Please Stop Working on Ads

It’s 2 pm. Alice goes to Facebook. Bob goes to Amazon. Charlie goes to Twitter. If you have to bet on one of them having spent money online by 3 pm, who would you pick?

Without any further information, Bob would be a no-brainer. Amazon is  an e-commerce sites, so at the very least Bob is “window shopping.” Who knows what Alice and Charlie are up to on Facebook or Twitter? This brings me to the point I want to make about these two companies. Facebook and Twitter are two examples of a model that’s been with us since the beginning of the commercial web, which is to:

  • create a site
  • attract many eyeballs as quickly as possible
  • figure out how to make money later

In other words, Twitter and Facebook (and Google, for that matter) did not start as businesses. They were experimental web services that became popular quickly, and once they had traction they had investors knocking on their doors. These investors were confident that they would find a business model (or be acquired by someone who already had one). Hotmail and Geocities were also examples of this, just to name a couple.

As it turned out, the only way online businesses with millions of users and thousands of employees were able to generate enough revenue to satisfy their investors was through advertisement. Google was an outlier among outliers: the #1 search box of the world will necessarily capture a large absolute number of searches for stuff to buy. If this number is large enough (which it is), the rest of the searches really don’t need to be monetized.

Facebook and Twitter are different. Not many people think: “I want to buy a Veeblefetzer. I’m going to Twitter / Facebook to start my research.” We go there to consume (and maybe produce) information. Because we happen to be paying attention to a screen it is possible to shove ads in our faces.

The implication of this is that Twitter and Facebook are forced to employ one or both of these strategies:

1) Mine the hell out of your behavior to guess if you may be interested in something that one of their advertisers is selling (preferably right now). You went there to see pictures of your second cousin’s new cat or something, but perhaps they can figure out that you’re desperate to buy some tickets to Hawaii. You are not telling Facebook what you want, so they are at a disadvantage. Google knows what you asked for, so it can do a better job at matching sellers and buyers. Therefore, Google can be more cost-effective. Knowing who you are and what you like is not enough for Facebook and Twitter. What if you are a 27-year old baseball fan who likes horror movies, but right now you are in dire need of a DUI lawyer? You could be a Republican soccer mom who drives an SUV, but secretly you are into S&M and swinger clubs (it’s between you and Google). You get the point.

2) Tempt you to buy shit you don’t need. That’s the time-tested TV model. This entails annoying most people who are there simply for entertainment purposes. The worst part about this is that the most desirable consumers are the first ones who will flee to an ad-free alternative if/when if presents itself (e.g. HBO, Netflix, iTunes).

There is another problem with becoming an ad-based company. The collective purchasing power of the planet is growing more slowly than the advertising space. In other words, Twitter, Facebook, and all other advertising-based business are waging a war for our attention. And if there was ever an unnecessary war, this is it; because of search engines, advertising is less necessary than ever.

Let’s imagine that advertising were made illegal tomorrow. Obviously search engines could not make money through ads, so they would start charging. Most of us would pay, like we pay for our cellphone plans (I couldn’t find any free cellphone plans that forced you to listen to ads). Because of search engines, people would still be able to find and buy the products they want or need. There may be less buying of stuff that we don’t need, especially what’s driven by impulse. Like the chocolate cake that seduces you when you open the fridge looking for a piece of fruit. In my book that’s a good thing. Another plus is that some industries would even save money by not needing to advertise, because advertising is an instance of the Prisioner’s Dilemma for many products.

TL;DR: Please don’t start a company that would need ads to make money. Please stop working on ads. Pick an unsolved problem instead. Not only it’s potentially better for the world, but also you’re more likely to raise funding and cash out. Oh, and if you’re an investor, don’t touch ad-based stuff. If you do, you’ll be very uncool 🙂

This post was brought to you by Happy Fun Ball.

Discussion on Hacker News

The Socialist Republic of Twitter

I attended a Machine Learning meetup at Twitter yesterday. It was a very good event, and I learned a few new things. The presentation about ads by Kumar Chellapilla was particularly interesting, and one point stuck with me. In a nutshell, Kumar explained that:

  • Twitter wants to remain freely accessible to everyone in the world.
  • The only known way to accomplish this is by monetizing the service with ads.

Let’s think about the implications of this. Because we live in a world where the distribution of income is highly skewed, a small number of users have most of the dinero. These users are the ones who keep ad-supported businesses alive. For the sake of simplicity, let’s assume ads do a good job at matching consumers and advertisers, and that everyone is equally permeable to them. Oversimplifying a bit, you could say that the small minority of Twitter’s users with the highest income are subsidizing the poor masses: the people who wouldn’t buy products or services even if they clicked on ads. In other words, my dear readers, Twitter (or Facebook, for that matter) is a socialized service provided globally by a corporation.

