Computers can do a lot, but they can’t do it all. This week, Nationwide’s Ted Williams shares platform stories, and the critical factor of human intervention in truly useful enterprise search. Ted serves as SharePoint Information Architect with Nationwide and is intimately familiar with what makes enterprise search tick, why enterprise search fails, and what he and his team are working on to make search smarter for their users.
SharePoint Information Architect
Pete: Welcome to “Shared Insights,” the podcast from BA Insight. My name is Pete Wright, and I’m joined today by Ted Williams. Ted serves as SharePoint Information Architect with Nationwide, and he’s here today to talk with us about the nuts and bolts of enterprise search, why it fails, what he and his team are working on to make search smarter for their users. Ted Williams, welcome to the show. Dare I say, “Play ball.”
Ted: Play ball. There you go. Out of the park.
Pete: Because, yeah. I was gonna make a subtle reference, but, you know, I feel like you and I are cursed in similar ways. You obviously have the legacy of Red Sox drive, Ted Williams, on your back, and I’m Mr. Wright. People, as you may guess, have been looking for me all their lives. I feel like we have, sort of, a comradery here we need to share at the outset.
Ted: That just means you’re to my right, right?
Pete: Yeah. You know, I’d much rather take that. So, let’s talk a little bit about search. To set the context, Will, first take on some of the pain points you’ve experienced in your enterprise search efforts. Obviously, there isn’t a single thing to point at when search goes South. Rather, you know, you have opportunities to resolve users needs and the challenges that companies have to overcome to do that. So, reflect for a bit for us on where enterprise search could use a hand in your experience.
Ted: Well, I think one of the things that we struggle with, and that is convincing leadership within an organization to the need to staff search to begin with. Because, you know, we’ve been led to believe that we can just buy this box and plug it in someplace and search is gonna be wonderful. It’s all gonna be like Google, and people are just gonna be able to go in and type what queries they looking for and engine’s gonna know who they are, and what department they’re working in, and what they specifically are looking for in some way.
And, we don’t need the staff or to think through the whole backend pieces of it. And, I guess right now, that’s where my head is at is trying to overcome that mindset, to realize that search needs to have a staff. It needs to have people who can sit at the table and have a voice and not just be a secondary thing we just hope to. We can plugin and make it work.
Pete: What is it then that the artificial intelligence can’t do that people can do? Because, at some point, these vendors who are promising the world are making a compelling case to people who hold the purse strings?
Ted: Yes, they are. And, I think they failed on to…what people fail to understand is you still have to have your content structured in some way. It has to be tagged, or some kind of formats, or are in some location. It has to be structured in some way for those artificial intelligence tools to understand what it is that it’s gonna be serving back to you. You know, I’ve seen some knowledge tools, you know, that provide us with the helpdesk type stuff where it can serve back some interesting results. I can ask it a question, it gives me back an answer. But what we have to understand there is those results have been formatted to be provided as answers.
You know, the engine then can know where to go and find that little bit of information based on that query and bring it back and present it in a way that’s usable. What I’m hearing now is people will say, “Well, we need Alexa in the enterprise.” Well, first of all, I just can’t see a cube farm where everybody’s just talking to their search. I really struggle to see that happening very well. You know, everybody’s got a little Alexa box on their desk, and you’ve got conversations going all over the place with the computers.
Pete: Yeah, we got one in our house, and it doesn’t take long to devolve into just strings of profanity at the echo for not getting it. Like, I don’t know how that…I can’t see it. I can’t see it yet.
Ted: Yeah, I’m the same way. I have Cortana plugged into my X-Box, you know, and I walked in the other day and tell it to turn on and it does nothing. And my grandsons were looking at me like, “Why do you waste your time and money on that thing?”
Pete: Right. How quickly we’ve become the insane person we all thought our grandparents to be.
