Home > Paid Traffic Sources > Facebook & Instagram

---

50% of clicks are BOTS [dafuq?] [GTFIH if you run FB] (26)

---

The last few weeks have been odd.

This week:

• FB shows 310,000 clicks paid for.
• Prosper shows 133,000 "real clicks".
• Prosper shows 160,000 " filtered clicks".
• ... other clicks: cloaker bleed rate (3.5%), and visitors who do not have JavaScript? It shouldn't be more than 1%. (20% of my clicks come from Germany, whose government agencies oficially recommend having "noscript" installed... I know, I should upgrade to the STM Prosper which uses cookies/variables instead of JS, but the 1-2% is not a priority, it's definitely not more than 1.5% at most for DE.)

Other notes:

• Real/fake click ratio is similar in all my 15+ geos, Germany being slightly higher than the rest.
• I can't compare the real/filtered ratio to times past, because I regularly wipe my P202 database. I remember it being 25% though (I have a gmail trail that shows me checking the stat and being frustrated. Now it's twice as bad!)
• I'm sure you're all aware of the reports a year or two ago by an advertising agency saying 80% of FB clicks might be bots... Well it's never been higher for me then now, and I run so many geos.
• I know Prosper cannot work without visitor JS installed, but any bot worth its salt will use JS to appear real... There is no way 50% of FB clicks are bots.
• I don't think there are so many accounts clicking ads for bot purposes... Wtf? Wouldn't clicking lots of ads multiple times make FB give you the boot if they suspect you're not real? (Actually I just had a thought: for making personal FB accounts, clicking like 10 ads a day is probably going to let your account stay active more easily, as you're worth $50-100 per month to FB if they keep you... With a suspected 83 million fake FB accounts, it would make sense for FB to keep fake and ad-engaged accounts active *X-files music*)
• I know the clicks aren't FB-commissioned fake clicks. Either these are bots, or real users that Prosper is fucking up...
• Nearly half of the filtered clicks are unique.

QUESTIONS

1. From my understanding, Prosper has a list of bot UA strings or whatever. Which file is this located in? Does Prosper consider somebody a bot if they visit the page and leave before the Prosper JS loads fully? (How can it even know about the visitor unless the JS loads... these filtered visits must all have had JS enabled, there's no way Prosper can know bounce rate, right?)

2. What does Prosper do in case of a suspected bot click? It still generates subids, but does it still display the webpage and stuff? Out of all of my filtered clicks, 3% have a "campaign" value set (but never any conversions)... meaning they would have clicked through to the offer, right? The other 97% did not click through...

3. I suspect many of the filtered clicks are real users and not being served the landing page. Has anybody had similar experiences and how did you solve this problem? I have just checked the useragent strings from my server/cloaker log, for filtered clicks that were unique IPs in my Prosper database, and they all have normal-looking UAs. How can I find out why Prosper filtered them?

4. (Not a question) If these ARE bots, it's nearly impossible to avoid them in targeting (except with exclusion targeting: geo-check the city for all, and see which city has the worst ratio of filtered vs real clicks). Generally fake accounts will "like" hundreds of the biggest pages. There is no escaping them, is there? I WISH FB would allow us to only show ads to "users liking <100 Pages/groups". How hard is it, FB? If a FB SRM rep is reading this, please get this done. Thanks!


Any thoughts appreciated! Sorry for the long post. It's just very frustrating. I don't care if bots see my landers or not. I don't appear on spy tools, nor on search engines, so all visitors would have come through FB ads. I want to know why half are "filtered" and have not produced any conversions. If nothing changes in the next week or so, I will remove Prosper and see how I go without it in terms of overall profit.

There HAS to be a way to turn the filtering off in Prosper - can I get an expert to weigh in?

If this is a FB ad quality issue, I'd love to fix it. Perhaps bidding for clicks is bad? Since these bots obviously click a shitload of ads, I may have to start bidding for impressions/reach instead of clicks. I know people that see my ad will click, whether they are regular clickers or not... Damn this FB complexity.

---

Also I'm seeing other threads talking about traffic quality dropping. Can you guys give me your percentage of real/filtered clicks for FB? Let's pool some stats and figure this out together.

---

If you are on p202 1.7.2 the bot detection is very basic and done via ip ranges. Keep in mind version 1.7.2 is 3 years old, so there are missing prances and the current ranges are most likely out of date. Clicks are filtered out for in the following situation

1. If the same ip hits a link in a 24hr span
2. If the ip is in the ip range of a known bot

P202 generates a subid for every click, pages are loaded etc. To a user nothing is different. The only thing different is that it will flag the click and show a different icon in spy view. This is mainly for you to see and know that you are not getting unique clicks. Also remember that all ad networks have their own idea of what's a repeated click.

