[This post was first published at HBR.org on 10th July 2015].
Why aren’t things better? This is something that I often feel these days when searching for information on the internet. And it is especially something I feel when I am searching for products where quality matters. This applies to restaurants, tradespeople, and books, for example. In each of these cases, I am looking for not just anything — but something that is good or even high quality. When I search on Google what I often find is a mess of links to sellers of those products and also lots of review sites. But I can also confine my search to something narrower. I can use Yelp, Zagat, or TripAdvisor for restaurants. I can use Amazon to search for books and read reviews even if I don’t purchase there, although the chances are surely high that I will make a purchase if I find what I want.
Not surprisingly then, the business of matching my search behavior with suppliers and information is a big one. There are people who will pay – usually suppliers themselves – for matches that work out. Google’s entire business model is based on that. People search and reveal their intent, giving advertisers who might satisfy that intent a clear in. All Google has to do is, literally, sit back, ask the advertisers to express just how much they are willing to pay for that connection and then connect accordingly.
There was a time, perhaps five or so years ago, when it was hard to imagine anyone being better at Google in finding stuff. But things have changed. Google is great for finding out the date of the Defenestration of Prague without even knowing what that is, or that there were actually two of them. But if I want a hairdresser who will do a decent job or someone to fix my air-conditioner, I’m never feeling lucky and know I’m in for more work in the form of “click – read – back arrow – click again.”
Many startups have realized this problem and have come into the breach to fix things. They offer specialized search (as distinct from Google and Bing’s universal search) which requires (a) consumers know they exist and (b) that their specialized search can actually give consumers the information they desire more efficiently.
Which brings me back to defenestration which, as Google’s luckiest result will tell you, involves the literal overthrow of rulers over the side of a building. But this isn’t “Game of Thrones.” It’s what search competitors hope to one day do to Google, and what some argue Google is doing to them in its own results in the meantime.
This is precisely what the specialized search companies are accusing Google of doing to them. A case in point is Yelp who, along with academics Tim Wu of Columbia Law School and Michael Luca of Harvard Business School have recently tried to measure what this action of Google has done to search consumers.
What Wu et al. do is consider Google’s search results for something quite specific like “restaurants near me.” This is the sort of thing someone might do from a mobile phone. Here is what I found when I tried it:
There are several options given but notice that they are ranked by reviews and ratings. What reviews are those? They are reviews that were posted on Google+, Google’s “Facebook killing” social network. What Google has done is taken a leaf from specialized search providers like Yelp and offered consumers search results based on their own specialized search product. In this case, it is Google+ but elsewhere it may be Zagat, which Google purchased in 2011. As I wrote at that time, this acquisition was a recognition by Google that having computers work things out wasn’t enough.
The problem, as Yelp sees it, is Google+ isn’t that good. Indeed, what Wu, Luca, and co. show is that if you replaced those Google+ reviews with reviews from other specialized search engines (including Yelp but there are others), then consumers would click on those results 50% more often. If the goal of Google is to provide people with the best results and if, as is plausible, clicking on them is an indication of that, the researchers have shown that Google is not achieving that goal when it favors its own specialized search product rather than being more neutral.
So what should be done about this? First of all, Google should thank the researchers for providing that information as it offers a guide on how to improve its product. But I suspect Google, famous for its experimentation, already knew that. So that means Google believes that it is better off using Google+ rather than its competitors. That is hardly surprising — but many have argued that Google isn’t quite free in its ability to defenestrate competitors and throw them out of search this way.
The reason is that Google has market power. In the U.S., some 80% of searches are on Google while in Europe and many other places, it is more like 95%. That is a challenge for specialized search engine providers because users of mobile devices, in particular, may not realize what they are being served up. From my screenshot above, it was far from obvious that the ratings were from Google+. The European competition authorities have been examining these practices and some other academics have argued that any favoring by Google of its own verticals should be considered illegal.
So what is the best way forward? Google often says it is providing services in the interests of customers. But this study challenges that, and Google’s defenestrations don’t give us comfort that every review is being given a fair go. Then again, this is the internet and so providers such as Yelp are just some good marketing and press away from being able to get people to bypass Google altogether.
One option is to partner with Google’s competitors. When I do the same “restaurants near me” search on Siri, this is what I got:
Notice that it is different from Google’s results but also tells me where the reviews are coming from (in this case, Yelp). But Apple isn’t running a neutral search engine and has just partnered with one that it believes best suits its own commercial needs.
