Facebook ad click numbers are inconsistent with other analytics | Hacker News Reader

Facebook ad click numbers are inconsistent with other analytics

medium.com

68 pointsadarsh_thampy37 comments

Facebook has a history of fraud in their other services. Veritasium fell victim to FB's scams as a publisher a couple years ago[1].

    I know first-hand that Facebook's advertising model is deeply flawed. When I paid to
    promote my page I gained 80,000 followers in developing countries who didn't care about
    Veritasium (but I wasn't aware of this at the time). They drove my reach and engagement
    numbers down, basically rendering the page useless.

They also have a history of encouraging video fraud[2][3].

[1] https://www.youtube.com/watch?v=oVfHeWTKjag

[2] https://www.youtube.com/watch?v=L6A1Lt0kvMA

[3] https://www.youtube.com/watch?v=t7tA3NNKF0Q

I think the explanation here is that Facebook tracks clicks on the banners on their site while the author of this post tracks page views using GA or other tools. These are two different metrics. In between lies (at least one) redirect (from Facebooks click tracker to your website) plus a full load of your website. There are a lot of visitors who click but never make it to the second link: Be it because they accidentally pressed the link (very likely) or because they decided otherwise and aborted.

I'd still go with your conclusion that it is difficult for the advertisers to track whether the ads pay of if you are paying per click. I would not go as far and say that Facebook is doing fraud or scam.

This is particularly true for mobile, where it's very easy to mistakenly tap on an ad while attempting to scroll/swipe.

The natural user reaction is to quickly go 'back' by swiping or hitting the back button, often before promoted site has time to load and fire off its mess of tracking pixels/javascript.

However, Facebook itself has tracking pixels, and are configurable such that you can optimize for visitors that reach conditions (e.g. complete a signup). It makes much more sense to look at costs and ad performance based on these downstream metrics, vs. looking at banner ad clicks. (As stated in the article)

> This is particularly true for mobile, where it's very easy to mistakenly tap on an ad while attempting to scroll/swipe.

Even if this is the case, I think it would be hugely unethical of Facebook to sell accidental clicks as deliberate engagement.

Watch out when using tracking pixels. I am sorry I've not been able to find the source yet but in the past Facebook has said that they use them to target ads at other users if they successfully hit your tracking pixel.

It makes sense that they would as they now know that user is likely to click ads and follow through, but will in the long run increase your advertising costs as you're now competing against others for that user

I would agree that, especially Javascript based tracking, numbers differ. What looks suspicious here is a factor of 10 though (19 to 190), which is hard to believe coming from people pressing back/mobile/slow loading etc.

As a mobile user of facebook I can attest that 100% of the ads I click are accidental and terminated before load completion.

Would be better if he'd looked into server logs and check referrer/utm.

There will always be a discrepancy. No doubt about it.

However, the magnitude of the discrepancy has never been so much in any of the networks I have advertised on.

As someone who has overseen spends to the tune of over 1 Million USD on various networks, I can attest to the fact that this does not happen often.

The discrepancy is why I like to have a conversation with my advertisers about the reporting I send them versus the GA activity they track on their website.

My numbers might be slightly higher a than what they see in traffic, but if they're still happy with the activity happening on their site, it is still a successful ad buy.

Edit: should also mention that my ads are priced monthly and not per click.

are you offering a flat CPM rate or RON on a specific site?

Comapring apples and oranges.

Facebook ad clicks are not the same thing as advertised website's uniques/visitors/hits/sessions/whatever. Especially if the latter is using external tracking that uses JS and tends to be blocked/opt-out by some.

Facebook clicks are really 'clicks', not 'unique people clicking', they are also not 'successfully redirected after click', maybe not even 'click successfully finished and not reverted by aborting'. It just means someone clicked or tapped on the thing used to advertise.

So yes, you get what you pay for, and in the case of Facebook it's their definition of a 'click' you are paying for.

By the way: You get pretty close to their numbers when using a redirect script that loads in a few ms and just writes 'hits' to a database and then again redirects to you to your normal target website. Not 100%, but much better than using GA or any other external tracking on the target website.

As I mentioned in one of the earlier comments, the difference in reported numbers is just too big. Some amount of discrepancy is acceptable.

If people are clicking on FB ads and not landing on our website for whatever reason, then those are probably bot clicks.

Repeat the test with a small PHP script in between that does +1 every time it is called to a value in the database of a text file. Then redirect to the real website that you track using GA or your own tool.

Until then, unfortunately, your argument is invalid. I think you will see this number will be quite close to what Facebook reports. What's left then, still isn't 100% fraud done by Facebook as you somewhat imply, but also aborted clicks etc, but a number we can actually discuss.

