I stopped in at Radio Shack last week to take advantage of foursquare’s 20% off newbie special on an iPod Touch. (It’s a great deal. 20% off a current generation Apple product is tough to find.)
The clerk I talked to had no idea what I was talking about. He reluctantly brought over the manager. She had no idea what I was talking about. She stared at the offer on the screen and couldn’t figure it out. She tried calling another store. Again, no idea what was going on. Then she called Radio Shack’s POS support line and was on hold for about 20 minutes.
If I were an ordinary customer, I would have been fed up and left. But I like to see how these things play out and consider it market research, so I let it go on. I amused myself as the manager spent her time on hold trying to sell me batteries, extended warranties, screen protectors, armbands and pretty much anything else that was within reach.
I asked if she could just override the system and add the discount. No, store managers don’t have that discount. Finally, she randomly entered promotion codes and figured it out.
Thirty minutes after I entered the store I left with my iPod. During that time she couldn’t help other customers. It was’t a great experience for me, the other customers, the store or foursquare.
Clearly the offer code was not a single use code or she wouldn’t have been able to guess it. Either Radio Shack needs to get much better at training or they need to put POS instructions right on the foursquare offer. (They also need to better staff their POS support desk. No one ever picked up.)
This isn’t limited to foursquare or Radio Shack. I run into this all of the time when trying to redeem mobile offers. My default expectation is that it won’t go smoothly.
That’s one thing that appeals to me about Square’s Card Case and Register. Because the POS system is integrated with loyalty rewards and promotions, here shouldn’t be a disconnect between the offer that I see on my screen and what the merchant sees on hers.
There has been a lot of discussion of late on the pros and cons of running Groupon offers for businesses. Here are five cases where running Groupons make the most sense:
You have a new business. If you’re a brand new business and don’t have existing connections in the market, Groupon can help introduce you to your local community. A good example of this is the Globe, a pizza place in Portland that recently opened in a troubled location. A new business also avoids a significant cost that many businesses forget when doing the Groupon math: margin lost on existing customers who instead use a Groupon, turning a full-price sale into a 75% off sale.
Your business has a subscription model or high lifetime value. This category includes gyms, yoga studios, flight schools. Many of these businesses already regularly run introductory specials, such as the first month for $30 (regular price $120). Converting a onetime flight lesson into thousands of dollars in ongoing schooling makes a lot sense. In cases such as gyms and yoga studios, the incremental cost of serving customers is small. Even in these cases, you need to be sure to cap the deal at a manageable number. You don’t want overcrowding that frustrates your loyal full-price customers.
You have fake pricing to begin with. Many businesses have fake pricing. If you shop at FTD, furniture stores, mattress stores or the Gap and pay full price, you’re a chump. The regular discount levels at these sorts of outlets run 30%-50% off. Starwood Hotels offers a 50% off certificate that you can buy with 1,000 Starpoints. But the discount is valid only off the “rack rate,” not the prevailing rate. In 12+ years, I’ve only found the “discount” to be valuable once or twice. Even then, it amounted to 5-10% off the regularly offered rate. If you have fake pricing, Groupons can work for you. But don’t push it too far — FTD wound up not smelling so good when Groupon customers discovered that in order to use their discount, they had to go to a special site which had even higher prices than FTD’s regular site.
You have a third-party really footing the bills. Magazines and newspapers fit into this category. The real goal here is to sell eyeballs to advertisers. For newspapers, auditors will count it as paid circulation if the customer pays even one cent per issue.
You are running an event with excess capacity. If you have an event like a theatrical performance, concert, etc., with a fixed cost of production and you’re unlikely to sell out, Groupon or other deal providers can help fill the audience.
You are going broke anyway. A Groupon can be the ultimate Hail Mary. With Groupon, you get your share of the deal within 60 days. Maybe the extra cash flow will help you smooth over tough times. If it works, great. If it doesn’t, well, you were going out of business anyway.
Yesterday, I wrote about Color, a new app that has the promise to become a ubiquitous, location-aware sensor network. It’s first incarnation is as a photo sharing application available on iPhone and Android.
The initial launch has met with much criticism, including comparisons to Wave — a doomed social effort from Google.
Consider the similarities:
Enormous expectations. Wave was hyped by Google and given high profile executive attention at Google I/O. Color’s expectations have been set by having raised $41 million.
Poor out-of-the-box experience. Both Wave and Color have poor first experiences for the casual user. Even industry luminaries are scratching their heads.
Big change in user behavior. Both Wave and Color go against established patterns of user behavior. Wave tried to replace email. Color is challenging the notion of manually creating friend lists.
