Ann All spoke with Theresa Regli, a principal with CMS Watch, which evaluates content-oriented technologies and publishes head-to-head comparative reviews of leading solutions.
All: Though we are seeing more software-as-a-service delivery models for virtually every type of enterprise application, search appears to be a notable exception. CMS Watch research shows that one reason may be the popularity of search appliances, like Google's. Yet Google has a hosted search product as well. Are appliances and hosted search likely to remain "either/or" options, or will the popularity of both grow?
Regli: I think generally both fit similar scenarios - which is people who don't want to have a lot of maintenance issues with search - a smaller business or even a department within a larger business. Somebody might buy an appliance to work with a fairly small set of data within a large company. The appliances have cut into the SaaS model because they're sold to be a low maintenance, low-cost search product with a monthly fee that tends to be less than the SaaS model. It's very easy. My caution is that it doesn't work well in complex environments where there is a lot of content.
All: Can companies get more robust functionality with SaaS vs. a search appliance?
Regli: It depends on who is offering it. It varies from company to company. There are so many different options when it comes to SaaS. You can get quite a bit more with SaaS in a lot of scenarios, because you can pay for more services and more people who might modify or tweak the implementation to be more along the lines of what you might want. Whereas with most appliances, especially the Google Appliance, they don't have a lot of options to customize.
As part of our research, we do a lot of customer interviews. So we talk to a lot of people who have already worked with the appliance. We find that for the SMB or the departmental implementations, something like the Google Search appliance is pretty basic, pretty straightforward and works well in those scenarios. But for people who try to extend it and use it for multiple types of documents, where they are dealing with many different formats, or for people with complicated security scenarios, that's where it gets difficult to use it. I was speaking to someone last week at a conference who was trying to use the Google Search Appliance to secure or manage access to various documents that are quite secure. The appliance is limiting him as far as allowing access for certain search results only to certain people. Neither SaaS nor security appliances fit very well if you have stringent security requirements. You need more complex technology to be able to manage that.
All: So is security becoming a bigger concern?
Regli: Definitely. There have been too many incidents where documents have gotten into a certain person's hands when they shouldn't have. At the last few conferences where I have presented, most of the questions have been about security. People have been asking, "Can I do this on a repository basis? Can I do it on a document basis?" I think the vendors are really scrambling, because they don't have good answers right now. Big enterprise search players like Autonomy, Endeca and even some of the infrastructure companies offering search tools like SAP and Oracle, have a lot more experience dealing with the issues of secure enterprise information and those kinds of complications. They have more answers to security-type questions. Of course, you pay a lot more for that.
All: Are security concerns limiting the potential for search delivered via SaaS? What about appliances?
Regli: I think there is a place for hosted Web search. If you have a Web site, and don't want to deal with the search engine on that site, hosting it makes a lot of sense. The content is already public, it's already out there, and it makes sense to go outside your firewall and index it. If someone searches for it, boom, it's there, and it's off your plate. When you go to sites that say "search powered by Google," that tends to be a hosted solution. Or they might have the appliance in-house and it's just indexing Web sites.
But when you get into enterprise search scenarios where you have dozens of different repositories and strong security requirements, that's a scenario where you don't want to have an index of that information living somewhere outside your firewall. The larger enterprise search vendors are doing a better job of coming up with solutions that tie your search results into your existing security requirements. They have a compelling argument: "You're already using our database or CRM system. You might as well use our search." That's where I am seeing a line being drawn.
I am curious to see how far the appliances will go down another path. They might just want to fit the existing scenario. Why would the Google Appliance want to be Autonomy? The problem is, some consumers think that it is. There's corporate pressure from above. You've got the CIO maybe saying, "I want this to work like Google. So why don't we just buy Google?" The reputation is probably getting them more business than they deserve, at a scale they aren't yet ready to handle. I think there are people in middle management who realize the limitations of the technology. There are a lot of ways to execute search, and different search needs; one size doesn't fit all. That's the challenge that enterprises are facing.
All: It sounds like companies want the ability to search lots of different types of information sources. Is that another limitation for both SaaS delivery models and search appliances?
Regli: The Google Appliance is very good at doing Web-based searches. But a lot of times in enterprises, documents aren't on the Web or even on an intranet, they are sitting on a hard drive somewhere, or in a database somewhere. It's not as easy to index certain formats when the application is optimized for Web content. If you've got a lot of documents in PDFs sitting on Joe's hard drive or sitting in the HR department, it's very difficult to break through that.
All: That said, do you think companies are trying to move toward getting more of their data online? Wouldn't this make it easier for people to find it?
Regli: A lot of projects we see happening involve not only optimizing search technology but also cleaning up the content - figuring out where it is, what format it's in, can the search engine actually work with that format, and if not, can we convert it or should we just throw it out? That's not a technology issue; it's a business process and management issue. I get calls and people say, "I need you to help me select a search engine." But I'll say, "That's six months off. You have to look out your content first, figure out what you've got and where it is."
All: In addition to the ability to search different types of information sources, what else are companies seeking from search solutions?
Regli: From a user experience perspective, the idea of being able to cluster results is now considered a de facto standard. If I execute a search on a certain topic, I should be able to browse through subtopics and be able to get a more granular search. The idea is you might search a certain topic, but not know exactly what you are looking for. You don't what kinds of documents will come up. You might say, "I want a good restaurant in New York," but you wouldn't know the name of the restaurant. So the idea of being able to group documents into certain categories is a strong requirement. Even Google is starting to offer that now, which is kind of shocking when you think about it. The Google paradigm is you expect to be able to type in a search term and get a list of results with the most popular results on top. But in the enterprise context, popularity doesn't necessarily work. The most important document or the most "correct" document isn't necessarily linked to often.
All: Another interesting finding of your research was that companies didn't seem all that interested in so-called Web 2.0 search capabilities such as social results rankings. Are the value of such results too hard to quantify?
Regli: Vendors are pushing it. I think it's ahead of the consumers. The consumers are annoyed that they can't get the results they want now. They are still trying to solve the search fundamentals, where the vendor marketing is saying, "You can have people tag results so you can find more things you might be interested in." That's fine, but people just want the right results. We're seeing that as a challenge. It sounds good from a marketing perspective. But I don't think consumers are ready for it, because they are still trying to solve basic search problems. I think that so much of what is hip from a marketing perspective, what the kids are doing on MySpace, rarely works from an enterprise perspective. I see people getting all excited about it, but how many people going to work every day have time to spend doing stuff on the search engine if they aren't actually working on it? It's not something people are going to spend a lot of time doing.
That said, there is value to people ranking whether a search result is good or bad; that's helpful. They can weight the value of the search results and put that back in the index so the next time someone executes that same query, then great. But whether they'll have a discussion thread around the results, I question that. That may not be as valuable as the marketers would have you believe.