Che Tweetara

Is this bad? No, I think it’s awesome. If you are a teenager from a poor country with lots of interesting things to say, a bunch of fat Americans and Europeans are clicking on ads so that you can have a voice. Be thankful for that girl your age who clicked on #JustinBieberLuv4Ever or something, and ended up buying a concert ticket.

What is the problem, then? I’m glad I asked. The problem is that corporations respond to their shareholders, so most of us don’t have a say in their policy. This is a socialist plutocracy (!) If for some reason a month from now Twitter discovers that it can make a fortune by somehow charging for the service, it will happen. You and the richest Twitterer in the world may pay the same monthly fee. He may earn it in a second, it may take you two hours. You’ll tweet about uprisings in Egypt while she shares pictures of Louis Vuitton’s finest products. Same with censorship; if Europe threatens to block Twitter unless they filter out the word Greece, Twitter will probably do it (and probably Italy too, for good measure). And, worst of all, what if Twitter discovers that it simply cannot sustain its business with ads? Will they appeal to the US government for a subsidy? Even if they do, it won’t happen: they are not a public utility, and they are not Goldman Sachs either.

I’m not saying Twitter is going to fail. It will evolve. It may become a very profitable business. However, I believe that it won’t be the democratic, globally accessible, free-speech platform it is today for much longer. Enjoy it while it lasts.

Discuss on Hacker News.

Silicon Valley != The Internet (The CEO of Yammer says the darndest things)

This weekend David Sacks (CEO of Yammer, sold to Microsoft for $1.2B) made a statement that I consider extremely shortsighted:

I believe his judgment is affected by the “I sold my company, so long suckers” syndrome. The main flaw in his reasoning seems to have escaped investors like George Zachary and Marc Andreessen (dude, I always have to google your last name). To put it in a few words:

Silicon Valley and the internet industry are not one and the same.

I’m not going to make a long argument about this, because it’s plain and obvious. You could have made the exact same argument in 1995 but used the word “computer” instead of internet. It seems that Google and Facebook have made people forget that Silicon Valley has always been about solving problems with technology, and the technology has always been in flux. Yes, if you build a new Instagram it probably won’t become a “giant”: it will be acquired by Facebook for one billion seven hundred million dollars. In what amounts to journalistic surrealism, pundits will speculate about how they could have cut a better deal.

In the meantime there are tons of technology problems not making headlines. Healthcare for one hasn’t advanced that much in the past 50 years. Even in the first world people are dying of stupid, preventable diseases. Some of them could be monitored and controlled with cheap smartphone peripherals plus data analysis. Transportation is in many ways worse than before 9/11. We fly in planes from the 1960s, but the trip takes longer than it used to: security, slower speeds because of fuel costs. These are just two problems, there are countless more. Of course the internet is a tool to solve them, just like computers and calculators. However, the companies tackling them are NOT internet companies like Google and Facebook. Sure, Google is working on self-driving cars. Still, there are 99 other problems they can’t even begin to think about. There are Silicon Valley startups working on these problems. I know a few. Ten years from now there will be multi-billion dollar, non-internet-centered Silicon Valley companies that you haven’t heard of yet.

TL;DR: David Sacks, go put together a vintage car collection and stop making a fool of yourself on Facebook.

Hacker News discussion of this post.

Facebook and Twitter Don’t Know What You Want

Facebook and Twitter, after years of searching for the I-wanna-be-like-Google business model, have settled on ads. The oldest business model on the web, and perhaps the default. There is one pesky problem though: online ads are about matching intent: Alice sells Veeblefetzers. Bob wants to buy a Veeblefetzer today, and you know it. You put Alice and Bob in touch, and take a cut. Now suppose Carol comes to your site, and she has no interest whatsoever in Veeblefetzers. If you don’t know what Carol wants during a specific session, that session is not worth much.

What do Google, Ebay and Amazon have in common? You go there to ask for something. You type in “medical insurance”, “fake plastic trees”, “barack obama figurines.” These sites have something for you. Of course, Google has huge amounts of queries for which there’s nothing you’d want to buy. It doesn’t matter; the relatively small number of queries with intent to buy is large enough for them to make a fortune.

