socialmediocrity

According to a report on InsideFacebook, Facebook have filed for a patent on inferential ad targeting.

And according to the report this goes beyond friend-connections targeting, as is currently available in the Facebook ad platform (allowing advertisers to target *only* connected friends of existing “fans” or “likes”), and covers the possibility of broader inference. What does that mean? Well, in a nutshell, that if you have infirmation missing from your profile, but amongst your hundreds of connections on the network there is some commonality or popular data item for that missing information, then it may be infered that you also share that piece of information.

To give a similar example to that in InsideFacebook‘s article, if you don’t publish details of your education, but enough of your friends have indicated they are all connected to the same academic institution, then it may be infered that you also came from that same institution yourself, and advertisers may target you based on that inference.

All of which sounds teribly clever, and in keeping with the notion that you are defined by the company you keep. However, there are a number of reasons why this approach can only suppress advertising effectiveness if utilised by an advertiser.

Firstly, accuracy.

I won’t for one minute imply that Facebook’s data is not accurate (indeed they would counter that by declaring it is not their data, but that of the user), but maybe, just maybe, the absence of the data is just the start of the story. I can imagine, for example, and to continue the same example, that students of celebrated institutions will *tend* to have a greater propensity to publish their affiliation than those that attend less well known and less envied institutions.

So consequently, and especially when we have developed professional networks that include people from all sorts of backgrounds later in life, the celebrated institutions become, logically, over-represented (if anything). Inference Targeting in this example, then, would only serve to exaggerate that over-representation, and presumably those infered ads *should* work significantly less well (diluting the overall effectiveness substantially). Algorithms to the rescue? Maybe, but not for a while, and not before a lot of mis-attributed inferences had been made (and reported)

Secondly, User-Suspicion.

Even if the inference was accurate, it is easy to acknowledge that an advertisers’ use of data that I haven’t published about myself could “spook” a user – and is unlikely to solicit a positive response. I liken it to the email I got from Alliance & Leicester yesterday, warning me of a security issue and inviting me to reconfirm my details by following their link to the security test area. (I don’t bank with Alliance & Leicester) If I did bank with them though, it is to me clearly a con trick, and I can entirely understand why users might get caught out by it from time to time.

My final offering would be that, users who consciously publish these details will, in part at least, accept that such information may be used to deliver advertising to them. They will have already witnessed it, consciously or not, and will be familiar with its’ occurence. Arguably, they have given expresss permission for others to use that information. Users who have intentionally not published that data then, again at least in part, have specifically chosen not to publish that data, and consquently are likely to be less welcoming of those ads. And from a privacy perspective, this could be dangerous new ground for Facebook to be forging.

There are surely many other reasons why the inference itself may be inaccurate, but there is only very little prospect of better returns for advertisers. The only real purpose for this, then, is to dramatically increase the scale of a prospective campaign, and consequently drive ad revenues up whilst decreasing effectiveness.The same thing happens all the time, and we have seen it time and again, with some direct similies on the way paid-for Search has evolved.

If, or when, it becomes available, buyer-beware. Facebook’s focus on revenue will increasingly pit them against user experience, and this development might see another battle erupt between user and provider.

It appears that Facebook have changed the end dates / times for campaigns running, and at the moment this could cause a problem for existing campaigns.

If your campaigns have an end date set, check what time the campaign is scheduled to end on that date. It appears as though the default has shifted from midnight on that date (i.e. end of that day) to 00:00 (or the beginning of that day).

It bit me in the backside this morning, although I managed to catch it in time, but unless I missed this change when setting the campaign up some weeks ago, then you may lose a day of your campaign – and if that is the build up to a product launch, or movie release, or similarly orgasmic crescendo to a piece of activity, then it could be the one day you absolutely MUST have your campaign live that you miss.

