Tim Peter: Right, like, well, and even if–I mean yes, everything you just said is absolutely correct. But even if it is something where, you know what, maybe All State’s financials are set up differently, that they can service that customer more effectively, more efficiently, whatever. That was irrelevant. What they were saying was this is a customer that is not a good customer for us, so we’re happy to have them go somewhere else.
Tim Peter: So even if again that customer is a good customer, somebody else can have ’em. Not our customer, not our benefit to have them.
Steve Zakur: Right.
Tim Peter: Well it sounds like you’re kind of mirroring, there was a counter argument to this A16Z article and you know, I want to be really clear. It sounds like we’re agreeing with the A16Z article, but also it’s interesting the counter argument because I think we also agree with this. This guy Alex Iskold on Twitter, in replying to the empty promise, the data moats, did say that there are three examples that are interesting, a) massive databases b) realtime response and c) hyper personalized experiences. And those are kind of interesting. I thought it might be kind of interesting for folks to hear why you think they’re interesting.
Steve Zakur: First it’s this notion of the massive database. Right. And this is interesting. He cites some of the examples are kind of medical history databases. Real time data sets, like Waze has, and you know, I do think that there are some situations where data is an advantage. And by the way, maybe a defensive line, like Waze, who’s competing with Waze these days? I mean, I didn’t know anybody. Now I gotta tell ya on my trip yesterday coming back from Boston–I don’t know why I started the trip–oh, I do know why I started the trip using Apple. The reason I started the trip using Apple was for some reason Waze wouldn’t find the Starbucks I wanted to go to. So I used, I used Apple’s map to get me there.
Steve Zakur: And at one point, later in the day, I had both Apple and Waze running at the same time. And what was really odd was I heard Apple’s voice telling me about Waze’s map, which was kind of strange, but regardless, during that trip back, you know, it was clear to me that, you know, again, both massive databases, both collecting real time data. So when you think of like traffic data, you know, they’re measuring constantly, how are cell phones moving through the environment? And they’re gathering other data to bring it in to understand, oh, there’s a slow down on I84 for, you know, take you know, 191 and go around it or whatever. Right. And they both made that recommendation. So, here’s two competitors — and Google probably does the same thing — so here’s three companies that these are massive realtime databases. And I would argue, you know what? Waze doesn’t have a super defensible market position there because these other competitors do. Now where Waze does somewhat differentiate itself is all that social stuff. Oh, there’s a pothole here. Oh, there’s a police trap over there. Right. So that’s where I think Waze does differentiate yourself. So I’m not sure it’s a perfect example of that sort of situation.
Tim Peter: Actually, well, what they do, Waze is owned by Google and what they do that’s very interesting, and it does speak to what you’re talking about, is the reason Google maintains Google maps and Waze as separate apps is because they’re each optimized for different use cases. You know, if you think about Waze, Waze only works if you’re in the car, right? They don’t offer walking directions, they don’t offer transit directions. Waze is optimized to be the GPS replacement, whereas Google maps is optimized to be, okay, how do you get from point a to point b, regardless of how you–
Steve Zakur: Very handy in New York City when you’re trying to figure out what subway to take.
Steve Zakur: Right. But if you’re driving Waze is a far better experience, so that differentiator to your point is very much around that application layer of, what is the use case?
Tim Peter: Who is the customer in this case? Oh, you’re a driver? You want Waze. You’re looking for multimodal transportation? I’ve got to walk somewhere and also get a subway or you know, I’m going to drive partway and then I gotta walk. Google maps is going to be a better option for you because that’s what it’s designed to do from the ground up. But they are relying on, actually very much the same dataset, and then differentiating on the application layer. Yeah. Right?
Steve Zakur: Yeah.
Tim Peter: Yup.
Steve Zakur: You know, it’s interesting though when you think about this and Alex’s statement about massive databases, and I’ve referred to D & B, Dun and Bradstreet, earlier. If you go back even 15 years, there were no alternatives to Dun and Bradstreet, no viable alternatives. I mean, they were the game for companies looking for information about companies. That was the right game. And you could tell that there were no viable alternatives because they exhibited kind of monopoly behavior. Like for example, their product was hard to use. It had all of these hierarchies and relationships between data that, it was clear that they were the big dog. And you know, maybe in a pre-digital world, even in a kind of “Digital 1.0” world that was defensible. But when you look at the power of company information has moved from the data layer to the application layer.
Steve Zakur: And every website that wants to do a realtime personalization based upon say the industry that you’re in or the company that you’ve worked for or whatever, well, they’re not using a Dun and Bradstreet product anymore. They’re using Demandbase or Clearbit right? And I’m sure D&B is in that business, but they’re not the top name that you see in the Gartner quadrant, right? It’s going to be companies like Demandbase and Clearbit. And these are companies where, I gotta tell you, I’m sure somebody at D&B regrets not being the Demandbase or Clearbit of this new world. And so that’s an indication of where data in the past might have been a moat. It may even have been a wall, right? A big stone wall. But I think today we’re seeing that that data has commoditized or is relatively easy to replicate. And so, you know, Demandbase’s strength is not really in it’s data. It’s in the fact that it moved first. That is, that has got a really great integration layer and ability to deliver this data in ways that are useful to people and to applications. And that again is the differentiator. It’s in the doing, in the execution and not in the gathering of the information.
