This afternoon, my daddy blogger friend posted a status about grilling hot dogs. Facebook suggested “hot dogs” and “breakfast” as related interests for me to like.
I know that Facebook REALLY wants me to click “like”, adding peer pressure by showing how many others have done so– and showing the names of my friends who have as well.
But what does this have to do with increasing their revenues? Everything.
Like Google, or any other ad-supported network, Facebook’s earnings are determined by the equation, Traffic x eCPM. eCPM represents how well they monetize that traffic in terms of earnings per thousand ad impressions. Facebook has a ton of traffic, but is likely making only 30 or 40 cents per thousand impressions on it, about a tenth of what display networks get on lower quality traffic. How is it that Facebook, with ridiculously precise targeting, can’t make as much as some websites that sell untargeted traffic? Several reasons, which we’ll discuss in technical depth below:
- More data on users equals more inventory. If you were Mark Zuckerberg, would you prefer doubling Facebook’s user base or getting twice as much information on each user? It’s the former that counts, while the latter is just more remnant demographic inventory. By getting users to click on likes (50 million likes a day on Sponsored Stories alone, according to Carolyn Evers of Facebook), they are increasing their effective inventory. Every time you click “like” on something, consider how an advertiser might use that information. Sobering, isn’t it? But advertisers don’t yet know how to leverage this because….
- Facebook has released a spaceship to a populace that knows only how to drive cars. People are crashing their spaceships. They need training on how to use this targeting. The old-timer SEOs fare the worst in social, since they think their SEO experience carries over into social. They don’t realize that organic and paid components work together, as evidenced by Facebook’s EdgeRank algorithm. Yes, your ads can influence organic traffic– this is not Google.
- Agencies and software providers haven’t integrated their systems. Only a few folks even have access to the Ads API, so unless you are a whiz at manual operations or can hack the system (like we were doing 4 years ago before there were APIs), then you’re doomed to low performance or having several people full-time manually loading ads. Facebook’s ads team is moving as fast as they can (they’re great people, by the way), but there’s a long way to go. Even Marin, the biggest paid search platform, doesn’t yet have Sponsored Stories or optimization to a cost per fan target. New features are released in the Ads UI before they hit the API, which frustrates users of the API and hampers mass community adoption.
With me so far? Facebook does $2 billion a year without even trying, while Google is 15 times that. Do you think that Facebook can increase the amount of data on users by a factor of 3 to 5 in the next year? Will advertisers start running targeted ads that are truly social vs rehashing their old PPC ads? We’ve often gotten a 3x performance improvement on just doing this. Will a new breed of small business advertisers come onto the scene to extract value from localized inventory? Any one of these factors can close the gap for Facebook, but it requires massive education that they’re not prepared to do quite yet, evidenced by the difficulty of reaching folks at Facebook– even us.
Now for a technical analysis– you can stop reading now if you’re not a technical marketer. But if you don’t mind investing a few minutes on understanding the mechanics of eCPM, it will be well worth your time in understanding how Facebook makes money and how you can optimize Facebook traffic.
Last month, Efficient Frontier reported that average CPC prices on Facebook are up 40%. In addition, TechCrunch is saying that Facebook is on track to surpass Google in revenues if you project their earnings curve forward. It’s true that prices are slightly up, but for advertisers that see massive increases, it may be a sign that they’re not using the ad system as it was intended. We have still been able to garner clicks for fractions of a penny in entertainment verticals.
At BlitzMetrics, we don’t see a significant increase in our average CPC (it rose from 50 cents to 54 cents) during the first quarter, but that is a trick question. Here’s why. Facebook and Google both operate on the eCPM auction model, which Hal Varian explains here. The CPC you pay is affected by both the competition for that traffic as well as the CTR. We find that the CTR by far is the biggest driver in traffic cost. If you want to measure the cost of traffic, don’t look at CPC. Look at CPMs for each type of traffic– by country, demographic, interest target, and other variable.
As demonstrated in the study we released back at the end of January 2011, CPC by category ranged from 12 cents for tabloids to $1.27 in healthcare, and CTR was the primary determinant of the CPC. In that study, we found that the relationship between CTR and CPC is minus 0.846. In other words, it’s a direct relationship. So poorly managed ad campaigns burn out quickly and drive up the CPC, though CPM holds constant.
With the recent release of Sponsored Stories on January 24th of this year, we’ve seen CPC prices trend down significantly– in some cases by several orders of magnitude. For example, Gordmans was able to get nearly 0.400% for Sponsored Like Stories as described here, and we’ve seen this consistently for dozens of other companies.
Knowing that CTR is the primary determinant in cost per click, what are the other factors?
- Increased market competition: This is the commonly cited as the excuse for increased costs. But there’s no direct way to know how supply is pacing against demand. Anecdotally, from our client campaigns, we believe that CPMs as well as CPCs are flat to slightly decreasing, driven by the fact that there are new ad units (for higher CTR) and that most advertisers are getting smarter about how to create effective ads. Most agencies, from audits we’ve conducted, are not rotating ad creatives, allowing them to burn out quickly, causing CPCs to rise. The most guilty parties are traditional PPC management companies, who are used to the “set it and forget it” model in keyword-based search.
- Quality sometimes costs more: As advertisers are learning how to target, certain interests and demographics are being bid up. Three years ago, we used to be able to buy any traffic at the same price. But now, if you target college grads, Facebook employees, or premium status members of frequent flyer programs, the cost can be a LOT higher. We used to be able to target Facebook employees at 20 cent CPMs. Now it’s over a dollar. But we don’t see this increase applying uniformly across interest targets– mainly just categories adored by internet marketing companies who wish to bombard journalists, Facebook employees, and anyone with a senior sounding job title.