So facebook have removed the thumbs up, thumbs down vote on ads that appear on the right-hand-side (ASUs) – I thinkn they removed that function a few weeks ago – and have replaced them with single, thumbs up “Like” call to action. (see below)

So pushing the “like” button then displays that I like this, and I can only imagine the same ad will appear on my friends’ pages as “Richard likes this”, in a similar way that fan page ads declare the number of friends who are also fans.

What’s intriguing is how many users do we think will legitimately use this function, and what impact does it have on me as a user?

Click-through rates on sites like Facebook have been well-documented to be low, or at least it takes work to get them anything above 0.02%, so theoretically you would think that adding distractions would only serve to reduce this further.

There is of course a risk here too, insofar as larger advertisers could benefit from paid-for gaming of this particular system. Find a couple of hundred unrelated users to “like” your ad, even if you have to pay them, and the returns could be huge if higher CTRs are the result of friend recommendations of this type.

Of course, the initial reaction of user will be interesting as it is not clear how I stadn to benefit by “liking” certain ads, and if anything I was always more inclined to give a thumbs down to ads I didn’t want to see again, but it is not clear that this will influence the type of ad I might see again in the future. I for one would not have “liked” the first ad in the example here: it is poorly targeted in the first place, and breaches Facebook’s own guidelines. But equally I don’t want my ad space to be inundated with ads for just a small selection of products either, based on the few things I might want to see more of.

It seems it is another attempt to introduce some subjective quality scoring to the ad serving algorithm (if there is one), but I for one struggle to understand what is wrong with CTR as a measure of the quality of the ad – surely if people “like” the ads they see, then they might click on them, and that seems like a fairly robust measure for whether the ad itself has been effective in soliciting a response. For sure CTR can never be the only measure used, but for these purposes it would seem like a fairly useful one.

Are men less likely to declare their marrital status than women?

There are 1.58m Male UK Facebook users either “engaged” or “married”, but 2.37m Females with the same status. That’s 19.75% of UK Males, and 25.76% of Females, before the pedants chime up.

Meaning there’s (probably) about 800k engaged or married UK men out there that don’t want you to know.

Or someone’s taking more than their share of wives.

I’m a big fan of social advertising. I like the idea that advertisers can use that information to give me increasingly relevant ads.

However, as mush as I tolerate, nay, embrace this new model, I can’t abide lazy creative.

Take this example. It talks to me as a fan of Liverpool FC. Bingo. Targeting achieved. I am one. Every bit of one.

But where it fails is the message. Apart from ending the question in their title with a full-stop instead of a question mark, their copy doesn’t seem right somehow.

Call yourself a copy-writer?

Yes, I’ve got the official shirt (two of them, actually).

Yes, I’ve got the hat (even the missus has one, reluctantly)

And yes, the scarf too (again at least 2 of those)

And yet somehow, somewhere, in the mind of a senseless out-of-touch advertiser or copy-writer, the hat-trick has not yet been completed.

Now last time I checked, a hat-trick in football constituted three goals. I remember Robbie Fowler scoring the premierships quickest ever (just over 4 1/2minutes against Arsenal in ’94). Liverpool won 3-0 so he definitely didn’t score more than three.

I remember Peter “good-touch-for-a-big-man” Crouch’s perfect hat-trick, also against Arsenal, in ’07. Liverpool won 4-1, but Peter Crouch only scored 3 of them.

But our learned advertiser, which I trust is the banking partner concerned and not the club themselves, is determined to sell me the mythical fourth constituent of every fan’s memorabilia “hat-trick” – a club-branded credit card.

The Facebook ad platform has stopped accepting any ads, it seems, for ages 13 and up. Your ad, if you select the from age to be “any” or “13″ will default to 18 and upwards. Even if that’s from 18 to 17.

Could be nothing more than another glitch, or it could smack of something more sinister. Either way it highlights soem of the discrepancies in the marketplace right now.

By not allowing advertising to under-18s, the CPMs naturally rise if your intended target is “all” users. Media cost for over 18′s is 2-3 times more expensive than it is for children. In the UK at least, anyway.

