Google recently announced a new capability: Springboard. Some call it an “AI tool,” some call it the replacement for Google Search Appliance.
I disagree. Google is finally confessing that context is important for enterprise search.
And all I can say is “Finally!”
Springboard is targeted at those organizations that use Google Apps for Work. It provides the ability to find your content anywhere in the Google Apps for Work universe.
Because it’s only currently available to members of the Google Early Adopter Program, we don't know if it only searches your documents, or if it searches a larger set of corporate content you’re authorized to see. My guess is it’s the former: only content you create. Security and privacy issues could be difficult to handle until computers are much smarter.
Springboard is rumored to include a nice capability: the ability to connect the dots between data in your different Google apps. For example, an email that mentions a lunch meeting with a colleague can generate a reminder email a few minutes before the meeting. This could help on multiple fronts: Google can generate a message reminding me of a cross-town meeting, and also notify that traffic is heavy so I should plan on leaving early.
But Google (and Microsoft Office 365) already do that on my iPhone without any special subscriptions. In fact, public Google search uses my previous queries to suggest other sites, much like popular shopping sites can tell me “People like you” have purchased some product. Google even shows me fares for travel I’ve searched for recently.
Google Swallows Its Own Medicine
Google employs a lot of really smart people, and they are finally following their own advice: search is everywhere, so use all of the information available to deliver results.
We've advised clients to use context for years now. Search is no longer keywords and 10 blue links.
Context is relatively easy to understand in the enterprise. At a high level, three unique classes of context are available to enterprise search: the query, the user and the content. These interact, of course, and returning great results usually involves all three.
Let’s look at the these three types of context that organizations can implement in their search platforms.
3 Classes of Context for Enterprise Search
Most search platforms use only the search terms — the query — to bring back results generally based on how common the word is in documents, in the entire collation and in the stemming and any synonyms you may have defined.
It turns out there is much more context in the search term. Other clues for relevance ranking, known as "signals," include the document language, capitalization and punctuation. A really clever search could determine whether the query is a product name, a company division, or even an employee name and search accordingly.
With all of this extra context, results would be that much better.
You also have control over how to display results based on the query. If the query is the name of a company employee, why not display that person’s name, email and phone number at the top of the list. If it’s a product name, show the product description and link at the top of the result list.
In the enterprise, you also probably have context about the person submitting the query. Chances are he or she is logged in, and from that you know physical location, job title, previous queries and co-workers. You also may have a security level for the user, to determine which content can be displayed — search platforms are finally getting good at using this context in queries.
A user from Germany, logged into the company portal in the Germany, probably prefers German content. An employee visiting from the US will likely search in English and would likely expect results in English as well.
Query and document viewing history for each user lets you know what documents previous search results included and which documents the person viewed. You know what people "like" the user search for and view.
Enterprise search platforms return documents, but what is a document? Often it is a web page or a Word document. Sometimes it is a spreadsheet, or a database record. And within the enterprise, it most certainly has a security level. And each document has content in a particular language; and often with its own user-generated "properties" — think in Word, for example, the author can provide Titles, Dates and Author names.
Sadly, most enterprise search platforms don’t use context automatically, so you’re left to implement it on your own. While commercial and open source search platforms are improving, much of the success of a particular search implementation depends on how much effort goes into the initial installation and into ongoing maintenance.
In a recent survey of enterprise search users, Findwise reported that a large number of companies called enterprise search “important or critical” to the success of the organization. Sadly, a much larger percentage reported that "one or fewer" full time employees managed search.
Organizations regularly replace their enterprise search platform every few years. If an enterprise was wiling to spend money managing and improving enterprise search, including hooking in the contexts mentioned above, they could save a lot of money licensing and implementing the new search platforms.
Use the context — save some time and money, and deliver search your users will love.