Thus, we have a classic difficulty in antitrust law in a dynamic environment. Yes, there is currently a dominant firm that appears to be favoring its own verticals rather than being more open. But they are not a fortress. Even when they get thrown off the page, specialized search products can find some ways of reaching consumers. How this plays out is still uncertain.
Perhaps more importantly, we have to worry about what this type of behavior says about Google’s commitment to organizing the world’s information. It is certainly great to bring a human equation into search. But even if it’s fine from an antitrust perspective, for a brand built on neutrality, are the returns to favoring your own verticals really worth it?
Digitopoly 
I’ve been puzzled for a while about how you an antitrust settlement against Google or Facebook or other platforms would work. While I’m sympathetic to the case that Yelp is unfairly harmed by Google’s ability to supplant it with its own restaurant recommendations, I also really like being able to access that through Google Maps or if I were on Android by voice query without opening any app at all. Google’s ratings are not as good as Yelp’s, but if they got to a point where they were good enough, say 80% as accurate/comprehensive, I probably would rarely bother opening my Yelp app or visiting the site.
Partnering with Yelp instead of building their own in-house alternative is nice for Yelp, but what about Zagat, or Angie’s List, or Seamless? I’m not sure if it’s still there, but for a while Google had integrated Uber into Google Maps. Was this fair to Lyft? Halo? Gett?
It seems to me that one possible solution, which would likely be quite difficult to implement in practice, would be to require that Google (or Apple*, or Facebook) make these components modular and controllable via APIs which companies can hook themselves up to. Just as I can choose which search engine I want to use in Chrome, I’d need to be able to select which restaurant reviews I want Google sourcing in Maps. The same could apply to car share services, weather information, etc. Implementing and maintaining these APIs sounds like a nightmare, but how else to you get the benefit of a one stop shop like Google Now without the corresponding monopoly?
In some ways what I’m describing is how Android works relative to iOS, where you can replace the default email, browser, text messenger, etc. with your own preferred third party application.
An antitrust solution would be to break Google up into multiple companies, making them compete with each other and breaking down barriers to entry for new competition. Google+ Maps, Android, Adsense, Mail, types of search…. All on their own!
Is a product like Google Now possible if you split up all of the components?
There are two errors in the study. The first is the assumption that more clicks indicate better quality. The opposite may be true in that additional clicks indicates that the information sought is not readily available. Google provides direct contact information which allows consumers to contact vendors directly whereas directories require consumers to enter the site in order to find information. Second of all, it seems to be an invalid assumption that Yelp is even quality or a desirable source of information.
Yelp lists businesses without their permission, places as many as 10 competitor links on the business listing, then uses it superior SEO capabilities to rank above the natural/direct link to the business. When a consumer looks for a business by name, often the Yelp link will appear first. This distracts the consumer and some attrition occurs because consumers go to their preferred vendor on Yelp, then find a myriad of alternative links (this would increase clicks and might also lesson the quality of search). When the business contacts Yelp about its listing, it is offered the opportunity to level the playing field by buying advertising, and in the case of Yelp, the cost is 1000x times the charge of Google (See Zacks Yelp study). In addition, the reviews carried on Yelp are not moderated, uniform, source identified or in many cases even valid. Numerous times it has been proven that reviews were by competitors, former employees or individuals who never visited the business. This provides a false negative bias of business performance.
Yelp also promotes a system of Elite reviewers who receive free food and beverages from businesses. These Elite reviewers in turn review the business. Not only is this unethical, but it is again deceptive to the consumer because there is no indication of the compensation by the business to the Elite reviewers.
Google has in fact increased the quality of search by demoting directory site links and allowing consumers to directly interact with the businesses they seek.
In reaction to the study, an FTC commissioner indicated that the study would not stand up to any antitrust criteria because the study was limited, the assumptions were unproven and the replacement of elements that caused an increase in clicks did not have a right to be represented on Google in the first place.
It seems the ultimate hypocrisy that Yelp, which usurps 100,000 business listings so that the consumer has trouble finding them, would accuse Google of hiding its links. The motivation for Yelp is that without free listings from Google, it can not through SEO outrank free business links and subsequently is unable to charge businesses for advertising. In its most recent filings, Yelp indicated that reduced traffic is to blame for its inability to make a profit. It seems the attack on Google is not to improve search quality, but is motivated by the necessity of higher listings so Yelp can maintain a chokehold on businesses, thus ensuring a profitable enterprise.