Tim Peter: So that’s a great transition because we, you and I talked about two different examples, that had been announced in the last week or so, that really show what you’re talking about in practice. And one is very much a B2C, Eh, I shouldn’t say entirely B2C play, but it’s business to user anyway, right? If not consumer, they’re talking to an individual person, whereas the other is very much B2B for small businesses specifically. So I just thought it’d be fun to get your take on each of these separately. The first is, Word is now incorporating AI based features help improve writing within word. So why don’t we talk about that one first and then we’ll go to the second one, if that’s okay.
Steve Zakur: As we’ve rolled out machine learning based products over the past year or so, one of the things I’ve realized is all data is training data and you know, so as I saw this new feature that’s in Word, in Microsoft word, I said Bob, that that’s really handy. I get that, that’s useful. I mean, Google came out with a month or two ago, maybe it’s a little bit longer than that, but a month or two ago they kind of do a type ahead as you’re typing emails. And this thing is smart, right? I find myself accepting almost every recommendation. And by the way, when I do a word change, it seems to learn that word change so it gets better. So the first thing I thought is, oh, this is brilliant. This is some great training data that Microsoft’s getting as it makes recommendations.
Steve Zakur: It’s going to get better at making recommendations and they’re probably using this data elsewhere in their stack to make, you know, who knows, make Bing smarter or maybe come up with a new product. But regardless, I think this is really fascinating. If for no other reason, yes, people will be better writers because of it. So that’s good. But the other thing I really dig about it is the fact that yeah, it gives everyday humans exposure to how machine learning makes life better. Text analytics makes life better for us all. Right. It is, it is kind of a very practical and visible example of that. And I, and I gotta tell you, I talked to a lot of marketing execs. Yesterday was my hundred and third conversation. Not that I keep track or anything, but my hundred and third conversation with a marketing exec this year and you know, they all struggle.
Steve Zakur: Like they’ve learned the Buzzword Bingo. They know some of the stuff that works for them, but they are still kind of wondering, you know, how does this all work? And I love when something very practical like this appears because it shows humans how it works. And I think it kind of fosters in business because that’s where the leadership needs to come from. It fosters in business the, Oh, I get how this might work now in some other context, that might be a context that’s not said by the most recent sales rep who sat at their desk. So yeah, I dig this. I think this is very, very cool. I’ve loved Google’s kind of version of this. The Word version seems much more sophisticated. And if for no other reason than, hey, everybody’s going to be a better writer, that that can’t be a bad thing.
Tim Peter: That’s great. Yeah, exactly. Maybe we’ll see better written emails and things like that. So the other one that I thought was really interesting, and by the way, I love your point about, showing it in the real world.
Tim Peter: I, much like you, I have had conversations with probably a couple hundred folks over the course of this year where literally people say, “yeah, but you know, is this real or is this just hype” or you know, something buzz that’s not really going to go anywhere. And so it is great when you can point to these examples of no, this is real. And here’s a very real practical example.
Steve Zakur: And it would be wonderful. Like we have a new product called GuideBox. And you know, when I tried to describe how that works to people and, you know this, but for our listening audience, what it’s looking for is patterns in visitor behavior and then it’s making content recommendations. And boy, if everybody were using this Microsoft Word thing, it would be great to say, Oh, you know how Word figures out that the paragraph should be worded like this or that instead of talking about this topic, you should be talking about that topic.
Steve Zakur: You know how that works? Oh, well this works the same way. And people will go, Oh yeah, I get that. But sometimes when I talk to folks who are less technically savvy or kind of haven’t read up to it, I really have to step back to “see spot run” sort of language to, you know, talking through and, and I’m happy to because I want a) I want to sell them something, but b) I want them to understand how AI can be practical. And I, and again, I think this has got to be a boon for that.
Tim Peter: Absolutely. Well, so the other one that I thought it was interesting, this B2B one than I thought was really interesting. People probably remember a year ago, uh, Google IO. Google introduced, it’s AI that could go ahead and call restaurants for you, or call hair salons for you to make an appointment. So what they’ve done now is they’ve taken that AI, that same technology and they’ve turned it the other way round. It’s called Call Joy. And it’s actually designed to allow small businesses specifically to, instead of place those calls, answer those calls
Steve Zakur: My phone’s going to talk to your phone.
Tim Peter: Right, exactly. Exactly. I’ll have my phone, get in touch with your phone and we’ll do something. But I think it does build on what you were just talking about. So again, I’d just love your take on this and it’s potential in there as well.
Steve Zakur: When I first saw Google’s tech, and it literally has very specific use cases. I think there are three now, but making a hair appointment, which I don’t have to do all that often anymore, but making a hair appointment was one of the big use cases. And yeah, I thought that was kind of neat. I mean, it was a good kind of practical way of getting people to understand the technology because it’s something we all do and that, that’s pretty cool that they can, they have a use case that everybody can kind of connect with.