Google made a number of key execution mistakes in the launch of Wave. Fortunately, Color has avoided the biggest one: Wave was opened slowly on an invite-only basis. Despite the fact that the product was based on group interactions, you couldn’t get enough invites. I know. I tried to get my entire team to use Google Wave, but I couldn’t secure enough invites. Color doesn’t require a special invite.
There are a couple of other key mistakes Google made. Here’s how Color can avoid them:
Lack of notifications. When I would make a change in Google Wave, the other participants had no way of knowing that I made the change. (Short of logging back into the service.) After I did this a few times and got no response, I stopped using it. The same exists with Color. I will see people’s photos randomly added to my Color application, but I don’t get notified when it happens. Getting notified when other people nearby are using Color would increase usage because you wouldn’t feel like you were talking to an empty room. It would also make face-to-face interactions easier. Notifications are an important part of ramping up any social network. At some point, your product will become so popular that people will use it all day unprompted. (I shut off Facebook email notifications long ago.) Until then, you need to nudge people to use it.
Lack of clear use cases. Most people have a really difficult time adopting radically new product concepts. You need to hold their hands and show them how it could apply to their lives. Wave didn’t do that. You were dumped into a blank canvas with a lot of unfamiliar controls. Color is much the same. There’s little guidance as to how Color can improve your life. The initial controls are so tight that you also can’t easily see how other people are using the product.
Then there a few things that are specific to Color and its goals that should be improved:
Lack of location/time liquidity. Color matches you with people based on photos being taken at the same place at the same time. That’s overly restrictive. Outside of major events and cities like San Francisco, this is going to be infrequent at best in this stage of the product’s adoption. It’s as if you launched foursquare and could only see tips left by users in the last 5 minutes. Older content has value. A few years ago, I took a set of pictures at Liberty Tavern. These pictures are valuable even now. And they’re certainly better than showing nothing. Showing older content would also encourage more people to take pictures. If privacy is a concern, older pictures with faces could be excluded with face detection software.
Locations aren’t visible. For a product that is focused on location, it doesn’t do a good job of showing it. I have random people in my Color feed, but I don’t know where I might have bumped into them — I have to guess at that. It would be better if I could select a person and see a map of where I met them with the date and time. Someone commented on one of my pictures asking where it was taken. That’s not a question they should have to ask. The data is already in the network; it should be accessible. (With the caveat that private places should be obscured so that someone doesn’t follow you home.)
People can’t connect with experts. One of the big reasons for the success of Twitter is that it works even if you don’t have any friends. When I’ve done user research on social products in the past, I inevitably had people who said “I don’t have any friends” or “my friends are stupid.” Social products need to work even in these scenarios. In fact, most of my real friends aren’t on Twitter. But I can still derive value from the people who are. With any social product, you’ll have a few people who are on the bleeding edge who can seed content for you. Exploit that. To fit into Color’s model of not requiring explicit follows, they could be added automatically if someone browses their pictures.
Although on the surface, Color seems to be another mobile photo sharing app, it is really the first incarnation of a ubiquitous location-aware sensor network.
Today’s cell phones are in many ways more powerful than laptops and desktops because they are packed with sensors. A modern smartphone has GPS, WiFi, Bluetooth, compass, gryoscope, light sensor, microphone and camera — at a minimum. All of these data can capture data to be analyzed.
Color is trying to take all of those inputs and layer social networks on to them.
If Color’s vision is fully realized (or my vision of Color’s vision), we can expect to see applications like these:
Breaking news. By detecting abnormal usage spikes, Color could quickly identify where news is happening. Because the app is automatically location aware, it’s possible to distinguish between people who are actually at the scene and those elsewhere who may be reacting to the event. See my post Adding Color to breaking news.
Race finders. Marathons and similar events today use chips to track runners. Imagine that Color is able to identify all of the spectators and runners with the app during Bay to Breakers. Based on your previous social interactions, Color would know who your favorite runners are. Not only would you be able to track their position on a map, you’d be able to zero in on the pictures that are being taken in the vicinity of those runners. It would also be able to provide you a map to reconnect after the race.
Person-to-person transactions. Going to a game at AT&T Park, but don’t have a ticket? Fire up Color and see people nearby who have tickets for sale. Tickets from people you know would be prioritized. Instead of sitting next to strangers, you might end up next to friends who have an extra seat.
Person recognizer. This could be a huge boon to people with a poor memory for faces. The person at the party looks vaguely familiar. You know you’ve seen them before, but you’re too embarrassed to ask for the name. Pull up previous interactions and find out their name and the contexts in which you’ve met.
Bar finder. When I go out, I often have a mood in mind. I may want to be really social or I may want to chill. With Color, I could pull up a bar and see what the feel is right now by looking through the photostream. If there are no pictures, I could potentially ping someone there and ask them to take to a picture. (It gives new meaning to “Would you mind taking a picture for me?”) Foursquare is providing a variant of this with Foursquare 3.0’s recommendations.