There are 2 million UK users on Facebook aged 17 and under (about 11%), and the estimated CPM to target them is 0.05 – 0.10 USD. Compare that to the over 18s market, of which there are 15.5 million people, but more poignantly the estimated range of CPMs is significantly higher: 0.16-0.21 USD.

Is this the first sign of scarcity in the (adult) social networking market, as CPMs start nudging upwards? US inventory has been at a premium for some time now, and now the UK seems to be  catching up.

With new figures showing Facebook overtaking MySpace in the US we’ve begun thinking again about Facebook’s ad revenue model. Of course a recession is always going to cause trouble and concern for ad revenue business models, but with a public platform available, we can begin to get a feel for their actual income.

According to Compete.com, the US user base has just exceeded MySpace’s for the first time, with 59.7 million unique visitors, versus MySpace’s 59.5 million. Google’s AdPlanner appears to show MySpace barely holding on to it’s lead, but it clearly won’t be long before they too will be proclaiming the beginning fo the end for MySpace too.

Of course there’s been constant speculation about the real value of the business ever since the Microsoft investment, which was rumoured to value it at a remarkable $15 billion. Leaked memo’s, emails, and recounted conversations since have reportedly shown that even some shareholders placed some more realistic valuations of $3-5 billion on it.

Heck, we’ve even had the valuation estimated on the Whopper Sacrifice application, placing a tongue-in-cheek $1.8 billion valuation on the community. Before Facebook banned it , of course. And not for the low valuation it appeared to place on the business either.

But whilst I might be over-simplifying things enormously, there’s enough information in the public domain to make an educated estimate.

Google AdPlanner tells us that, worldwide, and to the nearest billion, Facebook attracts 29 billion page views per month, and this is rising steadily.

We also know that, whilst it may vary from country to country and page to page, there are normally 2-3 ads on each page viewed on Facebook. Some pages have no ads on them at all.  So for the purposes of a non-scientific estimate it is probably safe to assume an average of 2.5 thoughout.

That gives us an ad inventory estimate of 72.5 billion ad impressions. every month. And rising.

Of course all the page views in the world (and let’s face it, they do seem to have most of them) are only valuable if you can secure advertising for them all, and at high enough rates. So where would we find out the rates Facebook are receiving for their inventory?

OK, no more rhetorical questions. Although the ad platform shows lower figures rates right now, they do vary, and of coruse much of the ads are not placed through the public platform, but in guaranteed delivery deals. What we can see is that at an average CPM rate of $0.50, if that is what they were achieving around the world on average, would give them ad revenues in 12 months of $435 million. Even if the user base continues to double every 12 months, the next year will generate approximately $630 million.

Of course if they’re not attracting those CPM rates, and let’s be clear that these are not the rates quoted at the time of writing for any markets on the public ad platform, then revenues will be significantly lower, and the same goes for the growth estimates. Sooner or later the rate of growth has to slow down, even for Facebook.

So with an ad revenue income estimate of $630 million, the valuation becomes clear once you determine whether you cup is half-full or half-empty. And of course which version of their costs you are prepared to believe. Alternative income streams should not be ignored entirely, siuch as sponsorship deals, and broader content deals, but with renewed prominence of the virtual gifts business that generates c. $40m per year income, Facebook could be generating $700m in 2009 if they get it right, the vast majority still coming from ad revenues.

Let’s assume costs of $250 million per year, leaving $450 earnings, then a $4.5 billion valuation would not be so obscene. $15 billion, even now, would be.

The key to moving forwards however will be in attracting advertisers, and finding ways to deliver excellent ROI for those clients in order to keep the ad dollars rolling in, at a time when advertisers are looking to reduce costs and slash marketing budgets.

Can advertising on Facebook standup to dominant low cost channels such as paid-for Search, and begin to make itself a regular feature on a marketer’s media plan? The smaller advertisers and marketers are making it work, but not until the larger worldwide agency groups find their way, and bring their clients to the party will Facebook see any great acceleration in it’s ad revenues, and until then their valuation will be debated and speculated over repeatedly.

