“I feel happy. It made me realize the answer was already inside me. Wahanegi is extremely helpful, and I strongly recommend it. I like it.”
Are you facing a big life decision? Try the beta today, it’s anonymous, free, easy and effective!
Dear Facebook Advertisers: I just published a much shorter posting about how-to save money on Facebook advertising based on the information below. I also posted an article based on past BBC and TechCrunch articles that includes a more recent communication from Facebook customer support.
This is an email thread between me and a Facebook customer service agent named Neil about my request for a refund for 90 percent of the advertising we did with them recently. I am requesting the refund because I believe Facebook misled us about the nature and value of the clicks they delivered as part of our ‘Like’ campaigns.
If you read this and do not fully understand what I am saying in the discussion, then do not advertise on Facebook. If you read this and understand it, then you have fair warning and can proceed with caution as you see fit. In that case, I recommend you also read my posting about how to detect and mitigate the problem.
FWIW, I recommend against speculating in their stock one way or the other until they make a clear public statement about the situation. To see why I say that, read this post for some very rough estimates of how large the problem may be.
To help you understand what is going on, I need to define a word. A booklicant is a Facebook user who consistently clicks on dozens of ads in a day, perhaps 4 or 5 in a minute, adding up to hundreds or thousands of clicks per year. They do this for various reasons that are generally unrelated to typical click-based online advertiser expectations – to collect things they like, to increase their ‘like’ score, to express their interests, to help Facebook better target content to them, to while away the time. They are a small percentage of all Facebook users, but they represent a large percentage of Facebook ad clicks, especially in smaller ad campaigns.
From: Erik Larson
Date: Fri, May 25, 2012 at 1:47 PM
Subject: Re: Help with Your Ads Manager
To: The Facebook Ads Team
Thanks Neil, I greatly appreciate this much more rapid reply.
To short-cut another possible delay that may occur, I already know these people are real users based on my own research. That is not the problem.
The problem is that Facebook misled my company with respect to the inherent value of most of the clicks by claiming that these clicks were comparable in value to clicks in other CPC venues or to clicks by FB users with typical and expected interactions with CPC advertising (emphasis added). To be specific, I believe that about 90 percent clicks you charged us for were worth about 1/1000th of the price you charged us. It also seems likely that your company was aware of this disparity in value, because after looking at the data some more it feels like you must have some sort of algorithm in place that invalidates clicks from these users once they pass a certain threshold, but probably you only invalidate clicks after that threshold, not all the clicks that came before. Even more, that threshold appears to be WAY above what any reasonable advertiser would expect it to be.
Specifically, my gut feel is that you allow valid clicks as long as people click fewer than ~4-6 rapid-fire clicks in a minute, and then after that point you invalidate their clicks within a very short session period, probably around an hour or two, but you do not go back and do what everyone would expect you to do and invalidate the first ~4-6 clicks. Further, I imagine you have a very high threshold for clicks in a day, probably ~40-60 daily, and similar to the rapid-fire clicks you do not go back and invalidate the first ~40-60 clicks. It also seems like you have no algorithm at all based on how many annual or total likes a user has, which is somewhat shocking since combining such an long-term algorithm with a rapid-fire algorithm and a daily algorithm would be the best way to identify ‘booklicants’ or ‘likers’ or whatever you want to call this group of users. I’ll continue to call them booklicants because from an advertiser’s perspective they are as much a creation of your product interface as they are regular people with rational interests.
I am guessing at those numbers after looking at the data from our campaigns and imagining how to reconcile the results, but the real basis for my guess is empathy. If I was your product manager for these clicks, and I had data at least as good as that provided to me as an advertiser, that is roughly where I would put the set points if you asked me to stretch the validation algorithm such that it filtered out truly bizarre behavior (e.g. perhaps people who have OCD about the word ‘like,’ or people who appear to be in the employ of a click-fraudster, etc.) yet did not entirely destroy revenue from these people since they represent a large percentage of our CPC inventory. I would then create some sort of rationalization after the fact – for instance, “Setting the limit at 4-6 clicks per minute is completely reasonable because it takes the average person about 10-15 seconds to read an ad made of one small picture and fewer than 115 characters, thus the user could reasonably have read the ad and made a decision to click.”
As you can tell, I am not just talking to you, but to whomever else reads this on your product team or on my blog. And to be clear, I think that the rationalization above is false and misleading. So if I had been your product manager, I hope I would have refused to do it because at a minimum it is unethical from my standpoint. At least I hope I would have said no, but that would have been a very hard thing to do in the face of millions of dollars coming towards me and all my colleagues. On balance, I am glad I was not in that position. I am trying to earn my millions some other way.