Search and rescue. Missions could be tracked automatically, making for more efficient operations. Pictures from a location could be used to identify victims, discover who may still be missing and to notify next of kin.
CalTrain tracker. Instead of the horribly inaccurate data provided by CalTrain, Color users would automatically crowdsource the data. You wouldn’t even have to check manually for updates. They would be automatically pushed to you.
That’s the grand vision. In order for Color to accomplish any of these things, it will have to reach large scale. This is a challenge because Color is a seaparte application and not built in to the OS. Google can use Android phones to detect traffic because it’s baked into the OS. Likewise, Google and Apple get location and WiFi network information based on other things that people do on their devices.
Color needs to create an application that provides enough value that people launch it and enable all of those sensors. The application that’s out right now falls short of that goal. It doesn’t deliver an instant wow experience and by most accounts is confusing. Color has tremendous potential, we just need to see that demonstrated better.
CNN began testing QR codes on air this weekend to direct people to a site where they can help Japanese earthquake and tsunami victims.
The code was easy to scan, even without pausing the broadcast. It worked fine from across the room. Just launch a barcode scanner and it will decode the URL and give you the option to open it in the browser. If you have a scanner, you can scan it off the image above. If not, click to go to the Impact Your World mobile site.
This is a great implementation of the often over-hyped QR-code technology. Print ads have occasionally featured QR codes which take you to an advertiser’s URL.
Some other applications I’d like to see:
In movie trailers. Scan it and it gets added to your movies to see list, possibly with a calendar entry dropped on the release date. Or an option to add to your Netflix queue for movies that are less interesting.
In TV promos. Scan it and it gets added to your recordings list. (The better implementation would be that the DVR itself would recognize a tag and prompt you.)
In TV commercials and on billboards. Scan to go to the advertiser’s site.
On CNN. Scan to get more information on a story.
QR-codes have a number of advantages over other technologies. They are free to generate, don’t require any hardware beyond a camera, hold more data than a standard bar code, are easy to replicate, work across a distance and have a built-in call-to-action (scan me!). QR-codes can also hold structured data; scanning the QR code on Rakeshagrawal.com will load up my contact information.
But it’s not the ultimate technology for every application. As much as people in the technology industry like to claim that one technology will take everything, that rarely happens.
Here are some other applications where other technologies work just as well or better.
Identifying artwork. Many paintings in the MoMA’s collection can be identified just by taking a picture of it with Google Goggles. Let’s face it, QR codes are ugly. They’re designed to be easily readable by machines, not to be pretty. I should point out that the wacky kids in Dubai are trying to turn them into architecture with a QR-code hotel. Still, it’s not my taste in architecture.
Payments. Because they are easy to reproduce, QR codes (and bar codes in general) aren’t well suited for payment applications. They only work when you don’t really care about security.
Scanning books or products. One discussion that came up recently was using QR codes in stores like Barnes & Noble to identify whether a book is available in nook format. That’s overkill — you can do this perfectly well with the bar code already printed on the book. Heck, you can take a picture of the cover and that’ll work.
Print ads. URLs can be detected with simple OCR software. No need to clutter your creative with an ugly QR code. The key here is to use a simple font against a high contrast background and leave space around it. That’s a good practice anyway to ensure that human eyes can read it.
Checking in to a business. WiFi and GPS positioning do a reasonably good job of this without requiring businesses to do any extra work. This could be improved, but it works OK.
I remember seeing Dennis Crowley at SXSW in 2009 shortly after he had finished a panel on location services. He was on the phone and giddy that they had just reached 3,000 users. Two years and more than 7 million users later, foursquare’s latest pivot is its most important.
Foursquare 3.0 is the most efficient way to get recommendations for places to go in a city. The new release hits many of the themes that I’ve covered in my series on local search, especially user and merchant engagement and the importance of recommendations. (See How the battle for local search will be won for an overview.)
Lately, foursquare has been afflicted by a form of reverse network effects. It’s become so popular among some segments that the game elements of foursquare weren’t appealing to many users. Most people don’t want to play games that they will suck at. When you check in somewhere and see that you’ll need 28 more visits to become mayor, there’s less incentive to participate. (See this discussion of two-sided markets.) In my research, I found that the average number of check ins per unique user at businesses was 1.56. Given that some places require 12 or more checkins in 60 days to become mayor, that’s a clear indicator that repeat checkin activity is being driven by a very small number of people.