- Your goals: Creating exposure, gaining fans, generating leads, and driving revenue are different goals that require different creatives. The CTR on a sale-oriented ad vs an engagement ad is usually far lower. Likewise, driving to a page outside Facebook for the purposes of checkout is lower than keeping them on a Facebook page. But these tradeoffs of lower CTR for a higher CPC are often acceptable.
- Your connectedness: We demonstrated that when there was social endorsement in an ad, CTR was often double. The more Facebook grows, the greater the likelihood of fan intersection, even without their aggressive pushing of ad units that specifically leverage endorsement (Sponsored Like Stories). As average fan bases grow, it’s reasonable to think that the “peer pressure” effect in ads could be much stronger if advertisers understand how to wield this weapon.
The bottom line is that traffic costs should be benchmarked by CPM prices for the same inventory over time, since CPC prices are affected primarily by advertiser sophistication, targeting methods, and other factors that outweigh base CPM traffic costs. When your ads suffer a 50% decrease in CTR, then your clicks cost almost double, as we’ve shown.
In an eCPM auction, revenue is based on inventory vs CPMs generated on that inventory (not on clicks). As we’ll demonstrate below, Facebook has growing inventory and is earning a greater rate on that inventory. The primary factors are increased advertiser competition, higher CTR ad units, and the proportion of premium vs untargeted inventory available.
The raw inventory figures are easy to publicly determine, broken out by country and mobile or non-mobile (no ads on mobile yet). For the purposes of this discussion, we are not including Facebook Credits or Premium inventory, though we have consistently seen advertisers be disappointed with premium inventory when any form of measurement beyond pure impressions are applied.
Thus, the question is how CPMs are faring and what affects these numbers. Using the same advertiser base of our study from January 2011, we saw CPMs go from to 25 cents in 2010 to 32 cents in Q12011, which is a 28% increase. Though this is a relatively decent apples-to-apples comparison, we do have to note that these advertisers have become more sophisticated in their targeting, shifting goals from pure impression to metrics such as CPC, Cost Per Fan, and even Return On Ad Spend. Because the CPM paid is not affected by CTR, the CPM is a true reflection of Facebook’s monetization efforts. And increasing advertiser sophistication nearly always results in higher CPMs.
Facebook has been aggressively pushing users to like objects within Facebook and in the open graph. The effect is to convert remnant inventory that is worth 20 cents to interest targets that can worth upwards of a dollar. People are identifying who they are, who they are friends with, and what they like, as evidenced here. We believe this one factor to be an order of magnitude more important than inventory growth.
We have some ad campaigns that have run for more than a year, which can provide a partial baseline on pricing over time. It’s only a partially accurate view, since these units will suffer from ad burnout, leading to a distortion of the CPC, but not the CPM. It will, however, give us a sense of general competition for various types of traffic. We notice that traffic across the board (on a CPM basis) is getting more expensive, not because of the general growth of advertiser demand, but because users are eligible for multiple auctions, creating greater competition. This is akin to the curse of a dentist trying to advertise on Google in NYC. He can target his campaigns just for his borough or zip code, but the other dentists who don’t know any better are targeting the whole NYC area, which overlaps with his segment, driving up prices.
The more interests per user and the more advertisers bidding, the greater this effect. This effect is especially pronounced for ads that have user endorsements in them (Sponsored Likes and friend-of-fan ads), since the available audience increases dramatically with fan growth, even when interest targets are left unchanged.
There are some things we’d like to know, but for which only Facebook has answers:
- How much premium homepage inventory is actually sold and at what discount on the ratecard? We remember two years ago that fill rates were barely double digits, causing Facebook to run house ads in that space and to sell that inventory to affiliates at up to a 5x discount from the $10 CPM ratecard.
- How have average CTRs fared over time? We know that premium inventory and untargeted ads suffer badly here, but we don’t know what proportion of total inventory is represented by such campaigns.
- How many interest targets are available now? Facebook has not allowed us to target every potential interest that would technically be possible. For example, workplace targeting is not possible on most small businesses. You can target BlitzMetrics and Webtrends folks, but most social agencies and small businesses are not big enough to show up. 18 months ago, Facebook told us that there were 300,000 interest targets in the system. We don’t know what that is now, but it’s certainly less than the tens of millions available in search.
- How are ad sales faring internationally? We see traffic in Brazil, Turkey, and some of their top countries continue to be available for pennies, just like in late 2007. Our advertisers are primarily buying US traffic, so our numbers are not representative of global ad performance.
Net-net, we think Facebook is probably significantly ahead of plan, but has a nagging problem of how to teach advertisers to run effective social campaigns, partnering with ad agencies (who are the most egregious offenders of poor social ad buying tactics, and educating the small businesses on ad campaigns (even more difficult than the F100 and agency advertisers). Perhaps canned ad units via Sponsored Stories and other pre-made campaigns are the answer. On the supply side, we see them performing marvelously at generating traffic and getting that traffic to identify themselves by hitting the like button.
The hurdles that we mentioned here are not to throw cold water on Facebook advertising. Rather, we are saying that Facebook can generate $2 billion in revenue with everything as broken as it is. They are not even running ads on mobile devices yet– intentionally. When they turn their attention from user growth to monetization, you’ll see these gaps rapidly close in a way that the folks who blindly project curves forward would never predict.
I may have to eat these words, but I believe by the end of 2014, Facebook will surpass Google in revenue– just 3 years away. What say you?
Dennis Yu is the Chief Executive Officer of BlitzMetrics, a firm that specializes in managing Facebook ads and pages for major brands and local service businesses. He is an internationally sought after speaker and author of many Facebook articles, including, How To Rank #1 In Facebook Search In 60 Seconds For Any Term, Facebook Optimization Tips for Hosting companies, and Online Lead Generation.
You can read Dennis Yu’s full PubCon biography here