Gaming the Cheaters

The Problem

Users of internet video websites trick the system to inflate the number of views recorded for a particular video. They may do this by setting several browser windows to auto-refresh on their video’s web page, or rigging a number of different computers to do the same thing, over and over again. A group of friends may attack together to confuse any IP tracking measures.

Why It Happens

Fame & Fortune. Bragging Rights. 15 minutes of fame. Why else?

The main reason internet video websites are so popular is because people crave attention. We love to show off. “Hey, look what I made!” – “Look, Ma!!! No hands!” (The other reasons are that people love to be entertained and people are naturally curious).

We can divide the cheaters into two non-mutually exclusive groups: the attention whores and the fortune seekers. Their means are the same, their ends are different. I’ve made the distinction because they affect different websites in very different ways.

Services like YouTube, that thrive on visitor stats and video views, have no real incentive (yet) to do anything about the problem. It actually benefits them. So what if there are users on YouTube who claim to watch 70 videos per hour every single day? It only helps YouTube boost their numbers.

Services like Revver or Metacafe, on the other hand, who pay users for their content, do have a problem. Users who cheat the system into believing more people watched their videos, could steal a lot of money from them. For these services, it’s not the number of views that’s important, but the number of real, attention-paying viewers.

The Solution

So, how do we stop the cheating? Turns out there’s several ways. Most of the internet video services require you to press the PLAY button on the video player to start the video. A simple Page View should not be counted as a Video View. Mindlessly refreshing the web page should do nothing to your video’s view count.

Secondly, videos have length. So measure the amount of time people spend on the web page after they’ve pressed the PLAY button. (You can also write some code to track the video being streamed through the Flash player, but this just seems a lot easier).

This simple method will tell you two things in the long run: how many people actually clicked on your video and how long through it did they watch it.

Granted, some people may just click PLAY and just let the video play while they do something else. This is something we’ll have to live with for the time being. Advertisers don’t know whether you pay any attention to their ads on television, but they still pay for them. You adapt to the limits of the technology. (Advertisers also trust AGB Nielsen to have selected a reliable sample of viewers, but that’s another story).

One problem with the simple method (using time spent on the page) is that you can’t use it to reliably track videos embedded on other websites. This is one point in favor of developing a reliable system to track the actual video streamed through the Flash player. Or you can simply restrict your stats to those from within your website and group the rest as “unreliable off-site video views.”

Yes, you can always write a script to open a web page, click on a button, wait some time and do it again. But it’s about making it harder for the cheaters and reporting more accurate stats.

Thus, three easy steps:

  1. Track video views only AFTER the visitor has pressed PLAY on the flash player.
  2. Use the video’s length and the time spent on the page to measure how far through the video they watched.
  3. Filter IP addresses.

What do you think? Any foolproof techniques to stop the cheaters? Leave your comment below.

Technorati Tags: , , , , , , ,

  • interesting problem, except no one is interested in solving it…even revver probably. don’t they only pay if someone clicks through on the ad at the end of the video? they don’t pay for views as far as i understand it – they share the ad revenue.

    so knock up your view stats all you want (though i agree a page serve shouldn’t be counted as a video view), it’s the ad revenue that matters. if your friends are clicking through on ads for you, revver doesn’t care because they’re getting paid to.

    so the only losers are the companies paying for advertising. probably not going to find a lot of sympathy for them…

  • @Christopher, I’m not really sure how Revver actually pays (will look into it). But you’re right in that there seems to be no interest in solving this problem.

    I’m sure the advertisers would like accurate numbers, though. Why aren’t they doing something about it?

  • guess it depends on the model. for tv adverts, sure you want to know how many people are actually watching – at least with click-thrus you don’t care how many people are watching, you only care how many end up at your site.

    beyond that, some advert models only pay when someone clicks through a buy-link and actually purchases (so the compensation to the referrer is more like a finders fee).

    in either case, view metrics don’t matter, except as bragging rights. so the real question is: are view metrics tied to revenue?

    if not, then they’re only good for public appearances – much like gross numbers for film box office. who knows how accurate they are, and really, unless you’re a gross points holder, it’s all about the net – which isn’t readily available information.