Yes they could be real users, and if they are their conversion will track fine in Prosper202 when it happens.

I haven't heard of anyway to avoid this if you are getting botclicks from fb. The countries/geos you target could make a difference.
The whole idea behind filtered clicks in Prosper202 is to show you that the same people(ips) are clicking your ads in a short period of time(24hrs). As I said earlier, in terms of tracking etc, there's no difference between a filtered click and a real click except the icon in spy view.

50% is a high number, so I'd suspect there could be something weird going on here.
BTW: Have you done lookups on the ips to make sure they are from isps and not servers(aws, rackspace etc)

---

Nana, thanks for the great response! That cleared up a lot of things for me.

All of the duplicate clicks I checked had hostnames from residential ISPs in my geos.

I agree 50% is very high. Many people click 10 and even 15 times... I may have to try oCPM (optimising for Reach, not Clicks) or straight CPM... right now FB is delivering my ad to clickers... But they are fanatical with their clicking.

I might look into modifying the P202 filtered click formula to say, 3 clicks in a 24-hour period is legitimate... I sometimes expect dual visits within the day, and have a cookie set up (kind of like, if user already visited same day, display alternative landing page/offer). Know in which file I can modify the logic?

---

Facebook will introduce Frequency capping soon - this will help solve quality issues.
http://www.insidefacebook.com/2014/0...campaign-type/

In the meantime I'll be switching bid type, amounts etc. to effect lower frequency (and more unique reach) in my campaigns. Even on day 1, the number of dupe clicks is 60% for a new account's campaign launched today... not cool.

---

50% is really that bad...

Can we have an Facebook Employee comment on this problem here please?

---

I think for mobile FB ads the bots rate should drop

---

It does, but then you come into different issues entirely.

I really think FB has no idea what it's doing or how to combat this issue. It seems like they try some new shit every week or two, fuck a bunch of shit up and see a sharp decline in revenue and revert back and then just try something else until they find something sticks.

Seeing the most sporadic stuff from them constantly and none of it makes sense. I really hope they get a terrible quater with a decline just once, and their stock drops hard so they can start reverting back to stuff where revenue is coming in fast and furious and stuff is working for DR marketers.

/rant

---

Probably one of the most salient features FB will have ever rolled out.

---

wow cool article. That will be a great feature. I hope it comes out soon !
I've been trying some experiments to lower frequency but i can't get my finger on it.
can't wait for this ! Thanks for posting redrummr

---

Look.... There are some groups running spy tools on Facebook. They are scraping tons of data from Facebook. They click ads and you pay for them according to FB rules. Even if you block them, you will still pay for those clicks.

---

Issue has been resolved, things are back to normal. Thread can be closed.

---

Spy tools don't click ads. They can get the destination from the URL without clicking the ad link.

---

On Facebook perhaps, as the advert click URLs do contain the ad destination URL which makes it easy to extract them. Not necessarily the case on other traffic sources.

---

^ They do click the ads, and SocialAdNinja/LotsOfAds will display the entire chain of redirects.

However the spy tools don't see my ads, much less click them thousands of times per day! It was never the spy tools. My issue has now been resolved.

Even if the spy tools upgrade their processes to fix the 2-3 easy ways you can block your ads from appearing on them, there is still one very simple way:
1. random unique click ID generated by PHP and placed on final LP URL
2. check your ads on the spy tool - you will be able to see the click ID on the final lander
3. check your server logs for the click ID, get the IP address, find out the city of the spy tools user, exclude that city from your advertising

Usually the IP will not say "Sydney" but will say something like "Blacktown" (within Sydney) - so you don't lose too much volume. This is the worst way to block them, but is a good last resort.

The spy tools owners are lazy, use the same servers etc. and I will probably let the cat out of the bag on them in the next few weeks, so nobody on STM needs to appear there anymore. Right now they aren't very useful, most people aren't appearing on them anymore. Once you are off them, it allows you to be truly creative, knowing it's so much harder to spy/swipe your shit. To be honest it would make things a lot more interesting. When you get creative is when you make bank.

---

Depends on which Spy tool you taking about particualary. Some tools i know click ads and crawl through all redirects. Just my observation.

---

Been using exact trick. You can only hide destination/display links . You can't hide your ads from Spy tools.