Ted: Exactly. But anyway, I think that we have this idea that we can just go speak to our computers and our computers can immediately interpret the meaning behind what we’re trying to convey, the feelings behind that, you know, and it just doesn’t work that simple. I’ve tried to go in and try to write rules for tagging content for instance, and, you know, depending on where the repository is at, the type of content that it is, every rule has gotta be nuanced based upon the user’s expectation. And, that’s difficult to do intelligently without adding in that artificial intelligence to it.
Pete: So, what are some of the kinds of search engines you’re trying to string together to make this stuff work?
Ted: Currently, we work within the…I work within the SharePoint environment, and so, we’re focused at this point in time on SharePoint. In the past, we’ve had other search engines, and, you know, each one has had its shortcomings. This kinda goes back to the whole idea that we don’t staff it, we just plug it in. In the last 10 years, we’ve had, I think, four different search engines as our primary enterprise search.
And, now we’re at a point where they’re asking questions about our existing search engine. Do we need to be looking for another? So, we’re all talking to vendors again. And, I look at it and go, you know, “If you’ve had four different search engines, they’re all good products. And if they’re still not serving up the kind of results that you expect, maybe we need to be spending some money and focus on the backend of staffing that and, you know, doing the rankings, and the query rules, and all the other…and the content tagging and enrichment that needs to happen on the backend and not just think we can just plug it in and make it happen.”
You know, I was telling a story. I was on a project, it’s been a few years ago now, we were looking at moving from one search engine to another. And, we had, and I hate to, you know, name drop, but we had Google Search appliance, and we were gonna swap that out for SharePoint search. And during that process, one of the segments of our organization came to us and said, “We need new search results. And so, you know, we’re gonna stop using Google, we’re gonna start using your tool.” And, so we set that up.
But when we finally set it up, the requirements that came back to us were, “We want the exact same results we were getting with GSA to be delivered to us from your tool as well.” And they literally had people sitting there looking at search results from both search engines and making sure that they were getting the same results that they had previously got. And I’m sitting there thinking to myself, “You weren’t satisfied with the results you were getting using that tool. So, we unplug it, and we plug a new one in, but you want the same results.” Well, to me…
Pete: I’m trying to track, man. I’m trying to track.
Ted: You are trying to track this. You know, the old adage. What is the definition of insanity? Doing the same thing over, and over, and over again expecting different results. But in this case, they wanted the same results, but somehow they wanted it different. I don’t know.
Pete: Doc, it hurts when I do this. You know what I mean? Stop doing that.
Ted: Yes, yes. And it’s, like, there was no discussion about, you know, providing faceted search to this notice guide. There was no metadata being applied to the backend of that. It was just…
Pete: No discussion, changing the user experience for the better. Just give me the same thing.
Ted: None. It was simply unplugging GSA, which, as you know, now is defunct. But it was basically a box plugged into a rack, unplugging that, and then switching it over to SharePoint search, which operates quite a bit differently. We could have provided a whole different new experience and given it a try, but now they were comparing one page with another page and demanding the same results, though the results were never satisfactory.
Pete: I was gonna pivot to a question about intelligent search. And when I thought about that, it meant something different than it does now. I’m gonna try to stick with it. Let’s talk a little bit about what intelligent search means to you in your work over there. What are you doing to enable a more intelligent search?
Ted: I think that you need to, again, you need to get back to the basics before you can expect anything intelligent. I mean, there’s a lot of tools out there that can help us in our ability to make search a little more personal and practical. Microsoft’s doing some amazing things, of course, with the whole graph. We’ve got a lot of vendors who are promising that they’ve got tools that can understand, you know, simple language and that kinda thing.
I’m of the opinion it’s still, again, gonna go back to, are we structuring content? Are we curating that content? Are we looking for…you know, how do we determine what of that content is actually, you know, the truth, a body of truth, as opposed to just general trash that’s been stored out there in our environment? So, I mean, we’ve got thousands of sites that we’re out here crawling.
And, when you do a search for any given thing, you’re going to get, you know, hundreds of thousands of results back. And, how do we get through to what of those results is important and what isn’t? And so, I think, from that standpoint, I’m just not personally ready to move. I know that we need to stay abreast of all the technologies that are available to us, but I don’t think, for the most part, in most of our organizations, technology is the problem.