WTF?

Facebooks fan page ads have taken another turn.

Mitchell and Webb being bumed off

Instead of just asking me to become a fan, I am now told that one of boys is bumin off to this too.

Whatever that means.

Anyway, clicking on the link simply displays which of my “boys” is “bumin off” to this. Turns out it’s one of my be-yatches.

What a strange turn-of-phrase Facebook uses sometimes.

In something of a follow up to an earlier post, Facebook-tastic. Let’s all be fans of everything, it seems Facebook have taken things further.

Or it’s broken.

I was slightly bemused by this.

I don’t know if there is some undefined relationship between the Farm Tavern, a publilc house in Brighton popular with my friend Kate, and the artist Martha Herbst, also from Brighton it seems. But on further investigation, there doesn’t appear to be an obvious one, or at least not one involving Kate. Looking at Martha’s 63 fans, Kate is not amon gst them yet, even.

We’ve seen time and again that micro-targeting is far less effective at increasing response rates if you don’t also micro-message the audience, and once you remove the relevance in this manner, it is easy to see why. This is nothing more than an untargeted ad for Martha Herbst, with a random picture of one of my friends on.

It seems to be a backward step and somewhat puzzling if it is a Facebook initiative, though I would be intrigued to see the results these are generating. Of course this could be what happens when adding Social Actions to your campaign goes wrong, and if that’s the case it is perhaps something Facebook need to protect their users against.

This morning the apparent relaunch of Lexicon. It’s been clunky on Firefox, but I got it to work better on IE, and the link on the advertising homepage seems to be broken, but I finally managed to browse around here: Facebook Lexicon

The fundamentals remain the same, insofar as the list of terms remains restricted. Clearly there’s much more value in being able to type your own terms, but that doesn’t appear to have been reintroduced just yet. When Lexicon first hit the scenes you were able to create your own term or terms to monitor, but that was replaced with a predefined list not too long after launch.

What is interesting about New Lexicon is the greater depth of data and insight available. As well as being able to see the number of individual posts on any given subject, we can now see the number of individuals generating those posts, and also the proportion of posts containing that term.

Facebook has gone further though by also providing much deeper analysis of what those people look like.

Under demographics, we can break these stats down by Gender, Age and Country of the users concerned, which begins to get useful (especially if the terms themselves become unrestricted)

Age breakdowns are provided in bands, and country breakdown appears for now to be limited to the major English-speaking countries in terms of Facebook user numbers: US, Canada and UK.

On top of these fairly standard insights though are some far more interesting, and distinctly more web two-point-oh insights.

Associations goes someway to understanding the other topics being discussed alongside the topic in question. IN some respects similar to Amazon’s recommendations engine: People who discuss “Politics” also discussed “President”.

Keyword associations

Sentiment reports will be familiar to anyone that’s ever seen buzz-monitoring reports, and undoubtedly it’s limitations will be similar. However this report does at least go someway to giving us an extra dimension over the original lexicon, and some indication of positivity – essential to a brand if they want to understand the direction of the effect they are having on conversations amongst their consumers.

Topic sentiment

The new Pulse report begins to give brands and advertisers some guidance on what other keywords and interests might indicate interest in any given topic. This can easily become very useful, especially in utilising the micro-targeting features of the ad platform in refining which users you might want to target with your message.

Other interests of the users

The final addition, a US map displaying the geographic spread of the users by state, is of course interesting to US brands and campaigns, etc, and it will of course become increasingly useful as they extend this to cover other countries and regions.

All in all, this looks to be a hugely empowering move, exposing much greater insight to advertisers, although the real value will only come if and when they allow unrestricted terms to be searched and analysed. Perhaps instead a premium service for particular brands is in the offing, as this represents a huge revenue opportunity for Facebook themselves given the scale of the conversation-poll they are able to conduct in real-time.