Here is an ethical algorithm to use instead: retroactively invalidate all clicks by users as soon as they click more than twice in a minute, more than five times in an hour, or more than 500 times in a year. By retroactive, I mean both all the clicks within that particular timeframe, and also looking back at their behavior over the past five years to determine other times that behavior occurred.
That algorithm is ethical because the booklicant interactions clearly do not represent a meaningful CPC advertising interaction by any reasonable standard (emphasis added). The reason they are clicking the ads is to express their interests to Facebook so that your company can present them with more things that they like, and also because it is satisfying to them to like many things. The are essentially collecting likes and interests. However, you can’t say that is the same as consciously liking something any more than you can say a wild-haired Berkeley liberal who happens to collect political campaign buttons ’likes’ Joseph McCarthy or Sarah Palin just because they have several of their buttons in their collection. I am sure that is a big part of why your product interface makes the number of likes so prominent - your most active users probably asked for that feature, so they could keep better track of their collections. All the rest of us just wonder why that ‘Likes’ box is so prominent in our new Timeline interface.
The booklicant algorithm is important because Facebook implies that advertising on the site is a form of ‘CPC’ advertising, first of all by using that industry standard term, and also by using industry standard terms like ‘CTR’ and ‘invalid clicks.’ That is where you misled me, and it appears you have misled a large swath of your advertisers as well. The current context for CPC ad pricing is search (high value clicks because the user intent/goal is fairly clear) and display (lower value clicks because the user intent/goal is less clear but still represents intent to interact with the ad and advertiser). In both of those cases, no one would think that there are normal human beings out there clicking on ads to somehow ‘collect’ them in a sort of general interest inventory, NOT EVEN THE EXPERTS, BECAUSE OVER THE LAST FEW WEEKS I HAVE TALKED TO SOME OF THEM AND THEY WERE ALL SURE THESE ‘BOOKLICANT’ PROFILES MUST BE BOTS AND THAT YOU WOULD QUICKLY REFUND MY MONEY.
Blog Side Note Not From The Email: Neil, et al., I apologize for the ALL CAPS…my passions got away from me for a moment there.
I am not saying that such likes have zero value. I am estimating that the market rate for such clicks is probably about 1/1000th the current value you are charging, since these people seem to click on about 1000 times as many ads as typical users (emphasis added). You have two completely different classes of ad inventory on your network, people who collect likes and people who don’t, and you are charging us the same for each even though it is clear the behaviors represent categorically different levels of intent and engagement with the ads. I will happily pay 40 cents for a click from a UK resident who only clicks on things that they are curious about on their own merits, i.e. like what happens with display ads. But I will only pay 0.04 cents for a user who regularly likes 20 to 40 or maybe more than 100 things in a day. It is wrong to allow advertisers to think that all clicks on Facebook ads are the same even though you could have simply and easily communicated strong evidence and understanding to the contrary.
How you have managed to continue to collect money from advertisers in this way for the last several years is amazing to me, and to most people I have talked with about it. They just don’t believe it at first: “How is it that more people are not shocked about this?”
Since I fell into the trap I can tell the story and explain why. Most people don’t want to believe it, so they don’t. Here is my story:
I started out advertising my external landing pages like a typical CPC campaign. The CTRs and landing page conversions were terrible but my ads appeared to perform better than the average FB ads, so it felt OK, and the demographics of the clicks validated some of our segmentation assumptions. Then since your app and your marketing highly encourage use of ads to market internal FB content, and you promise much more detailed data about the users, I spiffed up our FB page and ran my first ‘like us’ campaign. WOW! Hundreds of people in less than an hour, and the CPCs were 20-30 percent lower! So I did it some more, right then, spent more money, eventually $170 instead of the $30-50 I’d planned on. It was exciting.
Then I did a little bit to try to get engagement from those users, a few posts to the page, and nothing happened. I read more about how building engagement is a skill that requires investment, and I also began to look into sponsored stories. Luckily, in parallel I was analyzing my fan base and discovered they were not what they appeared to be at first. They were mostly ’booklicants’ who like dozens of things a day. So I didn’t do sponsored stories, I sent you a note asking for my money back. If I had done sponsored stories and managed to get into their streams, then no doubt some of those users would have engaged, because they clearly use FB a lot and I would have been jamming my message in front of their faces…which is what I thought I was doing in the first place, btw. But I bet the competition for those particular streams is relatively intense, not because they are more valuable and more advertisers want to compete for them, but because those people typically like thousands of advertisers, and since I was a small advertiser with an audience comprised primarily of booklicants this would have hit me hardest.
