A New Google

[TL;DR: Google has gotten bad; we all know it; ideas for a startup making a better Google.]

In 2000, Google got popular because hackers realized it was better than Lycos or Excite. This effect is happening again. Early adopters aren’t using Google anymore.

They aren’t using DuckDuckGo either. They’re still using Google.com, but differently. To make Google usable, users are adding faux-query modifiers that to supress the “garbage Internet”.

You see this in the typeahead logs.

image

Products (Reddit)

Services (Reddit, Yelp)

image

Movies (Rotten Tomatoes)

Even code (Github)

Interestingly, this doesn’t work for all categories.

image

Recipes don’t have a “Reddit” equivalent

Query Operators Mean Something’s Broken

More advanced users use modifiers like site: filetype: intitle: because adding “reddit” isn’t strict enough, as spammy websites often manipulate content to win SEO.

How about those websites that stuff the year in the title? “Reviews UPDATED JANUARY 2020” are exploiting the fact that customers suffix queries with the year. What those people are trying to command is freshness, not a title match.

Something’s broken, and a tiny share of Google is open for the taking. Obviously attacking incumbents head-on doesn’t work. Here are two alternative ideas for bootstrapping next-generation search:

#1 Boogle, A Query Reformulator

Introducing Boogle, a proxy for Google that’s just Better Google Search. It’s a query expander. We predict the correct operators for your query, proxy Google’s results, and serve. For example:

query("stripe.js example") -> query("stripe.js example (site:github.com OR site:gitlab.com OR site:..."))

query("is anker charger") -> query("is anker charger (inurl:forum OR site:reddit.com OR ...))"

Query topic modeling is a rich science with plenty of examples.

You’d almost as fast as Google, never worse, and occasionally better. This will help build the reflex to use you instead. This approach isn’t that hard to get started with, and might work for the high-end users.

#2 Community Vertical

You could go after this vertical by vertical – build the best site for electric product search, for travel, for code, etc. A key question is how to build habitual recall to use your product over Google. Amazon and Airbnb both enjoy a huge amount of direct traffic. Some learnings from those:

  • Stellar mobile destination. Using Airbnb directly feels more fluid and fun than using Google for Airbnb.
  • The 90% Rule. To build reflex you need to give me what I want most of the time.** **With Prime, Amazon made it such that we stopped price comparing across other sites; Amazon would get it to us fastest, and that turns out to matter more than price. Most of the time you open Amazon.app, it satisfies.
  • Come for search, stay for something else. I don’t think of Airbnb or Amazon as search apps. They help me get things or book homes. You might want to your search app to be a destination for something else. This is where I think community comes in.

Picking a vertical that doesn’t have strong typeahead completions[1] would help you build community around your search engine. Recipes, fitness, fashion, etc. don’t have decacorn conglomerates like Github or Reddit. That might mean it’s easier to build community around them, and put your flagship search engine on top.

If you’re working on this kind of stuff, try out Pioneer or just shoot me an email – daniel@pioneer.app.

[1] In reality I’d pick the vertical _I _love most. If you’re a guitar player start with music; a movie buff should build better Rotten Tomatos. Etc.