I think that the problem comes back to people in preparing and determining what’s important and what isn’t. I sat in on a seminar at the Search & Discovery Conference last fall over in DC, and I heard someone speak from a very large Fortune 100 company. And he talked about his staff and what he does to lead the search aspect of their company.
And, one of the things I thought was most intriguing of all the things he proposed was he said he was putting together a scorecard so that they could rate the sites that they are crawling. And if the site does not measure up, if the people who own that site aren’t doing enough to tag their content, to structure their content, if they’re not taking some responsibility for it, they stop crawling it.
And, if you want your site crawled, you need to, you know, bring it back up to speed. And, I found that to be just intriguing because we’re so…we just want to crawl everything, and everything people throw out there, no matter how it is. And then they get mad because they can’t find it. And, this puts some of the onus back on the end-user who’s creating that content. What are you doing to make your content findable?
Pete: Well, that, I think, is an incredible point. I feel like it takes me back to my very first day…my first job, right? Which was, you know, anytime I’m doing data entry, for example, and I think there’s something comes out of that experience that is now probably axiomatic. We’ll see that users don’t care about, you know, structured data on input, but they care very deeply about it on output. How do you meet them in the middle? How do you meet your teams at a place where you can change their habit and their belief about the importance of well-structured data on input when they need to care very deeply about it?
Ted: And that’s where I think the scorecard concept comes in because you have a tool then that comes back and says, “Well, you’re not gonna find your stuff because you’re not measuring up. You’re just not doing the work that needs to be done.” You know, we talked here about…we’ve got a group of people that are creating all this content in one tool, and they want us to crawl it. And somehow we’re supposed to go through all of the content that’s in that repository and sort through it and find just the finished pieces, you know.
And I’m saying to them… I mean, basically what they’re doing is they’re putting everything into a kinda like a file folder. So, you might have pictures, you might have some text, and you might have numerous pieces, and they all go into that folder and then out of that comes a finished product, some kind of a brochure, or flyer, or handout, that kind of thing, right? And so, they are throwing all that in there, and we’re having to think through and go through and spend all the money to try to figure out, how do we make that finished piece findable out of all that stuff.
And, you know, my suggestion has been, why don’t you just take the extra minute or two and just publish it over to a gallery and tag it. And, someone said, “Well, we can’t add to the business process there.” And I’m like, “You’re spending 20 hours, 30 hours creating a brochure, doing all the creative work. What’s five more minutes at the end of this when all approvals are done to say, ‘Copy this over to a gallery,’ so that we can simply search that gallery and not have to spend all that money and all that energy sorting through all that other stuff that we don’t even wanna crawl.”
My view here is that the simpler is usually better. If you can simplify it…and by simplifying it I’m saying, we can create a gallery over here, you just post it there, you tag it, or we can auto-tag it for that matter. We can go in and write some rules for that specific body of content because now we know what that body is, and that becomes your official. Why are we crawling multiple repositories over here with a bunch of stuff in it and trying to filter through that stuff?
That takes a lot of thought, a lot of money, and a lot of time building huge indexes that we then have to somehow parse down. Where we…you know, simple truth is we could just move it to a gallery. And so often…and here’s the other thing. I’ve had talks with our team in IT here about boiling it down to what we’re really trying to accomplish. My wife works for another company, another insurance company.
And, she would tell me the story how back, you know, 15, 20 years ago, when a policy came through the company, it wasn’t a very big one, and they would tie a balloon, a purple balloon, to the policy so that they could track it through the offices. I’m serious. I am serious. And, you could tell. You know, if you come in in the morning and the balloon was at your cube, you have work to do.
All right? And, I’ve talked to my team about the fact that in IT basically what we’re doing is we’re trying to replicate that balloon. You know, we’re finding ways to track the work as it goes through the system, and we’re trying to make sure that the workflows are happening and that content is findable. You know, that balloon helped identify where that thing was at at that moment, that that content is findable.