In contrast, you go to Facebook and Twitter to browse, post, interact. Not to shop. These sites rely on the data they collect about you to show ads based on your behavior. There are several problems with this approach. One of them is that they don’t know who we really are. People’s behavior on Facebook and Twitter is meant to be seen by others. We have an incentive to “Like” things that we don’t necessarily like in real life, or viceversa. Say, would you like this?

Still, suppose you hired the best private investigator to find out everything about me, and he delivered a thick file including details such as my favorite brand of underwear (by the way, I have enough of them to last me twenty years so I’m not buying any in the foreseeable future). The million dollar question is: how would you know what I want at this exact moment in time? Remington Steele or Maddie Hayes wouldn’t know much about my conversations with Google unless they had access to Google’s treasure chest of data, or to my full browsing history. Even then, Google has little idea of what I’ll ask next. I could be looking for a climbing gym in Houston, a music video, a dentist, etc. They will have to wait and find out. And find out they will.

Google guesses what you want.

With all this in mind, it’s no surprise that Facebook’s click rates are dismal. Perhaps things will be better in the mobile world: both Facebook and Twitter claim to have much better click rates on mobile. However, I suspect that the rate of accidental clicks on mobile is also much higher than on the desktop. No word on the conversion rates of those clicks, which is ultimately what advertisers are seeking.

Why is this a problem for Facebook and Twitter? Because they have to sustain insanely high valuations demanded by their investors. If they try to do it with ads, they will be compared against Google and almost certainly look like losers. Maybe they will figure out a new business model one day, but I’m not betting on it.

Now come follow me on Twitter. Let’s enjoy their fine service and continue to pay them nothing! 🙂

Discuss on Hacker News if you please.

Enforcing Different Passwords for Different Sites

Over the past few months there has been significant discussion about what makes a password secure (relevant xkcd comic that you’ve probably seen already). One factor that’s often overlooked is the fact that people tend to use the same password across different sites. This is perhaps a more a serious problem than the cryptographic strength of a password itself. Why?

Imagine that your passphrase is “steward fan catholic impersonator.” You feel confident that it would be hard to guess it through brute force. You use it for Paypal, Facebook, … and also for Now, one fine day gets hacked. The next time you visit, the web page has malicious code that sends your password in plaintext to someone. There go your Paypal funds, your Facebook account, your online life.

It doesn’t matter how hard we try to educate people to not reuse passwords. It’s the path of least resistance for most people, so they (we) will continue doing it. How do we solve this problem?

One possibility would be to enforce it from the browser. Google could make Chrome remember one-way hashes of passwords you’ve used. When you try to use “steward fan catholic impersonator” for the second time, Chrome would recognize the hash and tell you “sorry Dave, I can’t let you do that. You’re already using that password somewhere else.”

Of course, we can’t wait for Google to implement this. On your own site, you could do the following:

  • When a user is signing up, generate a random string.
  • Tell the user: your password must contain the following word: “hzru”

Obviously this is not perfect: the user could start using this password (say it’s IHateHaving_hzru_InMyPassword) for other sites. But if enough sites had a similar policy, the risk of reuse would decrease.

This post was inspired by the fact that an acquaintance of mine woke up to an empty Paypal account today, and he has no idea how it happened. To be fair I don’t know if shared the password with other sites, but this must certainly happen every day to some people who do. Nobody should have to freak out because a password leak at a site they rarely use could cause them financial harm.

Do you have better ideas? Let’s hear them.

Edit: few people seem to be getting the point of this post. Of course you could use a password manager. However, most people won’t unless they are forced to. How to enforce something like that for the masses is the point of this post.

Discussion of this post on Hacker News. 

The Dangers of Being a Product Instead of a Customer

Gmail is almost certainly the most important web service I use. It has become the master key for many other tools and services. The scary part is that Google could close my account any given day, and I’d have no recourse because I’m not a paying customer.

Twitter is a service that I use perhaps too much. I have a relatively small but stable audience accumulated over the years. I broadcast a question or a thought whenever I feel like it, and it has become one of the main channels I use to exercise free speech. Again, Twitter could close my account on a whim because they owe me nothing.

Facebook would also fall into this category if I cared, it just happens that I don’t.