Foursquare added a new “Explore” tab that provides recommendations of nearby businesses. Recommendations provide a real incentive to use foursquare even if you have no shot at becoming mayor. Foursquare will filter through thousands of local businesses to identify those that you might be interested in. (See my post on recommendation engines.) Much like Google’s Hotpot, recommendations come from a number of different sources:
Your friends. If you have friends on foursquare, you’ll see tips that they’ve left at businesses. Because you can friend or follow organizations, this can also show you “expert” recommendations.
A place graph. Places will be suggested based on other places that you’ve been.
Popular places on foursquare.
Time of day.
It is really easy to scroll through the recommendations. Unlike Yelp and Hotpot, you don’t have to flip back and forth among pages. Unlike Hotpot, non-recommended places aren’t mixed into this list.
Are the recommendations perfect? No. Recommendations never are. I’ve gotten plenty of bad recommendations when talking to friends or hotel concierges. But they sure beat scrolling through an undifferentiated list with no guidance at all.
With the new recommendations, foursquare would benefit from a much broader social graph. Many people are stricter about who they will become friends with in foursquare because of its real-time nature. But they really don’t care if other friends, co-workers and acquaintances see tips from visits in the past. In my case, I’ve only friended half of the potential Facebook friends on foursquare. These additional people would dramatically improve coverage of social recommendations.
Four new deal structures allow for better serving both end user and business needs:
Flash specials. These are the equivalent of door busters. Once activated, the first X people who check in can claim the deal. These are also especially well suited for use as yield management tools. Slow night at the bar? Offer a special for one night only.
Friends specials. Check in with a set number of friends to get the deal. These promote behaviors that are important for foursquare’s growth. They serve as both user acquisition tools and product improvement tools. The more friends that you have actively using the platform, the more useful the recommendations become.
Swarm specials. These kick in after a certain number of foursquare users check in at a place. These could be important for future friend discovery tools.
Newbie specials. These specials allow businesses to incent trial by new customers. Because they are tied to a relatively persistent identity, they can be much more generous than paper coupons (which might be abused by unscrupulous customers).
As important as a deal platform is, having actual deals is critical. Foursquare is launching the new platform with deals from Barnes & Noble, H&M, Toys R Us, Arby’s, Coffee Bean, H&M, Sports Authority, Whole Foods, Chili’s and Radio Shack.
There’s no word on notification regarding specials. Foursquare currently requires users to actively seek them out within the app. Being able to get alerts of new specials (especially flash specials) will be important.
As much as I like foursquare 3.0, there is a lot of opportunity left. The biggest of this is collecting more data to make better, more intelligent recommendations. Foursquare will be launching a new partnership with American Express at SXSW. The initial scope is limited to special offers for AMEX cardholders.
AMEX has the absolute best data of any company in America on where a large portion of the population transacts. Consider these pieces:
Many businesses set up a merchant account ahead of time, so they know when a business is going to open.
They can estimate hours of operation with reasonable accuracy. (Based on swipe data.)
They know the average purchase amount.
They know the mix of locals vs. out-of-towners.
They know the mix of personal cards vs. corporate cards.
They know if a business goes out of business. (The swipes stop happening.)
They know which businesses people return to and which they don’t. (Implicit quality rating.)
All of this data could be fed into a recommendation engine like foursquare’s.
With the way AMEX issues secondary cards, it would be possible to create a card that would automatically check you in on foursquare when you swipe your card. (While protecting your privacy on other purchases.)
At the extreme, you could have a co-branded foursquare card designed around social features. These could include automatic check ins, cardmember specials and a rewards program based on purchase and check in activity.
Sound crazy? Maybe. But back in the mid 2000s, AMEX launched city-centric cards for New York (IN:NYC) and Los Angeles (IN:LA) in a bid to attract younger customers. These cards weren’t successful and are no longer available.
Google’s initial goal seems to be to get as many ratings as possible. To that end, it has made giving your opinion very easy. While Yelp encourages long-form reviews with a lot of detail, Google encourages basic star ratings. It’s primary Web interface makes it easy to quickly rate many places. Animations when you’ve completed a rating add a touch of fun to the process; once you’ve rated a business, the card flip over to allow you to write a review. The box is sized for about four sentences. Restaurants can also be sub-rated on Food, Service, Atmosphere and Value with a smiley face or frowny face.
On Android devices, a widget makes rating possible without launching the Google Places app.
Hotpot integrates with your search history on Google. This serves as a reminder to rate places you may have recently visited. Given Google’s vast query volume, this is another important differentiator.
Hotpot also shows ratings and reviews. While Google builds up its ratings and review corpus, the page focuses on aggregated reviews from other local sites, including Yelp, insiderpages, CitySearch and others. This has been a bone of contention for Yelp’s CEO, Jeremy Stoppelman.
Google hopes to make intelligent recommendations with all of your ratings data. Instead of having users sift through mounds of data to find the right business, Google does the lifting for you.
