There is much that is technical about search; how you measure relevance and improve it, how you do the indexing, how you process the queries, how you remove duplicates from the found, how you deal with spelling errors, how you put together the found and present it.
This post isn’t about any of these things. There are a lot of places where those dialogues are taking place already.
I am more interested in what people search for, why it differs from past paradigms of requesting information, what that means and why we can obtain immense value as a result.
My way or the highway
First and foremost, people are now getting used to asking for things the way they want to ask for things. Not the way the “system” expects, but the way consumers expect.
And guess what? No error codes or warnings when you do the asking. No “Invalid code entered” or garbage like that. It’s not validation but parsing and transformation.
I know it when I see it
Many years ago I read a book on quality by David Guaspari. A decade and a half later, I still enjoy the fable. Customers know what they want when they see it. This process, of discovering requirement by delivery, is to me nothing more than a variant of fast iteration. Al-Noor Ramji, who hired and mentored me in my early years, used to say “Don’t ask them what they want, they don’t necessarily know. Show them a Ford Escort and ask them what’s wrong with it. And keep improving it. Quickly”. Or words to that effect.
The shift from deterministic returning of “founds” to probabilistic and relevance-ranked is not trivial. And it will take time for our generation to move off the accuracy high horse, the digital yes-no approach. But tomorrow’s generation are used to relevance and ranking as a way of life.
Size doesn’t matter
We search for many things. Sometimes it is a precise answer, sometimes we’re browsing within a general area, sometimes it’s a number, sometimes it’s a book or article. But unlike the way we pull information down from today’s systems, the answer sizes vary enormously. And we have to cater for that, people want little things and big things, a small amount and a lot.
I remember many years ago when working with dumb terminals at Burroughs, I was amazed to see how much perception ruled customer satisfaction. Given a choice between painting the screen with one big chunk of information or doing it in little bits at a time, the technical guy would always opt for buffering it up and doing the one big chunk. Why, because it was faster overall. But the customer preferred to see some action early, even if the transaction took longer. Why, because it felt faster. Fast is as important as relevant.
There are so many other things to consider. How to learn from the Firefly and Amazon and StumbleUpon experiences in collaborative-filtering-meets-deep-preferences. How to use the clickstreams and patterns to aid new hires, provide people with a readymade “this is how she does it”.
How to savour the richness of diversity between two people apparently doing the same job, but doing it very differently. Celebrating that diversity and learning from it rather than grinding everyone into lowest-common-denominator submission. Learning how to use tags and feedback loops to transcend language and cultural barriers.