Every great technology wave has a paradigm, a very broad and deep movement that is only barely perceptible in its infancy. This paradigm typically blossoms into many forms over years, even decades, to come.
The entire internet technology wave, for example, was a communications paradigm rooted in sending messages. Decades ago, mundane email enabled virtually instantaneous communication around the world. And the thirty years of consumer internet thus far have represented the generalization of that paradigm. At a very high level, the internet itself can be thought of as a user sending “emails” to a server and the server sending response “emails” in the form of a webpage. Even Zoom video conversations can be thought of as sending thousands of email packets (of images and sound) between people (but, of course, with different protocols, etc) .
We are now in the midst of another major technology wave, the artificial intelligence wave. The first major paradigm of that wave is search and the use of computation to sift through and understand massive amounts of data.
In the early consumer internet, before Google, navigating the web meant navigating through a series of human “computations.” At that time, Yahoo! (which stands for “yet another hierarchical officious oracle”) was the default way to access the internet’s information, and the product relied on thousands of humans using human judgement to group pages into buckets, like “Arts,” “Business and Economy,” “Computers and Internet,” and so on.
In 1996, Larry and Sergey took a different approach at Google. Rather than relying on humans to make millions of decisions for a manual hierarchy, they trained their computers to “rank” web pages using their Page Rank Algorithm. This effort was based on the prescient insight that the importance of a page could be approximated by the links pointing to the page.
Google successfully translated millions of human decisions into billions of computer computations. And while it wasn’t apparent then, this technology elevated search to the position of “the killer app” of the first twenty-years of the AI revolution. (Though we may no longer consider search to be artificial intelligence.)
As mentioned, the manifestations of technological paradigms become clear over time as they shift into new forms. Email shifted over time. Its basic value, instantaneous communication, was unbundled into messaging apps and social networks. Purchases that once were made over email are now made through companies like Amazon, Facebook, Google, Apple. All of this represents the flowering of the email paradigm.
Similarly, search is unbundling in the AI revolution and we’re seeing a shift in search’s user interface and the emergence of dedicated search vertical:
INTERFACE
The user interface of search is changing. First, and most obviously, the voice paradigm is changing how we can search. For years now we have talked with Siri, Alexa and Google Assistant. What hasn’t been achieved yet is a really incredible application of all that technology. Sure, Alexa has started recommending products to me – but I haven’t bought any of them. The technology isn’t yet at the point where it clearly has a business case. If and when the business case is ultimately crystalized, however, the tech giants will likely lead here.
Second, we have seen visual search, the ability to make a query where the input itself is an image. This can be particularly relevant for online shopping . It will likely not be massive until AR happens in a big way. The giants would also be at an advantage here.
Third, we have disruptors to the classic search business model of Google and Bing. Companies including DuckDuckGo, Neeva and You.com are pioneering ad-free search. The interface in these disruptors is similar- but doesn’t need any screen “real estate” for advertising (because they do not monetize with ads).
THE UNBUNDLING OF SEARCH’S VERTICALS
The other big change in the unbundling of search is the emergence of more-specific kinds of search, often by industry or product. A fundamental reason for this is that the search process often involves digging through successive levels of linked information with greater and greater specificity.
Think about your own search journey: it proceeds in stages. When you look to buy a TV, you likely start on Google. You may start with “Best TVs” and read a couple of blogs. Then you get some more details and you wonder “what is 4k vs 8k vs Ultra HD.” Then you may go back and search for the “best 70-inch 4k TV.” This is called an “upper funnel” search, and it goes on for a while.
The next stage is likely to be making the purchase. At this point, you may have your preferred TV brand. You now search again in what’s called a “lower funnel” search application, which includes Amazon, BestBuy.com, etc, for the specific TV. This example represents a two-stage search funnel, but many funnels have more stages. Amazon has proven to be exceptionally effective at product search.
There is a lot of opportunity in search, especially in the lower stages of the funnel. Lower funnel search examples include:
- Travel: This is one of the oldest search verticals and it supports huge companies, in part because these companies have been able to own the full purchase (top to bottom). Booking, even during the pandemic, supports an $90-billion market cap, while Expedia is worth over $20 billion.
- eCommerce: Amazon has really captured a lot of value in this area. Amazon has exceptionally performant search at the core of what it does – just on a limited (but still massive) number of products. More and more of the buying decision is being made on Amazon alone. This supports a $1.5 trillion company.
- Professional Contact Search: LinkedIn has been a professional business directory for a long time, and it’s being unbundled at this very moment. Take ZoomInfo, which went public last summer and is worth almost $20 billion. ZoomInfo let’s you search for individual contacts’ information. Grata is another early example of a lower funnel search opportunity – it has built a modern company search engine for proprietary deal sourcing and targeted B2B campaigns. Company search is a big opportunity in unbundling LinkedIn. When you are looking for potential customers, how do you go about it effectively? There’s a lot of value in whoever can answer that question. ZoomInfo and others have pieces of the puzzle, but that’s a business that has a lot more potential to be mined.
- Function-Specific Search: Every function can benefit from tailored search product. One function that finds search particularly valuable is software developers. General questions about coding, specific bugs and general code repositories may begin on Google, but they often end up in specific code bases. Companies like Sourcegraph are bringing tools to search code in a much more specific fashion.
These examples are just the beginning. The opportunities will continue to abound. From a technical perspective, in order for search to be valuable there need to be three characteristics in place: a lot of information to sift through; value in getting to the right piece of information; and some way to quantify what is “right” in the search result.
Anyone building a search business should think about that business using a simple tripartite framework. First, can you really help with some decision through search in a way that is above-and-beyond existing modes? Second, is that decision likely to be valuable to some group of users? And finally, are there enough users facing that decision to make for a sufficiently large market?
When those three criteria are met, there is plenty of opportunity to apply the search paradigm. At its core, this new wave of AI is really about helping people make better decisions faster, and search is a proven means to do so.
originally posted on forbes.com by Konstantine Buhler
About Author: Konstantine is a Partner at Sequoia Capital, where he focuses on seed and early-stage investments. He supports a number of portfolio companies, including AppAnnie, CaptivateIQ, Ethos, and Verkada. Prior to Sequoia, Konstantine was an investor at Meritech Capital, where he focused on enterprise software and data-first businesses. He’s also worked at RelateIQ, McKinsey, and The Pritzker Group, and previously founded and ran a non-governmental organization for seven years. Konstantine earned his BS, MS (in AI), and MBA from Stanford University. He is passionate about artificial intelligence and the practical applications of machine learning.