  • interesting problem, except no one is interested in solving it…even revver probably. don’t they only pay if someone clicks through on the ad at the end of the video? they don’t pay for views as far as i understand it – they share the ad revenue.

    so knock up your view stats all you want (though i agree a page serve shouldn’t be counted as a video view), it’s the ad revenue that matters. if your friends are clicking through on ads for you, revver doesn’t care because they’re getting paid to.

    so the only losers are the companies paying for advertising. probably not going to find a lot of sympathy for them…

  • @Christopher, I’m not really sure how Revver actually pays (will look into it). But you’re right in that there seems to be no interest in solving this problem.

    I’m sure the advertisers would like accurate numbers, though. Why aren’t they doing something about it?

  • guess it depends on the model. for tv adverts, sure you want to know how many people are actually watching – at least with click-thrus you don’t care how many people are watching, you only care how many end up at your site.

    beyond that, some advert models only pay when someone clicks through a buy-link and actually purchases (so the compensation to the referrer is more like a finders fee).

    in either case, view metrics don’t matter, except as bragging rights. so the real question is: are view metrics tied to revenue?

    if not, then they’re only good for public appearances – much like gross numbers for film box office. who knows how accurate they are, and really, unless you’re a gross points holder, it’s all about the net – which isn’t readily available information.

  • @ Are view metrics tied to revenue?

    Certainly not in the way they are measured now. But they could be very valuable if done correctly.

    If click-thrus are your thing, then I guess it doesn’t hurt to measure views incorrectly… that’s not what you care about. But if we want internet video to generate real revenue and evolve into a serious medium, then something needs to be done.

    Video producers need to know a lot more than “how many people viewed the page hosting their video.” They need to know how many people watched through it and where -if at all- they got turned off by it. The same goes for advertisers doing product placement, and traditional advertisers as well who want to know how their ads are viewed.

    And that’s the important point here… we should be as interested in *how* people experience our content as in *how many* do.

    The tools to make these measurements are already available. Why aren’t we using them?

  • @ Are view metrics tied to revenue?

    Certainly not in the way they are measured now. But they could be very valuable if done correctly.

    If click-thrus are your thing, then I guess it doesn’t hurt to measure views incorrectly… that’s not what you care about. But if we want internet video to generate real revenue and evolve into a serious medium, then something needs to be done.

    Video producers need to know a lot more than “how many people viewed the page hosting their video.” They need to know how many people watched through it and where -if at all- they got turned off by it. The same goes for advertisers doing product placement, and traditional advertisers as well who want to know how their ads are viewed.

    And that’s the important point here… we should be as interested in *how* people experience our content as in *how many* do.

    The tools to make these measurements are already available. Why aren’t we using them?

  • good point – if metrics aren’t tied to revenue, why aren’t they? especially since they could be. it would be helpful to know exactly what people are watching.

    it seems to me (not being in advertising), that in some ways you’re dealing with an industry that is used to not knowing. they’ve never known how many people were really watching their ads – how many are dvr’ing, going to the bathroom, talking on the phone or turning off halfway through their program or simply lying on the nielsen reports.

    so you’re mentioning an extension of the knowingness of advertising. good idea obviously, but not surprising that the industry hasn’t charged forward with it.

  • good point – if metrics aren’t tied to revenue, why aren’t they? especially since they could be. it would be helpful to know exactly what people are watching.

    it seems to me (not being in advertising), that in some ways you’re dealing with an industry that is used to not knowing. they’ve never known how many people were really watching their ads – how many are dvr’ing, going to the bathroom, talking on the phone or turning off halfway through their program or simply lying on the nielsen reports.

    so you’re mentioning an extension of the knowingness of advertising. good idea obviously, but not surprising that the industry hasn’t charged forward with it.

  • Pingback: NBC and CBS Neck-and-Neck on YouTube ratings » Granier on Digital Media Strategy()