Real Story Group. Make Better Technology Decisions.

Adriaan Bloem

A tale of two search technology selections

18-Mar-2008

Tags: Enterprise Search, Marketplace at Large, Selecting Technology, , Google Search Appliance, IDOL Server, Vivisimo Velocity

CM Pros organized an event on the topic of enterprise search last week, with two case-study presentations and a fair bit of discussion among attendees. The key takeaways: website search is a lot harder than you'd think, and the attitude of a vendor really matters to your success.

When talking enterprise search, the "multi-repository enterprise" scenario is where all the really complex technology and fancy infrastructures reside today. We shouldn't forget, however, that website search -- powering that tiny form field in the right-hand top corner of your site -- remains one of the most common scenarios. Listening to the experiences of a large membership association was a clear reminder that it's certainly not as simple and straightforward to get right as many would think. (We can sympathize -- trying to get the cmswatch.com search right is an on-going struggle for us.)

The first case was presented someone who works for an integrator involved in the project. He is architect for their CMS and was therefore also put in charge of website search (which is a fairly typical arrangement, as the level of expert knowledge needed for website search is usually underestimated). He explained how it started out the same way many of these projects do -- "why don't we just get Google?" -- and how the organization decided to embark on that journey without much further discussion.

This led to exciting tales of a chase to find the lost Appliance somewhere in the server center when the "try-before-you-buy" clock was already ticking. It then turned out the Google Search Appliance (GSA) didn't actually match the original, rather straightforward 2-page requirements specification. Nor was it really as cheap as originally envisioned (one box wasn't enough). And, to boot, whereas the association imagined itself to carry some major clout with the vendor (their site is a very popular), it turned out to be nigh impossible to get Google to do customization. Before the web search goes live, they will need to bring in outside help to get some things ironed out on top of the Appliance.

Though it would be easy to blame Google for what went wrong, the presenter gave the company credit for delivering good support and a well-performing product. To the audience it seemed he handled it well, the product in itself was fine, but the main mistake the customer made was in underestimating what they were about to implement in the first place. As I overheard someone whisper next to me, "and to think this was a relatively simple site search project."

The second presentation was an account of the selection and implementation at a pharmaceutical company. They had twice before attempted to implement search functionality across multiple disparate repositories, and both times the project was aborted. This time, however, the project was officially declared a "success." As such, there was even an investigation of why the third attempt, almost against all odds, finally turned out right.

We were taken on a text-book product selection tour, but what stood out was there was definitely no one-horse race here. As someone who has uttered the caveat emptor "try against your own corpus of content" more than he cares to recount, I can certainly agree with the thorough proof-of-concept they put three vendors through. If CMS Watch ever awards a prize for heeding our advice, this POC might be nominated.

Though you should never underestimate the importance of how the technology functions, one of the main criteria was the attitude of vendors towards the project. Arrogance doesn't pay off in a well-executed test-drive, and in that respect it wasn't surprising the smallest vendor of the three scored points on cooperation and flexibility.

It was interesting to see the two opposites on one evening. One obviously struggled with what they expected to be a simple project, and wasn't much helped by their vendor. The other, on the other hand, braced itself for the worst but came out on top -- not in the least because they made sure the vendor they picked would be cooperative. So though it is easy to be cynical about the success rate of search projects, it can be done. Just heed the lessons learned, and come well prepared.