Recommendations come in two forms:
Recommendations based on your previous ratings. These span venue types. For example, Ikea was recommended for me because I rated Voodoo Donuts highly.
Recommendations based on the ratings of your friends.
The quality of recommendations seems to be hit-and-miss so far. Some seem entirely logical; others, like the Ikea recommendation were baffling.
A few examples to allow you to judge for yourself:
Recommendations don’t currently span metro areas. For example, if you rate places in San Francisco and then visit Chicago, the San Francisco data don’t seem to be used to make recommendations in Chicago. Google could use data like cuisine and price preferences to make at least a first cut at recommendations.
Recommendations are surfaced in a variety of places, such at the Google Places app, Google Maps and most importantly, Google search. In the screenshot above, you can see a recommendation embedded right in the search results.
This placement and personalization is an important differentiator that may drive users to Google Hotpot over Yelp and other competitors. Here, you can see a review from my friend Adam embedded in the search results:
Local is the perfect place for social search: It reflects how we do things In Real Life. Friends and family are often the first places we look for advice on restaurants and nightlife. Even reviews from people whose tastes we disagree with are helpful.
Google’s big challenge with social recommendations is the lack of a good social graph. I have exactly one friend feeding into my Hotpot recommendations. Other players such as Yelp and foursquare have piggy backed on Facebook’s social graph. Google can’t. And after last year’s Buzz privacy issues, Google is likely being more cautious in using other Google-collected data for a social graph.
A significant problem with the recommendations is that they aren’t used as a filter. This is especially important in mobile, where screen sizes are smaller and patience is usually shorter. In one search, the top results was a recommended place. The next results that were recommended were in positions 14 and 30. In between were places that were farther away and even some places that were closed.
When I searched for a restaurant in downtown San Francisco from my Android phone, the first personalized result was Adam’s Osha Thai restaurant, in position 16.
The stated purpose of Hotpot is a ratings and recommendations tool; the recommended places should be at the top of the list.
Google’s mobile search app (called Google Places) is in some ways comparable to Yelp, but Yelp’s mobile app is overall still a stronger experience.
Google Places provides a number of filters, including distance, rating, currently open, price and neighborhood. Additional filters (hidden behind the >>) allow you to search by cuisine or ambiance.
The “Open now” filter is especially important on mobile devices, where the focus is often on the here and now. In the listings, you can see annotations such as “Open until 10:00 pm” and “Opens at 4:30 pm” for places where Google has such data. Yelp’s hours data seems to be much more comprehensive.
There’s also a filter to see just the places that have been rated by friends. Oddly, there’s no way to see just the places that are recommended by Hotpot.
Places shows offers from the few businesses who are using Google Offers. There’s no way to show only businesses with offers.
Like Yelp, Facebook and foursquare, the Places app allows users to check in to a business.
The Places app doesn’t allow users to add new businesses or upload photos.
A digital to-do list
Hotpot allows you to “Save for later”, which is a great way to keep a list of places that you may want to visit later. These are integrated with Google Maps (on the desktop and in mobile) and shown as stars whenever you render a Google Map. It can be helpful when planning trips — you may discover that a shopping trip takes you near a restaurant that you’ve been meaning to visit.
I have been using Yelp’s bookmark feature (and Google My Maps before that) to track restaurants I want to visit; the integration with Google Maps may have me switch to Hotpot for that. It would be nice if Hotpot let you record why you saved it for later (e.g. recommended by Epicurious, have Groupon).
The biggest problem with Hotpot right now is that the overall experience doesn’t hold together. There are numerous brands being used, including Hotpot, Google, Maps and Places. In some places, clicking takes you to a map-based page, other places take you to a listings-based page. Icons and terminology are all over the board. The mobile app is similar. Maps, Latitude and Places all seem to point into similar experiences.
Listing freshness. Having up-to-date listings is an important part of the local search experience. Here, Google lags both Yelp and foursquare, especially when it comes to new businesses and non-standard places such as food carts. Hotpot doesn’t seem to be designed to address this problem. There isn’t an obvious way to add new listings. Hypothetically, Google could algorithmically find new businesses by looking at search patterns and traffic to sites like Yelp.
Engaged consumers. Yelp and foursquare have highly engaged users who significantly enhance the quality of the data.
Engaged businesses. Facebook and Twitter have engaged businesses who regularly update content about their businesses. Apparently Google offers a similar feature, but in more than a year since its launch, I’ve only seen one business use it. I found that while doing research for this post.
Tighter integration with Android. There are opportunities to improve the local experience by integrating better with the phone experience. For example, sending a message with location information could be more seamless.
Google definitely doesn’t have the most features or the most engaged audiences. It’s not (as far as I can tell) trying to build local communities centered around reviewing places.