Steve Zakur: I think that this now from–that was kind of the consumer version. Now you flip it around and you say, you know, cause again, if I’m the hair salon owner, now I’m thinking, wait a minute, I might be talking to computer and that’s kind of awful and I just might feel a little weird about that. But oh now there’s a technology for me, right. So that I can kind of be more productive and have a better, more consistent customer experience. Yeah. And somebody who’s not going to get annoyed the third time a customer reschedules. Right. So there you go. This is, this is a win for all those reasons. But again, I think it all comes back to the more that we can make, as technology leaders, the more we can make this technology visible, and in very practical ways, I think the more trust there will be in the technology in more sophisticated ways, in ways that we don’t understand.
Steve Zakur: Because while I applaud Watson and IBM for starting to try to cure cancer, the doctor patient relationship is one of the most intimate things out there. And yeah, I want the best information in the world, but I don’t pick the doctor with the best library. Right. I pick the doctor who has demonstrated clinical expertise and that’s the kind of doctor I want to go to. And so it’s hard for me to trust. I might be interested in that technology. If it cures me, I might then build the trust. But it’s kind of a weird place to start with. Oh, let the computer decide, not decide that’s the wrong word. Let the computer assist in life and death decisions. It’s like, well, no, no. Why don’t we start with scheduling your hair cut? Let’s start there and that works out good. You know, then we can, you know, turn over kind of more complex things, as long as they don’t become sentient and take over.
Tim Peter: Absolutely. I mean, there’s no argument here, although one, but what I also think is really interesting is it brings us full circle. When we talk about these two examples, the Word example and the Google example in that, in neither case so far as I’m aware, are they really relying heavily on your data in the first case, right? It’s not Call Joy needs to hear all your phone calls before it figures out how to work and it’s not, Word has to analyze all of your writing. In either case, I think it learns from yours, but it starts with, hey, these are use cases we see all the time and the data that Google has collected, the data that Microsoft has collected, provide plenty of guidance to say, okay, here’s how we improve on what exists right now. And again, you know, deliver it at the application layer.
Steve Zakur: I think that’s really interesting point. And it’s interesting bringing that kind of home. It’s one of the reasons Google search works so well is because it doesn’t need this company’s specific data to be able to bet because it knows the patterns. Right? It’s got 80 bazillion pages in its index. And it’s quite frankly, one of the reasons site search is so hard because that does rely upon your specific use case.
Tim Peter: Well, and notice Google got out of the business of site search. Google went “hm, search is tough when it’s this kind of search, we’re not going to do that.”
Steve Zakur: But yeah, it is interesting that where you have the opportunity to use those massive datasets, boy this is back to another thread in this, that might be, I don’t think it’s defensible, but , it’s there, there’s probably a moat that does that makes it more difficult in that situation.
Steve Zakur: So I think we’ve argued both sides of the massive dataset. Yeah. It could be or could not be a moat, but, boy it’s like most things in life, it depends.
Tim Peter: Spoken like a true consultant Steve.
Steve Zakur: With a deeply unsatisfying unstatement to end this.
Tim Peter: No, I think, I think you’ve made a couple of more than bold statements to route. And I think, given where we are, this is probably a good time to wrap. I just wanted to give you a chance to get the last word. You know, anything you’d tell people about thinking about data and how they use it as a moat for their business or more importantly the application layer.
Steve Zakur: Yeah. And that’s just it, right? It is. What are you doing with the data to make it, you know, make the application of that unique and defensible?
Steve Zakur: And you know, when I talk to people about how we think about search success, that that drives all our decision making about data. It’s not only do we want a good predictor of success, but I want to be able to do it in ways that other people can’t copy. And so that’s part of what we do is we look for that success, but we look for it in very unique ways. And so I think any business, no matter what industry you’re in, if you’re thinking about, creating a defensible business model, you can start with some data, but in the long run it really is going to be what do you do with that data? How do you bring it to the market in that unique and defensible way. And that’s really the key, to kind of competitive success.
Tim Peter: Perfect place to stop Steve as ever. Well great talking with you again. I’ll look forward to our next discussion and catch up with you next time.
Steve Zakur: Thanks very much Tim.
Tim Peter: All right, bye now.
Tim Peter: SearchChat has brought to you by SoloSegment. SoloSegment is a technology company focused on site search analytics and AI driven content discovery to improve search results, increase customer satisfaction and unlock revenue for your company SoloSegment: make your search smarter and learn more at solosegment.com. If you like what you’ve heard today, click on the subscribe links. You can find that at solosegment.com/podcast on iTunes, Google Podcasts, Stitchr radio, Spotify, or wherever fine podcasts can be found. You can also find us on linkedin at linkedin.com/company/solosegment. On facebook at facebook.com/solosegment. On Twitter using the Twitter handle @solosegment or you can drop us an email at firstname.lastname@example.org again, that’s email@example.com. For SearchChat. I’m Tim Peter. I hope you have a great rest of the week. Thanks so much for joining us and we’ll look forward to chatting with you next time here on SearchChat. Until then, take care everybody.