Big Data practitioners have a huge problem on their hands: a consistent failure to build solutions for actual users. Users don’t care about analytics pipelines. They don’t know what data science is. Or the fact that you can talk to half-a-dozen Big Data storage systems using something-like SQL. They don’t care about interesting ways to drag-and-drop analytics pipelines in place in a user interface. Sue in marketing certainly likes her user metrics. Tom in finance certainly loves his dashboard showing him where the money is going. I bet Ed in IT likes watching every packet flow he can get his hands on.
“A common pattern in successful relevance practices is a deep culture of collaboration between domain or business experts and the search developers”
But real users that pay the bills don’t sit down the hall from us like Ed, Tom, and Sue. They are a different crowd entirely. Big Data to our users is the product catalog they’re using to shop for a gift, the extensive medical knowledge base being used to save a patient’s life, the catalog of technical books being combed for answers.
These users don’t care about dashboards, SQL, or analytics. They want to know one thing when they find your app: where’s the search box? If its not fine-tuned to help them find what they’re looking for, they’ll leave. So while Tom is obsessing over his pretty financial dashboards backed by best-inbreed analytics technology, users are silently leaving. They’re frustrated by irrelevant search results, switching to the competition that seems to have intelligent search that “gets them” and what they want.
So many companies fail to focus their Big Data efforts on search. Or if they do, they don’t quite understand the engineering discipline of improving the quality of search results or search relevance. The search engine seems like an inscrutable, mystical black box. Yet if your job is to bring users to content, products, restaurants, (really any “thing”), then search is at the core of your value proposition. Relevant search is the user experience. Helping users find what matters is the most important Big Data task you could be engaged in. Not focusing on a relevant search experience for your data is like bringing your users to a giant pile of books and calling it a library.
Consider, as an example, an e-commerce search. We’ve had e-commerce search for decades. Amazon’s empire is in part due to its excellent search. Yet engineers continue to fine-tune search results. It’s precisely because it’s both challenging and lucrative. Why is that? E-Commerce search must do everything it can to replicate an in-store sales experience. The search engine effectively replaces the salesperson. The factors a salesperson would use to create a successful sale for the company and customer are expressed directly in the results ranking. Is this a product that sells well? Do other users rate it highly? Does the text match what the user is looking for? Do we understand what the customer is asking for? Is it in stock? More importantly for the company, is it a profitable sale? Does it satisfy our suppliers? Just as these factors play into the mind of a human salesperson, so do they factor into the ranking used by search engines.
Consider another application. A doctor saving a life needs to consult a medical knowledge base for answers. A completely different set of factors comes into play. To the doctor, the search engine represents an expert on standby ready to offer much needed advice. The search engine understands medical jargon. It’s programmed to understand medicine. It can relate connected medical concepts together semantically. It knows the up to date practices and research. It may even understand something about the doctor, they’re capabilities and skill. All of these factors -- these domain specific features of the content -- play together to create a ranking function specific to creating an expert user interface.
There is often no “one size fits all” here. Every search user experience differs mightily, including yours. What does a medical search look like that focuses on historical research done in a library instead of life-and-death situations at a patient's bedside? What changes in the business rules that dictate how data is retrieved? How does e-commerce change when the application is classified ads? Do we suddenly need to measure how likely it is the buyer can trust the seller? How can we even know what our users need from our search application?
Search is creeping up on you. Users want more and better search. When search doesn’t work, they’re not vocal. They simply leave, often without thought. As search’s slow, inexorable advance into our applications increase. what can your organization do to enhance your Big Data capabilities to meet the needs of real users?
Build or hire a search relevance center-for-excellence.
Hire expert search developers that understand the search engine inside-and-out -- particularly the art of relevance and the field of Information Retrieval. You need skilled developers that understand the subtleties of various ranking scenarios. And relevance isn’t just search engines, there are many Big Data and analytics technologies to extract signals and suss out true semantic meaning from text.
Focus your metrics on search.
Are you A/B testing search results? Do you know what “thrashing” and “pogo sticking” in search results are? Here the analytics instincts of the Big Data community can pay off, identifying broken searches and places where users struggle to find what they need. Users don’t leave informative notes on when they weren’t satisfied, they simply just leave.
A common pattern in successful relevance practices is a deep culture of collaboration between domain or business experts and the search developers. Relevance engineering is a social discipline, not a place where engineers tinker with the search engine in isolation. The developers won’t know what business rules should be driving the user experience without the help of experts in the domain, marketing, or business. Tear down artificial barriers between these groups to instill a workplace where interaction is commonplace not scheduled.
As you’re choosing how to apply your limited Big Data resources, ask yourself -- Should we be optimizing for internal users, or external users? For many organizations, a focus on relevant search nearly always helps the bottom line. If the value in your organization is helping users find content, products, restaurants, or anything else, then there’s a direct relationship with better search, user engagement, and monetary value. Are you prepared to take the steps needed to take your search to the next level?