So Long Posterous, and Thanks for All the Bits

If you follow this blog you may notice that I’ve migrated it from Posterous to my self-hosted installation of WordPress. I loved Posterous; I thought it was simple, pretty and convenient. However, the quality of the service had degraded noticeably after the Twitter acquisition (e.g. I couldn’t even reply to comments on my posts the last few times I tried). Still, it wasn’t that bad. I could have kept using it while waiting for an easy way to migrate my posts. Twitter promised that “over the coming weeks we’ll provide you with specific instructions for exporting your content to other services.” And it’s only been 21 weeks…

Interestingly, what prompted me to move was a terse email from Twitter asking me to remove a small dataset of public data that I’d shared. There was nothing wrong with the email itself, all it did was point out their terms of service. However, if I had written that email I could have made it much more friendly, as I understand the value of PR (especially with developers and bloggers). It gave me a bad feeling about the inner workings of Twitter as a company. I wondered if one day Twitter might kill my Posterous blog just because, and I figured that I could easily host it myself. Unfortunately there is no alternative to the Twitter service itself (I’ve backed but I don’t expect it to be even remotely equivalent).

My next step will be to associate my other services  with an email address at my own domain instead of I may still use Gmail behind the scenes, but I can always take my domain somewhere else. That is, assuming I don’t use *that* email address as the key to my registrar 🙂

For similar reasons I don’t contribute to Quora very often. The 15′ that I may spend on a question gives them much more value than I receive in return. As an aside, who knows how long they’ll be around for, or if they’ll ever come up with a business model. There is even a topic on Quora itself about this. How meta…

I’d much rather be a customer of web services than a product. I would pay a reasonable amount of money for a few things, especially if this removed undesired ads (I don’t care that much about that, because like most people I’ve learned to ignore the vast majority of them).

The problem is that there are not enough people like me. Most people don’t care enough about these issues to pay for the cost of the services they use (as opposed to being “product” and having no rights). There is also a perverse incentive: a service with hundreds of millions of free users is rewarded by the stock market or private investors because in theory they could find a way to sell something to all those “people.” For example, Facebook’s valuation is not justified by its current business model based on display ads. Even at today’s stock price, their price-to-earnings ratio is much higher than that of comparable companies with similar growth rates (see GOOG, YHOO).

Finally, one subtle point that bothers me about the current state of affairs is not that we are just product. That was already the case in the age of TV. Things have gone one step further; we now generate the very content that digital entertainment companies sell to advertisers and feed back to us. In other words, we are Digital Soylent Green.


Using Foursquare to Detect Tax Evasion / Money Laundering

I wonder if there are governments or private parties already doing it. The idea is pretty simple:

The Foursquare API allows anyone to keep track of the number of check-ins to date for any venue. For example, the Eiffel Tower (Foursquare login required) has 36429 checkins from 26358 users as I write this.

For fairly popular venues, you could correlate the daily / yearly Foursquare activity with the reported revenue. Of course you’d have to do some normalization: Foursquare’s popularity grows over time, and obviously it has different penetration rates across countries / regions. This shouldn’t be a big deal.

For venues  within a very specific category (e.g. to-go coffee shops or McDonald’s franchises), there should be a very high correlation between revenue and check-ins. If one venue reports significantly less revenue than their check-ins would predict, that’s a red flag for an audit. Conversely, if a shop reports relatively high revenue numbers but has few check-ins for its class, there’s a chance it may be laundering money. This alone is far from an indictment (e.g. mobile reception could be terrible in the area so patrons wouldn’t check in very often), but it’s a cheap starting point to investigate more.

Of course this is not limited to Foursquare, you could aggregate check-in data from other platforms (Facebook, Yelp, etc.) for a more robust signal.

You’re welcome, IRS 🙂

The “Taken Username” Effect on Twitter

It happens all the time to many of us. We try to sign up for a site, and choose a username. We’re late to the party so our preferred one is taken. We may add an underscore, or append the last two digits of the year of our birth at the end.

On Twitter having a short username is a plus: @replies from other people leave more space for what they want to say. Here is some data from my Twitter sample that seems to show the “taken username” effect over time.

Average username length for accounts created in a given year:

2007 8.83
2008 9.38
2009 10.00
2010 10.27
2011 10.68
2012 10.97

Average number of non-letter characters in a username:

2007 0.58
2008 0.67
2009 0.92
2010 1.10
2011 1.14
2012 1.38

Also, the average number of underscores in names created within a year doubled during that period (0.086 to 0.174). Same for usernames that end in a digit (0.17 to 0.35).

Moral of the story: maybe arbitrary usernames will become a thing of the past if universal logins like Facebook Connect become the norm. In the meantime, shorter names will continue to look “cooler.” There will still be land rushes like what happened this week at I secured my preferred username just in case, did you? Do you care? 🙂

Relevant SNL skit:

Discuss this post on Hacker News if you’d like.