And sometimes, you know, that balloon sounds ridiculously kind of funny and simple, but it worked. It’s just that it doesn’t scale. And, I think we can accomplish that without having to get so complicated that we spend hundreds of hours. And literally, you know as well as I do, in projects, it can take that long, hundreds of hours to figure out how to stable that balloon to that policy. A simple gallery would do that because there is one place I can go to find it.
Pete: Well, it starts to solve this problem of dispersed information, right, of making sure that the information that they need that’s stored across separate tools is available when they need it, right?
Ted: And that goes back to as well. I’ve had conversations with a number of people who think that we have to have one place that we can go to, one portal, just log into in the morning that’s gonna have all the information we need. We’re going to be able to find anything we want in a moment from that one location.
Pete: It’s called Unicorn 1.0.
Ted: Yeah. You know as well as I do that’s not our world outside of work. If I need a restaurant, I grab my phone, and I bring up Yelp, and I look for a restaurant in my location, right? If I’m looking to buy a new computer, I might do a Google search real quick. But 9 times out of 10, it’s gonna lead me to BestBuy site, or Walmart site, or Amazon site. And then from there, I’m gonna use their search tools to find the stuff based on the facets that they’ve created.
So, there’s notable…you know, this idea… I like this idea of search-based applications, meaning that I have certain tools that take me to certain types or bodies of content. I should have an HR application where I can go find out about the time-off policy in my company. But if I’m confusing time-off policy with an auto policy, which we sell, then it makes it much more difficult. So, again, I think that abandoning this idea that one place is gonna provide all the answers kinda like “Star Trek” where…what was the episode where Scotty’s talking into a mouse and computer?
Pete: “Hello computer? Keyboard. How quaint?”
Ted: Yeah. How quaint. Exactly. Exactly. But, you know, vendors don’t like to hear that because they want to sell us these products that we can just plug in, and it’s gonna be clairvoyance and understand the needs and desires of every one of our associates as we struggle through mountains, and mountains, and mountains of content trying to figure out what we’re looking for. And I’m of the opinion, we changed 5 in just over 10 years. We’d be changing another one in another two or three years.
Pete: This has been a fantastically pragmatic approach here, and I certainly appreciate your insights. Fill in the blanks for me as we wrap up. If we can use this to summarize, fill in this blank, and tell me why this matters. Machines will ultimately…?
Ted: Rule the world.
Pete: It’s all over, everybody. Put a fork in us. We’re done.
Ted: Machines will ultimately come up short.
Pete: Why is that? When you have these vendors and you have ultimately other organizations who are incredibly bullish on the ways machines can help us to automate, how do you square that?
Ted: Because there are too many nuances for algorithms to sort through. People are people. People have… I can’t read your mind, you can’t read mine, and thus machine certainly can’t. We all work differently. We all process information differently. I think machines are just going to come up short. You know, I’m very impressed with stuff I’ve seen, the self-driving vehicles for instance and that kinda thing, but I’m also very skeptical because I work with computers all the time and I know this about computers. You have to reboot them a lot, and they’re not that complex.
Pete: Ultimately, what we’ve uncovered here is Ted Williams, once legacy of the Red Sox, is now a beacon of the robot uprising. This has been a fantastic conversation, Ted. And, also a terrific reminder of a refrain we have talked about on the show again and again, staff your search, people. Staff your search.
Ted: Yes. Absolutely.
Pete: Thank you for carrying that torch for us Nationwide’s Ted Williams. We sure appreciate your time here.
Ted: You’re welcome. Thank you.
Pete: And thank you, everybody, for downloading and listening to this show. We hope you appreciate the conversation we’ve had here with Ted. For more of just these kinds of interviews and conversations, head over to bainsight.com. You can go over to our Resources tab and check out the podcast, or you can subscribe to us for free anywhere finer podcasts are served. Until next time, I’m Pete Wright. And on behalf of Ted Williams, we’ll catch you right here on “Shared Insights,” the podcast from BA Insight.