Skip to content

Searching a haystack

Enterprise search needs to be simple to use; designing it is anything but
Searching a haystack

Search is one of those wicked problems. Not “wicked” in the English vernacular sense though. And it’s not the OED definition of the word of “evil or morally bad” either. More, as Roger Martin of the Rotman School defines them: as “messy and reactive, and with no single solution.”

Search has many moving parts. Building a search application really is a messy business, and it requires an obsessive attention to detail. “Simple, fast and relevant” do not come easily in a search project.

Google makes it look so easy though. In less than a second, I can bring back millions of results on any topic I choose — from learning to spatchcock a chicken to finding the funniest meme about the Brexit fiasco (a submarine of cheese anyone?!). Search is often the very first resort for many on the Interwebs. It has changed the way we find everything: products, knowledge, answers, services, people and places. It shapes the way we interact with organizations and the way we think about the world.

Search within the enterprise is slightly less profound than that but no less challenging to get right. It is often a source of endless irritation and the complaints of “I can’t find anything” echo down the halls of more than just our law firms. And while there are those who think everything is just fine and dandy with what they have, they may not know what riches await with a better search engine, better data, a better interface or using better search query language.

As I move into the final stages of development for our new enterprise search system, I have been looking back over the years at the eight search implementations with which I have been involved — from early Verity and Autonomy to SharePoint, BA Insight and internally built search solutions and from Recommind to new Insight Search from iManage. I am reminded just how mature the market has become and how important it is to get the basics right (find a document, fast) but also to support the deep topic research journeys that also happen at our firms. Moreover, the new tech now allows us to go deeper to create an environment where search is central to information-seeking and decision-making at our firms, where search can uncover insights in the data, make serendipitous connections between things and inform you of things that you didn’t know you needed to know.

For “Search: The Next Generation” to happen, however, we need to understand information- and knowledge-seeking behaviours. We need to design search screens that promote instant retrieval, while also supporting the deep and cerebral research of our lawyers. We need to design screens that can help us see the patterns and trends for even deeper insights about the firm and its expertise. Oh, and all those screens need to be simple, clean and intuitive. Wicked search.

Between them, Apple and Google have set the standards for simplicity by design. And, luckily for us, the search vendors in our market have sought to mimic them. They may not deliver the same slick user experience as Amazon or Google, but given the difference in resources, we should be impressed all the same.

Back in the mid-2000s, I remember spending hours learning the verity query language to more deeply research and audit a law firm’s document and knowledge collections. As an expert researcher using VQL, you could spend quite some time carefully crafting syntax-heavy search strings. You could use a combination of proximity operators, wildcards and Boolean to be super-specific about what you were looking for (eg. (shareholders AND agreement) AND (NOT Rubble) AND “Flintstones Financing”~5). It was oddly pleasing.

The vendors realized, however, that you could push all that search complexity into the back of the engine to present a much cleaner experience on the front. This led to those “advanced search” links to a page that presented a combination of fields for the user to complete: one for exact terms and one where you could list the terms to be ignored, as well as for phrase searching, stemming, etc. This presented a cleaner and simpler screen for users but with links to other pages for those on a deeper search adventure. It also means that many people today have only the vaguest sense of the term “Boolean.”

Other “advanced search” forms would present lawyers with specific fields they could use to search the content — things such as title, document type, date, client/matter and author fields. If you know that the word “shareholders” will be in the title of the document and you know that the author was Wilma Flintstone, you could retrieve a more precise set of results to pore over. Most of these advanced search forms have (sadly) disappeared as the power of the algorithms improved and the ability to filter search results came along.

Faceted search changed the way we find music and books online, shop for hardware or buy second-hand goods. Suddenly, it meant that you could stick any old word in the search box and then filter the search results down to a more precise selection by using, for example, topic, department or date fields. Enterprise search vendors quickly followed suit. But if you chat to any content management expert at a law firm, you will quickly learn about the pain this change in search caused. We had to develop new schemas and normalize taxonomies across collections; we had to clean up poorly tagged documents and hide the rest using vague descriptors.

I have been personally promised on more than one occasion over the past 15 years that I need not do any of that cleanup, that no-one needs to classify documents anymore. That there’s this one fully-automated-luxury-unicorn that can do all that for me, automagically. But I am yet to be impressed.

Law firms have a variety of data, information and knowledge — some in an unstructured mess, some in a semi-structured semi-mess and some in a highly structured and curated thing of beauty. The more structured the content, the more that search can leverage these tags to reveal more meaningful and richer knowledge — but only if search can merge the very best aspects of knowledge engineering already present at our firms. The taxonomies, ontologies, linked data and knowledge graphs all need to be part of this wicked problem we’re trying to solve called search.

Kate Simpson is national director of knowledge management at Bennett Jones LLP. Opinions expressed are her own.