that's good point. It would be nice to see someone who does not have a vested interest in under or over reporting the number of bots on the web take a good crack at this number.
Per spider.io, bots are ripping off advertisers left and right. That being said, scared advertisers are the target market for their products.
That doesn't really tell us anything about whether or not they trigger javascript. Bots triggering javascript will inflate advertising and js-based analytics numbers, which would be more problematic.
Bots that don't trigger js, however, are more a drain on bandwidth than anything else.
I seem to remember from a job I used to have as an analyst that advertisers have a few ways to discern human and bot traffic. They don't pay for traffic that is known to be from bots. However there are clearly some bots that aren't accounted for this way, and some that are designed specifically to trigger ads.
That tidy solution only works if the fraud rate on the ads you place is less than or equal to the industry presumed/priced-in average. It seems that there are more than a few ad networks around these days that specialize in non-human traffic. As such, advertiser realized fraud rates can vary greatly from this average.
My company right now won't purchase RTB ads that display outside of walled gardens so that we can 1. manually verify that the ads are showing. 2. Minmize the number of bots seeing the ads (presumably it's hard for bots to infiltrate Facebook or the WSJ than Buzzfeed).
On my blogs I used to use a script that banned bad bots by adding their IP to a list on my .htaccess file. Until it crashed the banning script by adding over 5,000 IPs of bad bots in just 1 month. If my google analytics numbers were correct I'd say between 50% and 70% of my traffic were bots. So 61.5% sounds about right.
I've pretty much stopped looking at "number of impressions" when purchasing advertising too. It's always untrue. A recent blog I was going to advertise on said it had 35,000 unique visitors a month. But yet just a few weeks ago they had a "free give away contest for a jewelry brand (no purchase necessary, international readers allowed, free to apply, no shipping costs)" and only 100 people applied. That pretty much said it right there.
One more example: A site I was on allowed unregistered visitors to give their entire collection of corporate logos a rating between 1 and 10 stars. And they used radio buttons to vote for the numbers 1-10. As soon as you select a radio button it auto-submits and gives the logo that rating. EVERY SINGLE LOGO had a massive number of "1 star" ratings because all the bots would see a radio button, check it, and trigger the vote. All the logos on their site had a rating of 1.
> "free give away contest for a jewelry brand (no purchase necessary, international readers allowed, free to apply, no shipping costs)"
The internet conditioned me to ignore such claims as too-good-to-be-true. I would expect loads of people to treat it like "you're the 1000000th visitor! you won!" and other "free stuff here (if you register and fill out hundreds of forms)" offers.
Na, it was a very personal blog and the owner does things like that all the time. The name of the jeweler was disclosed and everything. I get what you're saying though, but it was legit.
I know monitoring services like ours (NodePing) create a huge amount of traffic by just keeping watch over all those servers... in case someone wants to see that cat video. <grin>
I remember during the heady days of SLIP and PPP dialups, my dad was trying to explain how the future of this new weird invention called the internet would have automated software robots that would perform automated tasks to help up out.
The Internet access cost tends to be higher than advertising cost. You can just compare revenue of AT&T to Google. I pay about $600/year for cable internet, and my advertising value is much less than that. As long as bots pay for the internet access, I think it is just fine.
I wondered how they distinguished bots. The only info I found on the study [1], on a quick search, says:
"For the purpose of this report we observed 1.45 Billion bot visits, which occurred over a 90 day period. The data was collected from a group of 20,000 sites on Incapsula’s network"
So, in the first place, not necessarily representative of the web generally. And it does not say how they detect bots. Of course some declare themselves [2]; otherwise I guess one would have to rely on captchas [3], which are known to be of limited reliability and subject to an arms race.
It's interesting to me that the author links to an article about how coding is useless for journalists, and then spends the rest of the article talking about all the web-scraping scripts she's written or could easily write.
For me, this seems like a success of the whole teach-everyone-to-code thing; the author's picked up a bit of simple scripting and it's made some things dramatically easier for her. It's certainly not professional-level, but that's normal; for example, many people think programmers should know how to write, but few think they should all be able to churn out high-caliber journalism. Is this a failure of setting expectations, or is that really where some people are setting the bar?
29 comments
[ 2.9 ms ] story [ 71.5 ms ] threadPer spider.io, bots are ripping off advertisers left and right. That being said, scared advertisers are the target market for their products.
http://www.spider.io/blog/
That means that bots who declare themselves make about 13.5% of requests to this site. The robots.txt has no restrictions.
I don't know if that's helpful.
Bots that don't trigger js, however, are more a drain on bandwidth than anything else.
My company right now won't purchase RTB ads that display outside of walled gardens so that we can 1. manually verify that the ads are showing. 2. Minmize the number of bots seeing the ads (presumably it's hard for bots to infiltrate Facebook or the WSJ than Buzzfeed).
The report [1] does not include any non-http traffic, which excludes all heavy content like torrent, Netflix, music streaming, etc.
[1] http://www.incapsula.com/the-incapsula-blog/item/820-bot-tra...
I've pretty much stopped looking at "number of impressions" when purchasing advertising too. It's always untrue. A recent blog I was going to advertise on said it had 35,000 unique visitors a month. But yet just a few weeks ago they had a "free give away contest for a jewelry brand (no purchase necessary, international readers allowed, free to apply, no shipping costs)" and only 100 people applied. That pretty much said it right there.
One more example: A site I was on allowed unregistered visitors to give their entire collection of corporate logos a rating between 1 and 10 stars. And they used radio buttons to vote for the numbers 1-10. As soon as you select a radio button it auto-submits and gives the logo that rating. EVERY SINGLE LOGO had a massive number of "1 star" ratings because all the bots would see a radio button, check it, and trigger the vote. All the logos on their site had a rating of 1.
The internet conditioned me to ignore such claims as too-good-to-be-true. I would expect loads of people to treat it like "you're the 1000000th visitor! you won!" and other "free stuff here (if you register and fill out hundreds of forms)" offers.
I had no idea what he was talking about.
"For the purpose of this report we observed 1.45 Billion bot visits, which occurred over a 90 day period. The data was collected from a group of 20,000 sites on Incapsula’s network"
So, in the first place, not necessarily representative of the web generally. And it does not say how they detect bots. Of course some declare themselves [2]; otherwise I guess one would have to rely on captchas [3], which are known to be of limited reliability and subject to an arms race.
1. http://www.incapsula.com/the-incapsula-blog/item/820-bot-tra...
2. kristopolous on this page, '(bot|spider|crawl|aggregator)'
3. which Incapsula does, as a condition of access to its summary of its previous similar research.
I, for one, am glad that human trafficking over the web is down to only 38.5%.
Also, I wonder what the definition for "traffic" is within the context of this research.
For me, this seems like a success of the whole teach-everyone-to-code thing; the author's picked up a bit of simple scripting and it's made some things dramatically easier for her. It's certainly not professional-level, but that's normal; for example, many people think programmers should know how to write, but few think they should all be able to churn out high-caliber journalism. Is this a failure of setting expectations, or is that really where some people are setting the bar?