Comment by notlisted | Hacker News Reader

Comment by notlisted | Hacker News Reader

Parent Full thread

Since it was my post, let me hijack the hijack and refocus the discussion.

FB advertising has major issues and not just fraud.

#1 - FB has a feedback loop issue

The main reason why (for us) Google works so much better, is that it spans the full sales funnel, from impression to click to registration and finally conversion (room bought).

Google optimizes ad serving for themselves AND for me, and I can see the results. As long as they generate more revenue for my client at an acceptable cost per conversion (sales, not likes) we will spend more on google. Google isn't just taking, it's giving (a lot). As long as the last segment, ie actual conversions affecting ad serving, is missing, google will remain our favorite marketing channel.

#2 - Facebook has a major intent issue.

Eyeballs are nice, but only if those eyeballs have the intent to convert/purchase, which is rarely the case for high price, non-luxury and non-lifestyle items on the FB site itself, unlike a lead generated by google's search engine.

I have read about the "tricks" to minimize useless clicks and learned about some more below, but it really shouldn't be that hard.

PS I do realize that my situation is quite specific - see comment above or below, hospitality industry, busy market, price-sensitive audience, not a big brand name, etc. It may work for you, it just doesn't work for us, and not just because of the fake likes.

Regarding #1 - Facebook conversion pixel (a JS script that has to be loaded the HTML header) solves the feedback loop problem.

https://developers.facebook.com/docs/reference/ads-api/offsi...

Is there a native mobile equivalent for this? Identifying our FB ad cohort is a major problem when trying to calculate the effectiveness of our ad spend on FB.

Yes, there are dozens of preferred marketing developers who provide this service.

In my experience, Facebook actually has the best performing ads with respect to measurable financial objectives.

I think facebook's official mobile SDK will let you do this.

Thanks. Quickly glanced over this page (I wasn't aware of this option, I admit) but can anyone tell me if the conversion pixel is just a tracker for conversion, or whether it actually affects ad serving? I know on google it does.

Anecdote: once had a Google advert that resulted in many more clicks, but another advert was served more. When I inquired about this, google staff explained that the other advert had fewer clicks, but resulted in more (or actual) revenue. I did not have to manually track the results of the campaign and adjust, it was adjusted for me.

You can optimize based on the expected rate of conversion.

Well, since you're "here"

Here's the question I'd want to ask: if the Facebook ad clicks are from click-farms but you don't pay for them, how did they get there?

One guess would be that click-farm people like stuff at random to "seem legitimate". Maybe Facebook itself incentivizes fake account to like stuff by not deleting fake accounts that behave "conveniently".

But that's just a guess, I've look for someone to do more research.

The parent comment author is not the original video author, given his appreciative comments for the video. (Not sure otherwise what you meant by 'Well, since you're "here"')

As to your question, that's exactly what the video covers. With a nice animation and everything. And really, I expect Facebook to post some useful numbers rather than forcing us to do the research.

Jesus, did you watch the video that's subject of the post, or are you just posting, since "you are here." This is exactly the information that the video covers, with animations, explanations, and even links to sources.

I wish we are sufficiently close to a time, when automated comment-comprehending and comment-curating software - like browser extensions and plugins - are widely available, thanks to advances made in the areas of (textual) Deep Learning and (textual) Artificial Intelligence.

Perhaps then, clumsy comments like the one made by joe_the_user:

  One guess would be that click-farm people like stuff at
  random to "seem legitimate".

could be neatly redacted by the extension, without affecting the rest of the comment.

I surely hope so.

Some of the comments that surface to the top of the heap, are head-achingly dreadful.

Including this one, ironically. x0054 called out the problem, no need to pile on with nothing more of any particular value to add.

My comment is less ironic than you might discern.

History is rife with far wilder inventions than the comment-curating 'contraption' that I describe, that have been conceived and crafted, drawing from a far smaller source of inspiration than the clumsy comment made by joe_the_user.

In all probability, the sort of thing, I describe, would never materialize in the foreseeable future. It could just be a figment of my imagination.

The leaps and bounds that are required for such comment-comprehension - especially the two step thing needed for the software to assimilate the contents of the video and then contextually-observe that joe_the_user has not viewed the video, just by reading his comment - are probably far too great for the current state of Deep Learning and AI (someone well versed in these areas could chime in).

Or perhaps such advents in content curation - unbeknownst to us - are already afoot.

Surely then your quick dismissing of my "value-less" comment could be the butt of jokes for posterity.

Time will decide.

You're writing a lot of words without actually saying anything. You can try and dress your original comment up all you want, but all you were doing was kicking someone while they were down to make yourself feel superior, while attempting to disguise it in the form of some highbrow but vague speculation of some future technology without providing anything of concrete value about it. This is HN, we're all aware of the possibilities of strong AI, but unless you've got a github account with a demo in it or some new algorithm, you're just spouting bs.

(Also, you really don't want to end the world just to get better hn comment threads. Strong AI is scary, man)

The idea floated in this earlier Next Web article cited in the video is that click-farmers make use of the Page Suggestion feature, i.e. each time they like a Page FB suggests a new page they might also like to "like". What was once a useful service for discovery for legit users may have turned into the click-farmers' best friend.

http://thenextweb.com/facebook/2014/01/23/likes-lies-perfect...

My understanding was that identifying fake accounts automatically was a pretty hard problem to solve. That's not making up for the likes being fake, but it does justify why they can't just delete all the fake accounts.

To me, it seems clear that if something's going on, it's not Facebook "conveniently" letting the fakes go. It's more the problem being hard to cope with.

Or somebody present me results that demonstrate how easy it is to identify fakes with low false positives.

As others pointed out I'm not the author of this wonderful video.

Reply on news.ycombinator.com