CIO 24/7 Podcast: Tapping into Accenture s rich content with a new search capability
CIO 24/7 Podcast: Tapping into Accenture s rich content with a new search capability Featuring Accenture managing directors Jason Warnke and Chris Crawford Jason Warnke: Welcome to the CIO 247 podcast bring you the latest updates from Accenture s internal IT organization. I am your host, Jason Warnke and today I have Accenture Managing Director, Chris Crawford with me to discuss Accenture s new and improved Enterprise Search. Thanks for joining me today, Chris. Chris Crawford: Hey, it s fun to be on this side of the microphone, Jason. Jason Warnke: Exactly. So, Chris, a key component in finding relevant information within a company of our size, 384,000 people, is the search capability. Can you explain what Accenture s search capability is to our audience and how Accenture people use it? Chris Crawford: Sure. We have a couple of major search capabilities. One it to search our working documents that we store in Office 365 and that s pretty much just box standard Microsoft SharePoint search. The search that I am responsible for is the broader search that we use across the enterprise and that s where the search that we use to search our two major, two major corpuses, the internal content that you need to be an employee, so things like policies or websites or announcements or, you know, category pages, group pages and then the information that you need to better serve your clients so that includes our communities, our, what we call the knowledge exchange, which are our reusable documents and ideas, white papers, deliverables, reusable deliverables but I think the important thing here is we keep those two types of searches separate. The search for your working papers and then the search for the knowledge that spans the entire company. And we are talking about the second one today. Jason Warnke: Very good. So, tell me a little bit about the main differences between what we used to use in the previous version of Enterprise Search and this new version that has recently rolled out at Accenture. Chris Crawford: Ok, yeah. The search has always been important with, with, you know, hundreds of thousands of documents of reusable documents to search through to find your best stuff, there is a common complaint that people knew that information was out there, they knew the experts were there, they knew the content was there but they weren t always able to find it effectively. And we had used a couple of different technologies over the years and although those things continue to get better and we saw small incremental improvements, you know, as we worked with content owners and working with the products themselves, we still needed a step change and that s the challenge that we had. I had kind of lobbied against these fixed search programs for a while just because I thought that they would bring an incremental improvement for a lot of money and then slowly results would degrade again over time as you weren t paying attention to them. So, a couple years back, we started a Search Center of Excellence that really focuses on the ongoing improvement and the processes of search, not just the technology of search. That program had been making headway and was doing a good job working with content owners and improving the results but we were stymied because we weren t always able to get the information that we wanted about why certain results were popping to the top or how that was happening and we were also frustrated with the, with the rate that we could test and implement changes into those search algorithms. You know, search isn t about finding stuff, it is about sorting stuff. It is about putting that most relevant document that matches your key terms to the top and our ability to do that was just kind of limited as more and more of these search engines wanted to go black box on us rather than on giving us the information we needed to make our search better.
Jason Warnke: Wow, that s very interesting. What are some of the new features in Search? You ve mentioned some of these to me in the past so a new crawler, there is a metadata management layer, the search engine itself, what would you say are some of the key new features? Chris Crawford: Yeah, I mean there s obviously, there are a lot of things that make up any given search implementation. But the big things that we did in this one was we looked at an open source search engine called Elastic Search, which is very popular and we had good experience in using that as, for application level search. But using it for enterprise search where we've got to bring together results of different types into a coherent whole. So, we wanted to bring in blog results and web pages results and PowerPoint documents all into the same search results if you are searching for retail thought leadership for example. There might be a nice collection of things that would appear there. Elastic is really well tuned if you are searching the same type of things, or it is easy to tune if you are searching the same type of things. What we did was added another layer into this, into the processing pipeline there that helped us normalize and enrich the data. That s really the, I think, what's unique about this implementation, was after we crawl the documents, crawling them into a dirty index, we go through that normalization and enrichment process that allows us to add information across the data sources and normalize some of the scoring and some of the fields that we index into what we call the clean index. And then from there, it is just a matter of we re almost back into the normal pipeline again, where we customize the search results. We added some additional features that people have come to expect in modern search engines like infocards. Also, giving you that quick information, for example, if I type