It is shameful that many in the press and academia were unable to see through Yelp’s ruse.
I’m an ex-Googler. Look at the larger picture with Yelp.
Originally, Google used information from Yelp pages (like Yelp’s star ratings) and showed them to Google users searching for restaurants. Yelp didn’t like this, they thought it was unfair. Yelp could have removed those pages from Google search, but then their traffic would have plummeted.
Yelp complained to Federal antitrust authorities about this (among other complaints). In response, Google agreed to stop using information from Yelp pages. Review information is useful to users though, so Google replaced the Yelp information with Google+ reviews.
Now comes the complaint that Google+ review information isn’t as good as Yelp’s. Debatable, in my experience, but okay. This complaint is ALSO described as an anti-competitive issue and Google’s fault.
Yelp is playing the same game as German newspapers. They want Google to be forced to include them, AND they want Google to have to pay. No, you get to choose one, mandatory payment OR mandatory carry.
Yelp is for sale. Apparently they don’t like the offers they’re getting, so they want a new guaranteed revenue stream from Google. Maybe Google will pay them just to make them go away. That would increase their price. Well, good luck to Yelp, maybe they’ll get their money.
Here, I think, is a good place to say that Watson may easily change everything — just as revolutionary as original internet search was. Watson (or similar from another company) will be the new search, and so much smarter and better. You will simply say to Watson, verbally, with amazing voice recognition, these are the symptoms of my plumbing problem, what do you think it is? Then, Watson will read far more than just a few sketchy plumbing websites that you might have googled in 2015. It will read thousands of entire plumbing books and manuals, analyze them, and give you a short report with diagrams and suggestions and recommendations (see Rise of the Robots (2015), a must read!), and it will do all of this in seconds. And, it will include recommendations for plumbers who are likely to be reasonably high quality and honest in your area. And it will make these recommendations not just be reading Google’s reviews, but by reading every sites reviews, and linked in, and any resumes out there of the plumbers, and constantly refining it’s algorithm for finding the best candidates via experience and feedback (using machine learning and other forms of AI).
Watson will be as revolutionary as the internet and original internet search. And this is just the tip of the iceberg of what it will do.
I’ll add to this:
Watson may also say, and I do mean say, in an oral, quite normal and pleasant sounding voice, “Given what I know about you [you will have a detailed and private profile with Watson, which will help it a lot to help you], you could probably do this job yourself without too much trouble in about two hours.” “Would you like me to walk you through it?”
And if you answer yes, Watson will ask you to survey the bathroom area in question with your smartphone video camera, saying things like, “A little to the right, please”. Watson will have the complete plans of your house; it will order parts to be delivered, or for the Home depot to gather up front, for you to pick up, and so on.
Now suppose you do hire a plumber. You can video the plumber with your smartphone so Watson can make sure he’s not ripping you off with very expensive work that’s not needed, or overcharging you on parts, or just misdiagnosing it. And Watson will inspect the parts he takes out, again via your smartphone, to make sure they were really broken. Suppose the plumber/contractor gets behind your shower and says my sensor says you have mold, we need to tear out everything,… Watson can say, hook your sensor up to me so I can read how it’s calibrated, and exactly what it says.
The contractor refuses? You find one who allows his equipment to be hooked up to, and monitored by, Watson; you find one who uses Watson-Certified equipment and censors. Or even a contractor that advertises, all of our professionals set up a mechanized tripod hooked up to your Watson account to monitor and assist us, so you can feel confident in our work – a smart businessman could think this could give him a big competitive advantage, making him look a lot more trustworthy and safe compared to competitors who “have something to hide”.
You go to your physician? You take your smartphone with so Watson can monitor a video of the whole thing, and interject questions, and later give a report on what the physician says. The physician says you need CT scans on the machine in his office costing $5,000. Watson says, either then, or later, it’s very unlikely CT scans are necessary, and nowhere near worth the massive radiation. Furthermore, there’s a provider in town which will do the same scans, just as well, for only $800. You physician doesn’t allow this videoing? Again, you find another one that does, like with the plumbing contractor, you only choose a physician who’s “Watson-certified”
And in countries with a strong government healthcare system, the government may require all physicians to have their offices filled with sensors so Watson can watch well, and all equipment to report its findings to Watson. This not only allows the patient to be much better protected from rip-off and mistake, but this incredible amount of data on health and medicine, crunched and analyzed by smart AI/machine learning computers will be revolutionary for advancing medicine.