What are your thoughts on "accidental clicks"? Say, I accidentally click a facebook ad while scrolling, then close the new browser window before your site's analytics scripts load. Would that account for a large difference?

I know that I thick-thumb at least 10 ads for every ad that I intentionally select.

Analytics scripts are at the very top of the HTML file usually, and are one of the first things loaded. In this case you'd be counted I think.

I doubt very much that your scenario matches the general public. My guess is your scenario is maybe 5% tops (based on no reproducible data!) :)

This is very interesting, especially when using two separate analytics software to see that Facebook is not producing accurate numbers. Has the author considered reaching out to a tech journalist and seeing if they'd be interested in pursuing an investigation? Seems if Facebook has a little heat on them, they'd at least explain the discrepancy.

I haven't reached out to any tech journalists. Sent it out ot a few people who covered FB. Didn't get any response. So didn't pursue the idea.

You may want to try leaving a tip for TechCrunch - http://techcrunch.com/got-a-tip/

Another one is Gawker (although are a bit more controversial, but they do like to stir the pot quite a bit) - tips@gawker.com

If nothing comes of it, thanks for taking the time to post about your experience with FB Ads.

Since the article title is linkbait (if not both misleading and linkbait), we changed it as the HN guidelines request (https://news.ycombinator.com/newsguidelines.html). But if anyone can suggest a more accurate and neutral title, we can change it again.

> tail /var/log/nginx/access_log | grep 'your_landing_site' | grep 'facebook' | wc -l

How about run something like this for an actual concrete number of visitors, rather than relying on JS tracking? That's much more concrete evidence.

Reminds me of this thread from a few years ago here on HN when a clothing company accused Facebook of similar: https://news.ycombinator.com/item?id=4312731 .. "Facebook was charging us for clicks, yet we could only verify about 20% of them actually showing up on our site." and other stories from HN users.

It should be taken as a given that these networks are going to inflate their stats and really be more expensive than they look. The conflict of interest these platforms have in presenting an accurate analysis of clicks v. making money for clicks is just too intense to expect anything else.

I've been working with Facebook's ad api since its alpha days, and I can tell you that this problem has always been there. This is not just with Facebook - its a many other mobile/display ad networks as well. That is why clients insist on using 3rd party tracking solutions - they trust those numbers better. We are used to seeing discrepancy between reported impressions and clicks ON the ad platform and those reported by 3rd party tools. Its just the way its always been, and I've never heard an explanation that sounded plausible enough to be repeated to a client.

Few thoughts:

* Was your GA reporting period covering the day? The reporting always defaults to up until yesterday for me

* AdBlockers! Facebook click tracking may use backend metrics while GA is frontend, and blocked by various adblockers.

I've really enjoyed Segment.io, you can track GA page views etc via the backend which could provide better metrics ¯\_(ツ)_/¯

We tried using FB ads and the differences in what FB was reporting and actual traffic were pretty big. It felt like one giant advertising scam and we shut it down a couple years ago. It may work very well for certain consumer segments but I greatly question the numbers they report. It's not a company that has earned my trust.

I am a fresh graduate and I am working for the digital marketing department. Any suggestions on tracking the clicks on Facebook? It would be better if the tool can integrate different social platforms.

If you want to track only link clicks, then use a URL shortener like bitly. They give you analytics for free.

If you want to track the actual users on your website, use any of the analytics products available- Google Analytics, Open Web Analytics, Piwik, etc.

Adding UTM parameters to your URLs is the ideal way to attribute website traffic. Just make sure you are consistent in your UTM values.

all of these can be faked. if you really want accurate information, server logs are the best. the only thing facebook or users can't block.

if facebook says you got all the clicks, use the server logs to show otherwise.

-- edit. I'm not saying that I trust facebook though. I don't.

Of course users can fake server logs, since Referrer info is totally up to the client. Also, I'd argue it's harder to distinguish a bot from a real user from the server logs than from js.

Google Analytics is undercounting by 20-30% in our experience. No reason to waste time analyzing the wrong data.

I experienced this fraud as well, and knowing their policy about default reaching... this is a real shame. We ended up doing less and less facebook ads.

Oh, look, it does't actually matter. What matters is whether or not advertising on Facebook, or anywhere else, actually makes you more money. Instrument that, calculate bang/buck et voila.

So you didn't make it to the end. That was the conclusion.

  Should you stop Facebook ads? Depends on actual ROI. Look at your business data and see if Facebook has been working for you.

I think that it sort of does matter though. You're paying for those clicks afterall.

It does actually matter because you're paying for the clicks. I, for one, sure as hell don't want to be paying for bullshit clicks even if I still make a profit.

Reply on news.ycombinator.com