The Thick Edge of Quora
In January 2011, I wrote a post, The Thin Wedge of Quora, in which I hypothesized that while on the surface Quora was a Q&A service, that it could, over time, use this interaction to draw out high-quality, structured content from its users in order to rebuild not only the way content is created online, but also how one searches for it.
For a bit of context, in that piece I wrote:
Over time, the content contributed to Quora will constantly be improved, refined, aggregated, and structured. Like a stone in a tumbler, the edges will get smoother. One effect of the tumbling and fine-tuning is that the site will become better optimized for search…And, this is where the other edge of the wedge, the thick edge— will come into play. Search has undergone tremendous change and will continue to do so. For Quora, [this may depend on] small segments of curious users who feed the system good questions; slightly bigger segments who contribute knowledge to the system in response; and hopefully an enormous segment that searches the Internet in a variety of ways and somehow end up on Quora for their answer.
Over the past few months, I’ve continued to talk to many smart people about Quora, and I’m often asked random questions about the site, as well. A few themes emerge from those discussions, mainly from people who are confused as to how Quora could legitimately be worth (in terms of venture valuation) north of $1 billion, if not more.
For instance, many people simply assume that it’s just a Q&A service, and that a young analyst can size that market and make approximations. Then they wonder about Yahoo! Answers, which collected a tremendous amount of pages views but weren’t able to convert that into cash. Some think Quora is a fad that will just go away as people lose interest in the medium. Others are weary of ad-model businesses in an age where its difficult to pinpoint just how users spend their time engaging with sites that aren’t named Facebook.
Finally, most people who scoff at a billion dollar valuation don’t really use the site and therefore can’t possibly begin to understand how this property is already adding tremendous value to the web. It’s simply not enough to look at Compete scores or Alexa rankings, unique visitors and other engagement metrics — much of the site’s value is currently hard to measure, and that’s why I wanted to sit down and explore more of what the “thick edge” of the wedge could be.
If you’re read this far, thank you, and bear with me as I try to convey my thoughts. I’m often asked this while meeting with people, and it’s hard to discuss conversationally, so here we go…
When Google was unveiled to the world, the timing couldn’t have been better. In those days, the web was growing unwieldy. Everyone had websites, the design was poor, the ads were intrusive, and it was hard to find relevant information. Google’s PageRank algorithm found a way to help us sort through the web via keyword search queries, dramatically reducing the time it took users to pinpoint relevant answers to their query. In return for this service, Google charged for keywords against these queries and, in the process, built a technology empire that is hyperactive and hyper-competitive in a range of industries.
Google search became (and still is) so dominant that it transformed search engine optimization (SEO) into a lucrative art. The PageRank algorithm compressed users’ search query times so much that it created an incentive for those managing websites to trick or game the algorithm. Web pages began to get tagged with keywords, or littered with keywords in the text itself. The game was (and still is) to show up on the first page of Google results. That is the prize, and folks who figured out how to do that made (and still make) a good chunk of change.
The PageRank gold rush also created a perverse incentive for content sites to build themselves around popular keyword searches and try to appear on the first page of Google results. Websites such as About.com and eHow, for instance, were successful content “farms,” where the site owners could justify paying for keyword-littered, highly-structured content written by faceless writers in exchange for the ad-revenue it received by luring users to their site for answers. The content farming business proved lucrative in a world where people instinctively went to Google to get their questions answered.
Lately, the efficacy of the PageRank algorithm has slowly come to into question. This topic is its own separate debate, but briefly — people began to question whether not only PageRank could be gamed at the expense of the user, but also that certain sites could be left out entirely altogether based on the corporate whims and interests of Google. This debate opened a dialog among many who cover the web closely. For instance, while a Google search query still remains the fastest, most efficient way to find information online, the question remains as to whether (1) the content that Google surfaces is authentic, relevant, and trustworthy, and (2) that once a user generates a Google search, that he/she will be able to pinpoint the answer to their question in a short period of time.
This is where Quora comes in.
The design and embedded controls within Quora help address parts of these two questions: (1) content on Quora is written by verified individuals who are monitored by machines and human reviewers to contribute this content with proper grammar, within a structured and specific question, and to be tagged and set in context using topics and sub-topics; and (2) the quality of content on Quora is ranked largely based on the votes of users monitoring the thread, using a PeopleRank algorithm to assign voting weights for specific users who have demonstrated knowledge in certain topics. In many cases, these controls help the most relevant answers surface to the top of the thread.