Watson will be absolutely revolutionary. Again, please see Rise of the Robots, to see what Watson is already capable of right now.
Of course, there’s so much to it. You could see this kind of Watson, and general machine learning and AI ability, just making a lot of work done by plumbers and others unnecessary, so then where will the 50% or 80% have the jobs, and therefore money, to buy plumbing parts in the first place? Or to have a roof over their heads in the first place? With these kinds of machines able to do just about everything a low skilled person can do, and many higher skilled people too, as they might in 10-30 years, then the demand for lower skilled workers at the minimum wage may be nowhere near the supply of such people.
But 10-30 years? What, does that come from your unserious book Rise of the Robots? Well, first Rise of the Robots has been endorsed by some serious people like Jeffery Sachs, and reviewed by some serious media like The New York Times. Its author has a lot of technical expertise in this area. But it also comes largely from many other sources, like Oxford’s Benedict and Frey’s big 2013 paper, which I’ve read in detail. And that paper, at its root, is from qualitative questioning of scientists and engineers at Oxford, not from some vague overblown, simple-minded use of statistics. And, FWIW, machine learning expert Jeremey Howard, in this video from the recent Milken Conference, claims that most machine learning experts think that the time until smart machines can do the bulk of what humans do is in the 10-30 year range. See starting at minute 23 for his exact words, at:
http://economistsview.typepad.com/economistsview/2015/04/video-technology-and-jobs-should-workers-worry.html
So, this is obviously humongous for the coming generation or two. Lot’s to it, but at least, Watson, or the same kind of thing from some other company, or companies, will be the new search. And it will be revolutionary.
Ok (or sorry), not done yet. So, back to being more on subject, with Watson too there is the issue of neutrality in your recommendations; or in the information you give to the user; or biasing to your products, or the products of who pays you.
It’s a very serious issue, and yet another example of how as the world gets more and more advanced, market failure issues get bigger and bigger, and more and more ubiquitous and important. As, I say all the time, it’s not 1810 anymore, the good ol’ days of small gub’ment the Republicans want to take us back to; you know, when the average lifespan was in the 30’s, and the 99% were dirt poor and mostly illiterate.
So to overcome the enormous externalities and asymmetric information here, a large government, or governmental body, like the United States or the EU, might provide a public option Watson.
Or, there may arise an open source Watson. Fifteen years ago I had my doubts about large open source projects. Will there really be something that can compete with multi-billion dollar private projects? People will, for free, provide that kind of work and money? But as we’ve seen, yes, in important cases. Wikipedia is vastly better then Encyclopedia Britannica, in so many ways; just unbelievably larger, and containing content from an incredible number of specific advanced experts that Britannica could never get, or afford. And Firefox is in the same class as the best browsers.
And there could be a public project Watson, from say the EU, that’s also open source. This would be a fantastic way to get an objective Watson, and to put competitive pressure on private Watsons to be objective.
And this, of course, extends to our google issue now. If they’re going to start being a costly problem with unobjectivity, then maybe it’s time for an open source search engine to put pressure on them. Or, an EU, free, no-advertising, search engine that’s open source (Forget about the US for now with the Republican death grip on the House, unless it can be done by administrative action.)
duckduckgo.com gives you private, neutral search. It’s on all my devices.
I read the Luca/Wu paper and have to say that I am as impressed as you by the results. There is very little space given to a description of the experimental design, most of that discussing the A/B methodology and very little on how they controlled for variables that could have affected the results.
That and the paper’s theme, to me, seems focused on making a case for antitrust, not whether Google is doing a good job of providing search results that make sense for its users– which is a good proxy for whether they are either incompetent (not likely) or taking advantage of their users.
For these reasons, I am skeptical. I am not saying the conclusions are wrong, just that I can think of a lot of reasons for the differences in click-through rates:
— the number of items returned from the search might affect how people react. A large number
might be perceived as “better” as it often is in this culture;
— I can imagine the layout and content of what is returned having an affect on action, maybe there
is enough information on the search page (in the Google alternative) to make that second click
unnecessary.
In closing, if as you say Google isn’t doing a good job, there is lots of money chasing good B2C applications (not to mention well funded entities like Microsoft and Amazon), why isn’t Google failing or at least being challenged in some reasonable way?