But Google has three things that are hard to match: incredible distribution from Google search, deep pockets for promotions and Android. Facebook is the only company that can really come close to Google when it comes to distribution.
Google can surely solve the branding and consistency issues that make the current product experience frustrating. The bigger question is whether Google can develop a social graph that will really drive home the benefits of Hotpot.
Thanks to Mike Blumenthal for the pointer to Google’s business status updates.
Disclosure: I have several good friends who work at Google and went to high school with co-founder and CEO Larry Page. I’ve benefited from free drinks and other Google schwag at various Google promotions in Portland.
Local search has changed dramatically in the last decade. Gone are the days when you could buy a generic database from a mailing list provider, slap maps on it and have a local search solution. Social networks, mobile phones and businesses themselves are changing and enriching local search.
These are the key factors that will define success in local search going forward:
User generated content and engagement
The best local search databases are content rich. They include attributes such as hours of operation, friendliness of the place to kids and pets, whether there is outdoor seating, etc. Many of these attributes are collected by users themselves. Increasingly, this is being done on mobile phones — people can update data before they’ve even left the place.
Users also help to maintain the quality of the databases. In my research, there wasn’t a single case where Yelp or foursquare didn’t have a place I was looking for. There were quite a few that I couldn’t find in Google Places and Facebook. For the U.S., the best database of restaurants and bars is at Yelp. New places are often in Yelp’s database as soon as the place opens. (Sometimes even before the official open, as people participate in friends and family dinners and soft launches.) Foursquare’s data are also comprehensive, but are cluttered by users who try to exploit the service’s game mechanics by creating extraneous venues.
Users can also report businesses that have closed, helping alleviate the frustration of driving to a business only to find out that it is no longer around. Check in data on foursquare and Yelp can also identify anomalies. (e.g. a check in stream that suddenly stops can indicate that a business has closed.)
Photos are key components of some of these databases. The growth of smartphones will only further this trend. Some venues on Yelp have a hundred or more photos. Yelp reports that its users are uploading a photo on average once every 30 seconds. Foursquare recently introduced photos. Google is sending professional photographers out to take pictures of top places. Specialized applications like foodspotting have small but loyal audiences who upload pictures of specific dishes at restauranst.
When it comes to mobile data collection, Google’s Hotpot is weak. New places can’t be added and photos can’t be uploaded. It supports ratings, reviews and identifying problems such as closed businesses and duplicate venues.
But providing tools isn’t enough; it’s important to provide the right incentives.
Yelp has done a great job of providing ordinary users incentives to contribute to the maintenance of its database. It uses both social reinforcement and more tangible rewards. Yelp makes it easy for members to thank and compliment each other for reviews. Selected reviews are featured in weekly newsletters. Review often enough and you become a Yelp elite and have a badge on your profile.
Yelp employs community managers in its markets to help reinforce the community. Frequent events (including Yelp-hosted parties) provide more incentives to review and create adhesion among the community. Only a small proportion of the Yelp user base does any of these things. But you only need 1 person to provide value to millions. Yelp’s dedicated, engaged user base will be a significant barrier to other competitors in the space.
Businesses are using these tools to communicate specials, announce closings (e.g. for private events), run promotions, have contests and just engage with their customers. I’ve even seen businesses helping businesses; one business had electrical problems and another business offered her electrician’s number in response. Here is a snippet from a Twitter list I created of restaurants in Portland:
This sort of real-time information can help sway a decision or prompt users to go out on a night when they would otherwise have stayed in. Radio Room in Portland does a great job of this with their Twitter feed.
The image to the right is an image from the Hops Cam at Beachwood BBQ in Seal Beach, Calif. It allows users to instantly see what’s on tap now. What spot now?, an iPhone app, allows users to see real-time cameras from various restaurants.
Although Google, Yelp and foursquare allow businesses the opportunity to claim their page, there is no mechanism to communicate with customers through their platform.
Businesses are claiming pages and providing enhanced attribute information. Nearly 2/3 of businesses I looked at have claimed their Google and Yelp pages.
To date, no one has done a great job of making recommendations based on a user’s preferences or social network. Local search has required users to sift through mounds of data or just go for serendipity (like in UrbanSpoon). Yelp and foursquare have had some form of social recommendations. Both will highlight recommendations from friends, but their social graphs haven’t been large or relevant enough.
This is a key focus area for Google Hotpot.
When you do a search, you might see recommendations based on other places you’ve rated.
Or you might see that a friend has rated the place. Unlike recommendations from strangers, this provides immediate context. I know some friends whose tastes are similar. If they like a place, I know the chances are good that I will like a place. Negative affinity can be helpful, too. There are a few people whose tastes are so divergent that I know not to go someplace they rate highly.
