Thursday, March 29, 2007

My Click Fraud Theory

With the astounding growth of auction-based pay-per-click (PPC) advertising led by Google and Yahoo over the past several years, there is a great deal of discussion about click fraud. I think most people are completely missing the point. Those who are complaining about the issue really have nothing to complain about. Bizarrely, those who are getting screwed haven’t even uttered a peep.

Click fraud takes place when someone clicks on a PPC ad with no real intent of actually following the link. The result, of course, is that an advertiser has to pay for a bogus click. Many advertisers are complaining about being charged for bogus clicks and are even demanding refunds. Somehow they managed to force Yahoo to settle a lawsuit for about $5 million and Google actually coughed up $90 million in a similar instance.

This baffles me because any reasonably sophisticated advertiser shouldn’t really care about click fraud. Why? Because if they’re using any of the dozens of off-the-shelf tools to measure their PPC ad conversion rates, they would be automatically lowering their PPC bids to maintain an acceptable ROI. An advertiser simply shouldn’t care if he pays $1 per click for 10 clicks that yield 10% conversion and a single customer or if he pays $0.50 per click for 20 clicks that yield 5% conversion and a single customer. In either case, the advertiser has paid $10 to land a new customer.

Given the real-time tracking capabilities inherent in auction-based PPC, click fraud gets priced into the equation automatically. In my example above, the bogus clicks drove conversion down by 50% (from 10% to 5%), and advertisers adjusted their bids from $1 to $0.50 per click. Yes, in theory the advertiser may suffer temporarily from fraudulent clicks before he has a chance to adjust his bid downward to compensate for the lower conversion rate, but today’s automated systems figure this out pretty quickly. Any real damage to the advertiser is inconsequential.

So why are advertisers complaining? I don’t get it. Perhaps they haven’t all figured out how easy it is to use tools from search engine marketing (SEM) experts such as Efficient Frontier, SearchRev, iCrossing, iProspect (or many, many others). These SEM experts offer reasonably cheap software to solve the problem through automated bidding.

Perhaps some advertisers believe only their ads are being clicked fraudulently. Click fraud targeting a specific advertiser, does, in fact, hurt that advertiser. He can reduce his bid to maintain his conversion rate and ROI, but now he will get a much lower number of clicks because his competitors can afford to bid higher if they aren't suffering from the same click fraud. Though I acknowledge that advertiser-targeted click fraud is possible, I believe most click fraudsters are just trying to make money for a certain publisher rather than trying to deplete the ad budget of a certain competitor. So, I'm at a loss to explain why advertisers seem to care so much about click fraud.

Even more confounding, however, is that the companies who are really getting screwed haven't started screaming about it. In fact, we haven't heard anything from them. The companies on the losing end of click fraud are high quality web site publishers. When an unscrupulous publisher engages in click fraud to increase revenues, the result is reduced conversion rates for advertisers, who, naturally, lower their bids. Because of the way most advertisers participate in Google and Yahoo PPC auctions, when they reduce their bids, the reduction applies to every publisher in the Google and Yahoo networks.

If publisher A is engaging in click fraud, which causes lower PPC bids from advertisers, publisher B gets screwed. Publisher B doesn't generate any additional clicks, but now he's getting less revenue for each click. In a sense, publisher A just stole money from publisher B. My theory is that the publisher getting screwed the most is probably AOL.

I suppose it's not too surprising that high quality publishers aren't complaining because it's virtually impossible to detect this phenomenon. This is especially true because Google provides publishers with such a minimal amount of information about their advertising performance. Perhaps one reason Google is holding on so tightly to the lack of transparency in its system relates to keeping click fraud off the agenda of high quality publishers.

Sunday, March 25, 2007

My Simple Multiple Choice Test For Consumer Internet Startups

Anyone thinking about starting or joining a consumer Internet startup should be able to answer the following question, and if the answer is "d) none of the above," then I'd suggest looking for a new job.

So, here's the question: does your company attract users with any of these things?
  1. an inherently viral idea
  2. search engine optimized content that grows naturally with usage
  3. the ability to spend money to acquire users with a very fast payback
  4. none of the above
Why do I think those three things are critical? Because without at least one of them, you're building a Field of Dreams. Though it worked for Ray Kinsella (played by Kevin Costner), it simply doesn't work for Internet entrepreneurs. I have met too many people under the delusion that if they build it, people will come. With tens on millions of web sites competing for eyeballs, it just doesn't work that way online.

There's also a group of startups who think they have passed the multiple choice test, but that's because they have misunderstood the answers. This blog post is my attempt to clarify them.

1. An inherently viral idea
Very, very few sites can correctly offer this answer because very few concepts are inherently viral. You may have a great site and may rightfully think users will tell their friends about it, but you probably don't have a viral site. You're not benefiting from a viral idea, you're benefiting from word of mouth. Word of mouth is nice, but it's not nice enough to form the foundation of a business. To be viral, the utility of your site/service needs to grow substantially as your friends use the service. Great examples of inherently viral services include Skype (with no friends to call, it's useless), Facebook (you can't trade messages with your friends if they're not on the site) and LinkedIn (you can't access your business network if you haven't established it on the site). Bessemer had one fantastic investment success with Skype and is hoping for a second with LinkedIn. Too bad we missed out on Facebook.

Viral ideas are the most powerful of all because they grow exponentially. Skype had 500,000 users less than a month after beta launch in the summer of 2003 and has reached more than 100,000,000 people since.

2. Search engine optimized content that grows naturally with usage
Almost everything starts with Google and its search engine brethren these days. If you can build a service with lots of web pages that get indexed by Google and show up naturally as top results for a large number of searches, you will get a lot of free traffic. It sounds simple. But it's not.

There are dozens of techniques and tricks to building search engine optimized (SEO) web sites, and there are countless firms who will sell you advice on how to do it correctly. Most importantly, though, the idea underlying your site must lend itself naturally to SEO. First, consumers have to be doing a lot of searches for content on your site. Second, your content has to grow -- ideally through contributions from users. Bessemer had a very successful investment in Site Advisor (acquired by McAfee) that leveraged SEO distribution. Several of our current investments -- OLX, Wikia and Yelp are SEO-driven.

3. The ability to spend money to acquire users with a very fast payback
If you can spend money to make money, you can control your own destiny. Generally, you have to be in the business of selling something to consumers. Online retailers are great examples of companies in this category. If 2% of visitors complete a purchase and if the average purchase is a $50 with a 30% gross margin, then you can afford to spend $0.30 for a visitor. Assuming a small fraction of the eventual purchasers will become repeat buyers, you will have a profitable business. Retailers and subscription sites like Improvement Direct and Match are great examples. In fact, most of the IAC properties fit in this category. Online media properties that monetize through advertising can rarely spend money to make money. It just costs too much to acquire a customer.

There are a few companies like LinkedIn and Yelp that benefit from more than one of the user acquisition techniques mentioned above. LinkedIn is primarily viral but leverages SEO through its personal profile pages. Yelp is primarily SEO driven but has a viral element among its core contributors who invite friends to the service to share and compare reviews.

There may be one other compelling customer acquisition strategy for consumer Internet companies, but in my mind, the jury is still out. This fourth technique leverages the open APIs for widgets offered by MySpace and many of the leading social networks. Although it's clear you can drive tremendous usage through widget-based distribution, I'm not yet convinced that these widget companies will develop viable business models.