---

With FB and Google, both will provide the destination in the url so you don't have to trigger a click. I know this 100% since we built our own also.

You will see redirect chains, but those are the chains of the link they provide in the URL, not from what they click and get.

The lots of ads guy is on here I think, and I'm sure he can verify this.

Clicking ads 100% or even 50% of the time is not sustainable and it'll flag you straight away. It'll cause headaches for them and the advertisers and that's why they started putting the destination in the urls.

Here are two examples (edited the strings):

https://www.facebook.com/a.php?u=http%3A%2F%2Fdimitri.co.uk%2Fbeauty%2Fbeauty.htm l&__tn__=%dgB&xt=2.AQJxrc6_ijnAACNZNdSVm7uTHRasdgas dgasdgasdgsdg4i_wM6VlhbJM3TithLPg2_Oxy-gteZ5phSOxHiBDtdCasdgasdgvZ8PmY0xu0T_GSRYpz3g6DnEj v--zmGi_icqRklc7PUFkNCaWrw5hw5W83LN12LcgB_HGRtXWxy_qH EJY-W3IBouz6p511bxAK-lLuxOzWy1ytZLuIUq4FuK2I3AA8IiruccJNsCYq-MSVY-ELMlYnT19yS7otgRIhb-K7DAVXE7z6ZfH1BB8ZZzKVBQZ-K4iFNOzrvw2ra5GEKcmePk7cx5asdgasdgasdgKGjiGStYuPnM cGY83x9VAaIITalrKQZ_CT_LBQT497CzY1rvvCKQ9nx3wXLGdO lNREMZmu-q-rM45y5_YbTC3Ai_vA5Jw9PTL1u11G7baLZKpxtdBNDeZE4PXOn 9dXv8wloDC3jF1T4uI6Cn3RlGyja75AON4fe9-A1Os5Hs5D8D&mac=AQIGAcDwj6yRqjXy&sig=129780

http://www.google.co.uk/aclk?sa=l&ai...s&ved=0CC8Q0Qw&adurl=http://clickserve.dartsearch.net/link/click%3Flid%3D43700000988571109%26ds_s_kwgid%3D587 00000119948475%26ds_e_adid%3D30097204837%26ds_e_ma tchtype%3Dsearch%26ds_url_v%3D2

---

How did you resolve the issue?

---

I am having same issue. Somebody help me ((((

---

Buddy. SAN crawls link by clicking it. Not by looking at destination urls chained. No need to argue. Ask the masters and keep quite. I also built scraper.

Someone please post some unbiased opinion.

---

---

David, Janglib is right - the destination URL is in the ad's link - you don't need to click it to see it, you can hover over the ad (or right-click and copy its URL). This means scrapers don't need to click links from Facebook, they parse the URL and can click separately. Your argument of having your own scraper means that yours just isn't as smart as SocialAdNinja and others would be (it probably makes sense to not click 500 ads per day).

Janglib, the elementary method I described still works, even if SAN is not clicking the link from FB.com...

AD URL: hello.com/page.php
// page.php generates a random code and sets the server variable 'goodbye']
AD URL redirects to: hello.com/?goodbye=77014
// Spy tools will show this unique URL (as a link in the redirect chain)
// Redrummr inspects SocialAdNinja and sees that the ad's redirect chain includes the 77014 number
// server log will reveal the IP address that accessed that particular (unique) URL
// then geo-check the IP, exclude the town from future FB targeting

Like I said this isn't the best method, it's a bit manual, but in 2014 I don't think spy tools should be a concern for anybody. After all, as others in this forum have pointed out, just having the ads is not enough. But sometimes, you don't want to inspire other affiliates with footprints of your massive scale operation, or with your unique landers and cloaking angles etc.

---

Some misunderstanding dude.

My
Crawler locates desination link in ad.
Then it stores it in DB.
Then other crawler crawls through this link, stores every redirect link in DB.

My scraper only have 5,000+ Fb accounts. I am thinking about selling subscription.

http://i.imgur.com/fBRmjbZ.png


Yes, i agree cralwer is not clicking links like browser do. It just extracts the destination link and visits it later. Its similar to copying ad's link then visiting it in inconginto mode in chrome.

---

But nobody is paying for these clicks... Why did you say they were?

You have a scraper and have transitively admitted that you are costing advertisers money... by using the exact ad URL (with FB's metadata/tracking)... with 5000 accounts... Now I know where my ROI went bro.

Unless of course, you are full of shit.

---

Hate me :P

---

Home > Paid Traffic Sources > Facebook & Instagram