Pictures also play a big part in decision making. Local search has long relied on textual data because it’s been easy and available. But visuals are a key part of the experience when it comes to dining and nightlife. They can answer questions like “Is this place fancy or a dive?” and “Would this place be a great place for an anniversary dinner?” much quicker than text reviews can. See Picturing a new vision for local search. Pictures are also much easier to go through on a mobile device.
Making intelligent recommendations requires having a lot of data. The easier you can make it, the more participation you will have. Few people will go through the trouble of writing detailed reviews, but 1-click ratings can provide important signals and will have a higher participation rate. See more on recommendation engines for local search.
No matter how good your content is, it doesn’t matter if you can’t get it in front of people. Here, Google has an indisputable advantage. Google sites serve 170 million people in the U.S. Yelp reaches 26 million. (Many of these come through the help of Google’s search results.) Foursquare claims several million downloads. The difference in scale is enormous.
Google’s distribution advantage extends to mobile with prominent applications on iPhone and deep integration within the Android OS. Facebook is also a large player here, with more than 150 million unique mobile users worldwide. When they set their eyes to local, they will be a big player to watch.
Local search often involves a shared experience. Plans are made and coordinated. So far, no one has really provided a great solution for this. Here’s a simplified version of how the process often works:
Step 1: Person A looks up a place on a local search site.
Step 2: Person A sends the place name via SMS to Person B (and C, D…).
Step 3: Person B gets the text message and looks it up in a local search site to find the address and look up information.
Step 4: Person B responds to Person A that it’s acceptable. (Or not, back to Step 1.)
Step 5: Person B then uses the site to generate driving directions.
This could be greatly simplified. Again, Google’s deep integration into Android provides an advantage. Person A could find and text the place information. The receiving phone would identify that the link is specially formatted and instead of presenting it at as an SMS, would present a Places page with pictures and reviews and an accept/reject button. Such sharing could also help Google build out a social graph.
Social search has been talked about for several years now as the wave of the future. We’ll get better information with the help of our friends. Local is the ideal place to prove that out:
Most people tend to have a lot of friends in their immediate area.
Local search revolves around everyday experience.
The “answers” are based on opinions.
Google’s entry into the space is Hotpot, a local ratings and recommendations tool. It is spending a significant amount of money to promote Hotpot in the Portland and Austin markets.
Hotpot is clearly meant to compete with Yelp. To a lesser degree, it competes with Facebook Places and foursquare. (It’s funny how much the local search space has changed in the last few years. AOL, Mapquest, CitySearch and newspaper Web sites have largely dropped off the local map in recent years.)
It’s important to set the context in this fight: Google is already the undisputed leader in local search. Despite the attention that other sites get, Google is the number one place people go to get local information. More than 20% of Google queries are local in nature. Google Search serves about 170 million users. I bet 99.99% of them have done a local query. Yelp serves 26 million users in the United States. But many of those users come through the Google front door. (Partly because Yelp is one of the best SEOs out there.)
There are two core problems to be solved in local search. Providing someone additional information on a business whose name they know and providing guidance to those who are open to suggestions on a business.
Business name searches
The first problem is largely solved, despite the fact that the scope of the problem has increased. Just a few years ago, it meant providing someone the name, address, phone number and a map for a business. Today, it increasingly includes providing hours of operation, attributes such as romantic, kid-friendly, links to make reservations and menu information.
Distribution and integration helps Google capture business name searches. You can use the browser’s search box and Google.com to get your answer. With an Android phone, it’s even simpler. Press a button, speak your search and the answer appears.
Google can answer most of the basic questions about many businesses in the United States. Yelp has the best data out there for restaurants and bars in the United States. I’ll get to the reasons why later.
Google has difficulty with non-standard venues. For example, in Portland, it does poorly with food carts. In most cases, I don’t advocate manually updating a database to address localized concerns. But given the amount of money that Google is spending on promotion in Portland and the importance of food carts in the city’s dining scene, they should follow the advice of an Oregon company and “Just do it.” A basic effort could be done in a day by using online resources. A street team could hit all of the major food cart areas and provide enhanced data such as hours and pictures in a few days. (While also handing out Google stickers.)
The other core problem in local search is discovery — helping to find an appropriate answer when they only have a few parameters or no clue what they’re looking for. These are the questions like “I want a kid-friendly pizza place nearby.” “I want to go to some place fancy,” “I’m looking for a special night out on the town.”
This is an area that Yelp excels at but Google generally sucks at. The problem with Yelp (and the opportunity for Google) is that getting the most out of Yelp requires a lot of work from the user. Yelp has an incredible amount of rich data on local businesses. But it’s too much. It’s overwhelming to see hundreds of reviews. Using Yelp also means trusting people you don’t know, whose tastes may be very different from yours. And it means dealing with the snarkiness of reviewers who often spend more time talking about their life stories and girlfriend problems than the business they’re supposedly reviewing.