somebody s name I see a quick infocard about them without having to go to the complete results and other things like autocomplete and suggested alternative searches and things like that, that were just a little more challenging for us to do in the old engines. Jason Warnke: That s great. I mean, it sounds like we now have a capability that allows us to open up the hood and, you know, to tune the engine the way we want it to work and in the past we had some black boxes that didn t allow us to do that and it, we have much more control and tunability which is really cool. What were some of the challenges faced when coming up with the processes to improve Search? You mentioned some of the process aspects of this earlier but maybe expand on that just a little bit. Chris Crawford: Sure. Understanding the overall process of Search, you ve got to understand, you ve got to come up with the metrics and the measurements as to how your users, whether your users are satisfied or not with the results that they are getting. So, having a way of gauging user satisfaction is key. Having a way of scoring your search results is key as well. You can do that with a couple of different techniques. We use, we use different analytic reports that we've developed over the years to identify searches that people are using commonly that aren t giving them the results and then we use other, we use other measurements like searches where people don t click on any of the results and we look for the common terms there. We have the ability then to sort some of those out, look for some of the most common terms where we think the user satisfaction is very low and then we can use that either to fix them, if it s a particular point problem or we can look for trends, more importantly, we look for trends about why people, what people are searching for and are they not finding it, is it a gap in our, in the content we are providing or is it just a simple matter of not getting that content to appear in the searches as well. The other thing we look at too is new search terms. We look at search terms that are rising, that are increasing in popularity. So, for example, when Accenture implemented a activity, a fitness tracker program, we saw a number of searches for the word Fitbit but we weren t, but that, the word Fitbit didn t appear anywhere on the website so people were frustrated when they would search for Fitbit and they weren t getting any results. Well, we knew what they wanted was the Accenture Active program and so we had the ability to force that into the results. We added some metadata and
now when people search for Fitbit, they find the Accenture Active program even though the word Fitbit doesn t appear on the site. Jason Warnke: Very interesting. It is a comprehensive look at or a reimaging way search is done. It is not just the implemented tool but there are a significant, there s significant aspects of the process change mindset that is really important to be committed to if you're going to, getting the results that we re getting which is sort of the next question is, you know, we ve been working on this for, for I guess ten months now and have recently rolled it out and been getting very good feedback. Tell me a little bit about that feedback. What have people been saying? Chris Crawford: Yeah, the anecdotal feedback is great. We implemented feedback options right on the site so we can always capture the immediate questions that people have and we get some good suggestions there too. What we find is that people certainly expect things like quoted phrases and pluses and minuses in their search terms to work like they do on the popular search engine and that is something that we underestimated going in and something that we are resolving. But what we are, the feedback is overwhelmingly positive. People are delighted with the speed of the results. Things are coming in, in well under a second whereas the old results were coming in at about two or three seconds. And it is just that extra responsiveness, just makes the, improves the use perception of quality all around. They love the infocard. They loved seeing the things as they type. They loved the best bets, the things that we can force to the top when we know that there is a definitive answer that they are most likely looking for and what we've seen in the measures of our portal, the satisfaction measures of our portal and our KX have both gone up about 20-25 percent in the past three months and those are two things that are highly dependent on search, on the quality of search in there and so we are really pleased with that. We are also seeing the number of abandoned searches where people aren t clicking on any of the results, were really, we've seen that one drop from, you know, somewhere in the upper 40s to the low 30s so we are really, we are really pleased so far with the results but I think we are even more excited about the capabilities that the platform gives us to continue to improve Search not just as a one-time fix but as something that we've implemented the processes for going forward. Jason Warnke: Excellent. Very, very good story and as a user of Search myself, I can say that I have been seeing those positive results in my experience as well. So, very good. I look forward to the continuing journey because it really is that. Not something you close the book on, but rather it s an ongoing journey that we ll continue to tune and make sure it gets better and better which is great. Well, thank you for listening to another CIO 247 podcast. I am Jason Warnke and with me has been Accenture Managing Director, Chris Crawford, to discuss Accenture s new and improved Enterprise Search. Be sure to check out previous episodes on Accenture.com. # # # #
Copyright 2017 Accenture All rights reserved. Accenture, its logo, and High Performance Delivered are trademarks of Accenture.