And that is the #1 reason I think what you are doing is wrong. It hits small and unsophisticated advertisers the hardest (emphasis added).
You could have given our money back after my first inquiry saying something like “All of those users are real based on our initial investigations, but we will continue looking into the matter. We value our customers and in particular we appreciate your detailed feedback, so we would like to offer you a $300 credit as a thank you.” I would have been happy, I would have spent a lot more that $300 over the next few months, and I would have enthused about it on my blog, etc. That is not what happened, I suspect because you have internal policies in place that require you to take certain steps to give you legal standing in case of lawsuit, and probably the last thing you want is for people like me to be publicly enthusiastic about large refunds from Facebook as a result of strange click behaviors.
I really want my $150 back. I now doubt you will give it back to me any time soon, because doing so would effectively require you to admit to what I have described, or explicitly explain why I am wrong. If I cared about the money, that means I made a tactical error by sharing all this information with you and the readers of my blog postings. However, what I really care about is the truth, and figuring out how to use data and social networks and so forth to make people happy. I am happy that the booklicants like ‘liking’ things, and I am happy that Facebook exists, and I am happy to spend money on advertising. I am even happy that you misled me and caused me to waste so much time and energy on this, because it helped me better understand how people interact on social networks and it gave me some great blog postings that were fun to write. But I am not happy that you continue to mislead other small advertisers who do not have the skills, education and experience that I have. So I will continue to explain my understanding about how Facebook ads work to other people until you guys take responsibility and start to do it yourselves.
One final suggestion. Just come out with it, right now, all at once, while the clouds are already gathered, and try to turn it into a pivot point. Maybe do it this way:
“We have discovered an amazing type of user in our Facebook community, they are people who like lots of things and as a result they provide us with amazing data about general human preferences. No other company has this information, it is unprecedented in the history of advertising, it will allow us to create both relevance and serendipity at a scale never before imagined. From this point forward, we will not be charging advertisers for the clicks from such people, but instead we will use that information to dramatically improve our ability to target meaningful awareness advertising to other users just like them, primarily through very high value multi-media CPM sponsored stories, first on mobile devices, then on our website. This is a material change and will have a short-term negative impact on our revenue of XX% effective immediately, and we will also take a one-time loss of $X billion in the form of ad credits refunded to all advertisers who have paid for the clicks from these people over the past five years. Our FAQ describes the algorithm we will use for this refund. This is a new industry, but as you can see we are an agile company that keeps the interests of our users and our advertisers front and center. This is just one step in our march towards the future.”
If your company has the fortitude to make that sort of decision in the face of all the likely consequences, then you have a chance to become a boon of unprecedented proportions, make a lot of money in various ways and have me as a shareholder.
If not, then I’d say all bets are off.
I know it is hard, but I’m rooting for you to do the right thing.
On Fri, May 25, 2012 at 12:24 AM, The Facebook Ads Team wrote:
I might get delayed in responding to you, but certainly wouldn’t forget. I’ve reached out to our products team to investigate into the two campaigns you mentioned below. We’ll look at the delivery as well as the kind of users who have been clicking on your ads, but at first glance, these clicks appear to be coming from real users with real likes and interests.
Will keep you posted on this. Thank you for your patience,
Global Marketing Solutions
From: Erik Larson
Date: Thu, May 24, 2012 at 9:18 AM
Subject: Re: Help with Your Ads Manager
To: The Facebook Team
Thanks for the reply, I appreciate it. Thought you’d forgotten about me ;).
In the intervening three weeks I’ve done my own research, which makes this easier.
To keep things simple, I would like to shrink my request for refund and limit it only to the money we spent to advertise ‘liking’ our page at www.facebook.com/wahanegi. I am requesting the refund because a large number of the clicks appear to be as a result of Facebook users engaging in behavior that has very little to do with a meaningful response to our ads and is mostly about their attempts to enhance or enjoy their own experience using Facebook as suggested by your product interface. You can read more about the problem here: http://wp.me/p2dKfK-5s.
If you also want to keep it simple too, then please just refund all of our money from the following two campaigns on our Wahanegi account:
$109.91 spent on We Need Your Help – ANZAC
+$62.40 spent on We Need Your Help – Philippines
If you want to be more specific, then you can easily do the required reporting, because all of that traffic stayed on your servers and was driven by users authenticated into your app. Further, since Facebook does not provide sufficient reports for advertisers to allow us to determine how much money was spent to generate any particular ‘like’ by a specific user, it is literally impossible for me to provide you with the necessary data to get a more accurate sum. You must do it.