Yelp has introduced a number of tools over the years to alleviate this problem. It does data summarization across reviews so that you can see at a glance what are the things most frequently mentioned about the restaurant (e.g. popular menu items). You can see a distribution of the ratings to see how consistent a restaurant is. You can also see ratings trends to see if the restaurant is getting better or trending downward.
But often, people just want a few options. Too much choice and too much data is overwhelming. People don’t want to spend 30 minutes figuring out where to go. We’ve been getting recommendations from Amazon and Netflix for decades. “People who liked X also liked Y.” “Based on your previous ratings, here are places we think you’ll like.” This is especially important in mobile, where people are often more hurried and the screen real estate in which to read is limited. That’s what Google is trying to do with Hotpot.
In some ways, this is an easier problem to solve than Web search. If you’re looking up answers for Jeopardy, there is usually only one right answer. And if Google can’t find it, you know right away. For a discovery-oriented local search, there is more than one right answer. And if the answer isn’t what you were expecting, you won’t know for hours and you might not blame Google. (The restaurant might have had an off night.) For more details, see my earlier post about making intelligent recommendations in local search.
Picking the right social graph
In order to make the best recommendations, you need data. You need data from the user about their preferences and you need a good social graph from which to present options. The more data the better.
This is a significant challenge for Google. Other companies in the social space such as foursquare, Gowalla, Quora and Instagram, have piggy-backed off Facebook’s social graph. That’s not an option for Google. And I’m not willing to spam all of my friends to invite them to use Google Hotpot. The advent of Facebook Connect has made such spamming less socially acceptable. As a result, I have exactly one friend on Hotpot — and he’s a Google employee.
Foursquare’s social graph is OK, but it’s a bit small given the current focus on check ins. The number of people who I want to be able to see where I’m at in realtime is fairly small. But I’d be comfortable sharing historical data on reviews and ratings with a much larger audience.
Facebook’s social graph is ideal for this application. It has a lot of personal connections, including both close and loose connections. The loose connections are important because they help provide coverage that you might not have in your tighter friend circle. For example, the data to make recommendations for Indian restaurants in Paris might be from a former colleague who now lives in Paris.
In the next part of this series, we’ll look at some of the key success factors for local search.
I frequently tout Yelp as the company that has the best local database in the United States focused on restaurants and entertainment. With thousands of Yelpers around the country who aggressively review businesses in their cities, Yelp manages to stay well ahead of their competitors. Where a new restaurant can take months to make it into Google Maps, it’s often listed on Yelp before it opens thanks to devoted Yelpers who keep an eye on what’s going on in their neighborhoods.
Yelp also has another key asset, which has long been hidden: a large volume of pictures uploaded by Yelpers. While these have been available on the Web site, they haven’t been the focus. Yelp’s iPad app puts them front and center.
Local search has long been optimized around the data sources that are available and the way computers best process information, not the way consumers look for information. Looking for the address of Lovejoy Bakers? Piece of cake. Local search will find it for you. Looking for a romantic restaurant that’s not too crowded but has a modern feel? Good luck with that.
Here’s where pictures can play a big part. Solving such queries is incredibly hard because they require value judgments and computers aren’t good at making such judgments. Even among different people, those judgments vary. Romantic, crowded and modern mean different things to different people. If you read dozens of reviews, perhaps you could get a good sense for whether a business meets your definition of these words. But that’s work that very few people are willing to do.
It’s much easier (and more fun) to flip through dozens of pictures.
Pictures provide easier and faster answers to:
Is this place a dive?
Does this place cater to people like me?
It this place kid friendly? I never would’ve guessed that a brewpub near me was kid-friendly until I flipped past a picture of a play area with kids in it.
What does this place feel like?
Is the food pretentious?
Pictures also help with another problem that many user reviews have: too much time spent talking about the reviewer rather than the place being reviewed.
Popular venues in major cities such as flour+water and 21st Amendment in San Francisco can have more than 100 pictures. In smaller cities, it might be just one or two.
We’re just at the beginnings of truly using images in local search. I imagine that we’ll soon see image recognition algorithms that will sort the uploaded pictures into categories such as food, interior, exterior, etc.
Cell phones are increasingly becoming data collection devices and Yelp users are at the vanguard. Yelp claims 3.5 million monthly unique users on mobile devices. If only a small fraction of them are contributing content, that’s still thousands of people providing ground truth. Yelp reports that a photo is uploaded every 30 seconds via mobile devices. With check ins, photos and real-time data corrections, local search is becoming a much richer experience.