Mechanically, I’ll briefly lay out what happens in the formation of a new question thread on Quora to illustrate:
A user has a question and wants an answer. The user types in the question into Quora. If a similar question exists, the user can visit that particular thread or elect to create a new question. The machines make sure the new question meets the requirements (e.g. no “polling” question, etc.) and encourages the question-writer to set the query in context by assigning tags. While the machines work, a cadre of human reviewers and “Quora Admins” try to make sure the grammar of the question is correct, that there is relevant context, and that the topics tagged on the thread are appropriate. The question is then distributed to the news feeds of users who follow the person who asked the question as well as to people who follow the topics the question has been tagged with. In these feeds, other users can elect to either answer and/or “follow” the question and receive notifications when an answer is submitted. As those answers are submitted, anyone can vote (up or down) on an answer or flag it for a variety of reasons. If a Quora thread has enough answers and votes, the idea is that the most relevant answers will bubble up toward the top of the thread, though anyone can review the entire thread for other information. On occasion, someone will take a thread a step further by writing an answer “summary,” which is done anonymously and therefore can be edited by anyone, much like the tension in Wikipedia editing goes. If a question thread is rich with information, users can “star” it and mark it as a “best source” on that specific topic, which increases its authority in the eyes of users and comes up higher within native Quora searches.
Phew, that was a mouthful.
It’s very complex when you break out all the little things that go into a thread. It’s important to consider these controls when thinking about the future value of the site. For instance, advertisements that are targeted against specific Quora threads will know everyone who has subscribed to that thread, their explicit interests, and related questions. That alone on an ad-model basis could be worth billions of dollars. Furthermore, while some content on Quora can have some properties of the old content farms, but by using the crowd to contribute and rank answers, it works toward making the web a bit more manageable and thereby helping the person who posed the question (and those who chose to subscribe to it) get an answer that’s more accurate. And, once those answers are there, the question thread lives on, able to mutate in the future as facts change or when someone is searching for something similar.
Even if it takes a long time to shift behavior for people to search natively within Quora, the site has already demonstrated it can not only surface up content on the first page of a Google search, but it can also be in the top three sites generated by PageRank. That is an astonishing feat for such a young company that collects and organizes contents from its users, who contribute that content freely. Part of why this is happening is because the content on Quora is so highly-structured to begin with that the PageRank algorithm picks up on these tags and relevancy.
The other reason these threads surface on Google is that, while Google has trained everyone online to search by a string of keywords, the true nature of how humans search offline among people is through questions. “Where’s a good place to vacation in the Caribbean?” “How much butter should I use to make a batch of cookies?” You get the idea. In real life, when we were with others in conversation, we are searching all the time by asking questions, oftentimes specific questions. Yet online, the queries we pump into the Google search bar are deconstructed questions written in an anti-conversational way: “beach vacation caribbean” and “butter batch cookies.”
While all this is slowly unfolding, Google — despite its hammer-lock on intent-based search, is more than concerned about the slow trend where users move from traditional search to discovery, largely fueled by the fact that users are leveraging a variety of social networks to send a variety of signals about their preferences, tastes, and purchases. As a basic example, this means that for my last camera purchase, I would conduct a thorough web research project in order to find the best device. My intent to buy a specific camera resulted in a specific query and purchase. For the next camera I buy, I’m much more likely to simply go on the recommendation of a trusted group of friends and/or camera experts, to instead search through the filter of specific people or to stumble upon it and discover my next purchase.
Quora is an enviable position because it’s set up for both for discovery and search. For discovery, I can research new cameras on Quora and connect directly with people who are knowledgeable about these devices, or while browsing through the site, I can stumble upon a thread about how to buy a new camera and start expressing my intent to purchase right then and there. If I chose to search specifically for a certain camera, it’s possible that Quora has already tagged enough relevant pages within its own system with the proper level of metadata that its machines may learn to answer queries over time.
This is where Quora has the potential to move into the semantic web, where a question posed by a human generates relevant, specific answers produced by machines.
This is the “thick edge of the wedge” in the world of Quora.
And, I’ll leave you with one final thought, the key takeaway and why I wrote this to begin with…
When all of these Quora threads are tagged in context within topics and subtopics, it builds out the site’s ultimate secret weapon: Topic Ontology. The ontology built so far within Quora is staggering. (Take a minute to browse your favorite topic in Quora and then search for the topic ontology.) For many topics in traditional verticals, the site has already mapped out all the relevant topics and subtopics, tagged them against other relevant pages, and created an entire hidden architecture of related pages that are all built into its own system with little to no contamination. Think of these topics as plates on a planet, rubbing against one another and moving over time to form entire new land masses — this is how fundamentally groundbreaking Quora could be for the web.
These are big-picture reasons why Quora is poised to be one of the web’s most cutting-edge technology companies. What the average user sees is a pretty site with the ability to ask and answer questions. It appears to be a highly functional service, and it is. To be fair, it has a lot of work to do to keep users interested and make sure the service works for every user — I don’t mean to suggest that is a trivial task. At the same time, no matter what, behind the scenes, Quora is slowly learning about our interests (both explicit and implicit), they way we use language, and our intent through search, following, and voting, using all of this information to perhaps reorganize the web and lay a new foundation for years to come.