If you want to go that route, then I am specifically asking for a refund for any clicks that we paid for that came from Facebook users that hit the following filters:
*Refund money for any paid clicks from users who have liked more than 499 pages in the past 12 months, OR that liked more than 9 pages on the same day they liked our page, OR that liked more than 149 pages in the past 12 months AND more than 5 pages on the same day they liked our page.*
When you run that report, I would appreciate it if you could provide the full results to me so that I can review them for accuracy and decide if I want to remove or block any of those users.
Finally, if you don’t want to do that work but feel that I am asking for too much since these clicks are probably legally valid even if they are meaningless as a measure of advertising effectiveness, then maybe we can compromise. I estimate that the filter above encompasses about 90% of the clicks we paid for. So you could also just refund $150 and we’ll call it even.
Please reply either way and keep me in the loop wrt your progress and any questions you have.
On Thu, May 24, 2012 at 8:08 AM, The Facebook Ads Team wrote:
Thanks for following up and giving a detailed background to this issue. I really appreciate your inputs and will make sure your feedback is heard by the products team. Meanwhile, we are happy to look into this further if you can provide us with detailed click logs documenting the activity you’re concerned about. We aren’t able to investigate this further without actual traffic logs. We understand this may be a frustrating process, and we apologize for the inconvenience. Please contact your web hosting company directly if you have questions about how to obtain server logs detailing individual clicks or visits to your site, as we aren’t able to verify data collected through third-party tracking systems. In order to assist you further, please include the following information from your server logs:
(A) Raw server logs of all clicks coming to your website, or the total amount of all clicks coming from Facebook, with an explanation of how you filtered them. These server logs should, at the very least, include:
(1) Timestamp of page load
(2) User agent string
(3) User IP
(4) Exact page loaded, with the parameters passed to the page load if you are doing URL tagging. Please note: a popular tracking method is to link your Facebook Ad(s) to unique URLs that are only used for specific Facebook Ad campaigns. Another tracking method is to add an extra, identifying parameter to your URL. By doing so, you’ll be able to isolate visitors who reach your site through Facebook Ads rather than other traffic sources.
(B) Aggregated counts of your clicks.
If possible, please also include the following:
(1) The total number of clicks you received from Facebook, split by day, for the specific time period where you have noticed the click issues.
(2) The total number of clicks you were billed for, by Facebook, also by billable day for the period in question.
(3) A screenshot of your external reporting system showing the total number of clicks received from Facebook.
Please note that we do filter out any clicks that appear to be invalid, including repetitive clicks from a single user, clicks that appear to be from an automated program or bot, or clicks that are otherwise abusive. We also cap the number of times any user can view or click on an ad. Rest assured that we take discrepancy reports very seriously and will be happy to take a look through our and your logs for any indication of invalid click activity.
Thank you for your continued patience,
Global Marketing Solutions
—–Original Message to Facebook—–
From: Erik Larson
To: The Facebook Team
Subject: Help with Your Ads Manager
Email address: …
Describe your question or issue: Two weeks ago I inquired about the validity of a large percentage of the clicks from campaigns that we ran via this account … and my personal account … . I also followed up with an extensive description of the problem, but have heard nothing since. I just made a blog posting about it (http://wahanegi.com/dont-
—–End Original Message to Facebook—–
NOTE: I hope this post is helpful. However, we are not in the advertising business…we are a start-up building an app that helps people make wise life decisions. So if you have a chance please help us out by signing up for our beta. Also, you can follow me on Twitter @erikdlarson. Thanks.
Journalists – Especially ABC News and Yahoo! Finance: If you link to this blog posting, you should read what it says and make sure you represent it accurately to your readers. I do not think these clicks are from ‘bots,’ I think they are from real people who behave in an unexpected way. And I am not a blogger, I am the CEO of this start-up and just happen to blog about some of his experiences along the way. That said, I still think Facebook has an obligation to be forthcoming to advertisers about this user behavior, and to make adjustments in their business to end the misleading and unethical nature of their current advertising algorithms. As you can see, I am not a blogger trying to make a big press splash, or else I wouldn’t use the words ‘algorithm’ and ‘unethical’ in the same sentence ;-). You can spin all this however you want of course, as long as it is not a false representation.