Local Discovery Is Broken, Here’s Why

I think local discovery is one of the big software problems we haven’t solved yet. The problem is akin to this – Google helps you find what you need online by putting in a parameter and drives traffic to the online world. No one has perfectly cracked the nut of find what you need in the real world by entering in a parameter and therefore driving traffic to the real world. Approximately 5% of spend is done online, yet 95% of spend is done offline, so through that lens we’re looking a pretty worthy problem. With everyone now having a smartphone in their pocket, it’s a problem we need to see solved. PCs helped us find and make things happen in the digital world (Google), smartphones help us find and make things happen in the real world (Uber). Below are some of the problems I see:

Not enough intent data points, therefore no “page-rank”

This is the biggest problem in my mind. Google has data to make relevant search results to show “authority”. They can look at which pages are linking to each other and use the page rank algorithm to determine what’s the best result for said query. Right now the only data points we have to determine the best result are as follows:

Reviews

Reviews are very subjective and often not a valid indicator of authenticity. Imagine if Google’s ranking were based upon reviews of websites. You would have a system that was clearly gamed and results that likely weren’t the best. Results would be based upon which websites could game the system and get the best+most reviews. This scenario is a mixture of cynicism meets reality of what’s happening today on sites like TripAdvisor or Yelp!. Businesses hawk their customers to write reviews and the small few that actually write reviews help determine the ranking of “what’s best locally”. Reviews at best, show which places might happen to deliver a good one-off experience and base this upon the small % of experiences that are actually persuaded to write a review.

Even with that said, reviews don’t focus on the two most important lens’ – personalized recommendations and popularity. Review sites don’t collect enough signal of intent from their users’ activity within their city. If we knew where every Yelp user went in the city, we could of course tell what’s the most popular, but there’s no way of getting that data. We also don’t have intent data for personalized recommendations. Recommendation systems require a lot of data and a lot of signal. I’m “reading”, not “writing” on review sites for the most part, so it’s tough to tell what I might be interested in. I could state my interests like American restaurants, burgers, mexican food, seafood, but that isn’t granular enough. The key to personalization is really: “I’ve been to x,y,z, look at the rest of the data from the network, and tell me places a,b,c that I should go to.”

Review sites are great when you know where you want to go and want to see if it’s any good. They’re not good for when I don’t know where I want to go.

Check-ins

I’ve been a Foursquare user since the early days and still check-in on Swarm. I personally love it, but the truth is that the majority of people don’t want to check-in. In a perfect world, the check-in would be a good signal indicator of my interests. The problem is that it’s not something masses are going to do. Check-ins also need to be consistent. One of the most important signal indicators in an algorithm would likely be “loyalty”. When looking at data network wide, a location should rank higher if it gets more visits from a person. i.e. – simple cohort analysis to show what % of people come back >1, >2, etc. If check-ins aren’t consistent, there’s going to be lapse in the data set. If I used the check-in system on a daily basis, BUT not consistently across everywhere I visited, then you might miss some interesting signals of what I actually do. You might miss the Yoga class I go to or the bar that I’ve now visited a second time. Losing random data points messes up the picture of my tastes.

My friends’ tastes

Just because my friend likes something really has no bearing on whether I’d like something. Someone might have odd tastes in food, be a picky eater, and allergic to seafood. Their friends love seafood and therefore would potentially kill them. Seeing where my friends have been is interesting to see, but not a good indicator of where I want to go personally.

Lack of simplicity and too many results

There are too many results in local discovery. It presents paradox of choice and analysis paralysis.
Hunter Walk has a great post about “Hotel Tonight for Local”. One of the key selling points of HotelTonight is that it doesn’t provide an endless stream of hotels to choose from, it simplifies the list. Look at their language in the picture below.

Google might provide hundreds of pages of results for a query, but the UX is clearly setup where most people don’t go beyond the first page. We are all simple creatures and less is more. Hotel Tonight is successful from a UX standpoint, because it gives us less choices.

No purchase attribution in the business model

Now I want to start to talk about the merchant side of things. Not only should local discovery be valuable to the user, but it needs to be useful to the merchant itself. Google offers advertising where you pay per click on an ad and as a merchant, you can place a tracking pixel that shows whether that person made a purchase and how much that purchase was for. Yelp makes it’s money by offering advertising on the site too, but it can’t offer attribution. Just because you clicked on a Yelp ad, there’s no way to attribute a click on that ad to an actual purchase and what the purchase amount was. Real World Merchants should be able to deliver digital campaigns and track the results, the same way that digital merchants do. That’s a very tough nut to crack as there’s been no way to do it. Your digital world isn’t really linked to your physical world. The closest match to this is Facebook’s partnership with datalogix. Once again, it’s not complete as not enough people purchasing give their info such as phone number to a store. I actually don’t like when I’m asked that at check out.

Helping foster relationships and rediscovery

Great local discovery isn’t just about initial discovery, but rediscovery of businesses by helping them build a relationship with the consumer. The relationship between review sites, the merchants, and the users is contentious at best. That’s a big opportunity. If I’m going into a business frequently, I likely want to hear from them or follow them – The right local discovery platform isn’t going to just help me discover new businesses, it’s going to help me build relationships with those businesses and build more loyalty. My favorite place in NYC is Sticky’s Finger Joint. I go there frequently and I want to hear new updates from them about new stores or new seasonal chicken fingers. I get this info on Twitter, but it’s lost in the stream.

Solution Overview

Get a clear consistent stream of signal from every user to show what’s popular and deliver great recommendations. Do it with a simple and limited list of results. Make the business model something that can have a purchase attributed back to the platform and let merchants build more relationships with their customers. Someone should go build that, I think it would be a monstrous company if built right.