6 Mistakes You’re Probably Making When Interpreting Facebook Ad Data
You’ve probably been doing Facebook ads for a while now. I’ve had the luxury of seeing the interface evolve since early 2007, back when it was called Flyers– CPM-based ads targeted by schools. By now, I’m used to things getting switched without warning. All of those switches have had some effects on advertisers, one of them being the widespread misinterpretation of data. Here are six areas that require special attention — and mistakes that get made otherwise:
So you want to know how many fans you got from your ads. Facebook used to call people who became fans of the ads themselves “actions”– more specifically “in-line actions” if you ran an ad that was for driving fans. This was before Sponsored Stories. Then they introduced a metric called “connections” that measured the number of “actions” (which occurred immediately– you could see it within 60 seconds) plus folks who clicked through the ad and became fans later. The “connections” were calculated with a 2-3 day delay, which royally messed up our Cost Per Fan calculations.
You see, we often charged on a performance basis. For example, we might charge $1 per fan for 100,000 fans on a particular advertiser. If we know that the number of connections was usually 30% greater than just the actions, we could run to about 77,000 fans and hope the rest would trickle in to hit 100,000 fans by the time the ad reports updated.
The same issue is still true for Facebook Insights, which is typically on a 2-4 day delay.
Nowadays, Facebook is reporting on the metric “actions” again, but it stands for nearly any type of click– not just what you’re optimizing for. So if you’re using to reporting as you’ve done the last couple years, you might mistakenly be over-reporting your ad-driven fan numbers.
We used to often get fans for a tenth of a penny (an extreme example), because the traffic was 10-cent CPMs all day, and we’d get 80% click-to-fan conversion rates. Now it’s more like 60 cents in the US and 30-40% conversion rates. We’ve not found viral likes to really help boost overall fan counts except for in entertainment.
So if you want to get truly ad-driven fans, you can mouse-over the ad unit to get the slice of the pie for fans, or you can download the ad performance report to get what are now called “page likes.”
And here you thought getting fans on Facebook was easy!
Ignore the latest reports that other agencies publish about the cost per click rising.
Wrong on so many levels.
Because Facebook operates on an eCPM auction, your CPC is inversely related to your CTR, just like on Google. Remember when Facebook gave us their Quality Score three years ago, but then hid it from us? It’s still there– just not shown to us anymore. If your CTRs are declining (perhaps because the agency isn’t watching the ads carefully for burnout), then don’t take the excuse that traffic is now more expensive. They’re mismanaging your campaigns.
Your CPC should be getting better over time (except compared with 3 years ago). Why? As your fan base gets bigger and more engaged, you have more opportunity for social connections. If your fan base is small, then you’re just not going to be able to run Sponsored Stories, which will get you 2X the CTR. And if you can run mobile Sponsored Stories, you will often get 5-6% CTR.
Comparing against “industry” CPCs is useless, anyway. It’s not like Google, where you can use PPC tools to spy on the competition or see how many people are searching for a particular keyword at an estimated price. And you can’t compare ads that are for fan growth vs those for engagement or conversion. Of course, an ad touting low interest rates for a credit card company will perform more poorly than one about the Visigoths asking you what’s in your wallet.
We published research on 120 billion ad impressions we ran on Facebook last year during a 6-month period. CTR and CPC varied wildly, but now that we’ve seen half a trillion impressions, the same law holds true with how relevancy and CTR affect CPC.
3. Cost baseline
The real baseline to measure your traffic cost is to look at the CPM you’re paying by placement, even though you’re still bidding CPC.
Mobile will be about $10 CPMs, but it’s so worth it, even when backed out to a CPC. So don’t get hung up on CPMs either. You can’t compare a newsfeed impression (which is huge) against one tiny impression sitting among 6 other ads on the right side.
If you’re worried about bot click fraud driving tons of fake clicks, I wouldn’t be. Facebook has caught such guys in the past and promptly refunded the dollars. You may have noticed ad credits in your account, which are usually small. But more likely, they are just not charging you for these fake clicks.
Guess what? They’re missing! Facebook now shows “reach,” which is the unique number of folks who saw your ad. Arguably, reach is more important than raw impressions, since it’s easy to squander a ton of impressions against a narrow audience.
Reach x frequency is your impression figure. Half of Facebook users log in daily, so you better watch out for ad burnout. Above a frequency of 10 and your CTR will suffer from ad burnout. You can still download the reports, of course, but if you’re doing live optimization, you don’t have time for that.
Promoted Posts (see screenshot below) make sure your fans see your messages. I’d recommend this for small business owners who have exciting content. Hence, the restaurant example. But if you’re a brand, you’ll likely see far more impact by running page post ads against finely tuned targets bid carefully. Some will argue with me, but I believe that most promoted posts shouldn’t really rely on CTR, since newsfeed impressions carry so much power. In the case of La Gondola here, the business owner swears by the ROI– he knows when it brings customers into the store, even if they don’t click on the post.
In the UI, you see it at the campaign level but not at the ad level. So you have to do some quick math (clicks x CPC) to get there. Budgets are set at the campaign level anyway, but we like to see how ads within a campaign are spending.
There are no ad groups in Facebook, of course. Some folks advocate creating one ad per campaign (generating MANY campaigns) to be able to tightly control spend at the ad level or force a version of ad rotation, but we see that as often excessive. We do believe you should have a “production” campaign and test campaigns, since a new ad in a high budget campaign can suck the daily budget away from all the other ads.
This is the most difficult of metrics to measure, not because Facebook took away conversion tracking two years ago, but because word of mouth is nearly impossible to track. Google has view-through conversions to track folks who saw your ad, didn’t click on it, but converted later anyway in another channel. But in social it’s worse than that, since you’re not talking about one person’s click path. You’re looking at how other people influenced that user. When people are talking on the phone or at the dinner table, you can’t cookie those conversations and award credit.
Last-click conversions in your Omniture and Google Analytics will definitely mis-report the power of your social. Even the new multi-channel funnels in Google Analytics will not capture it accurately, since they don’t track conversations until it hits the website. Facebook, regardless of what the developer relations folks say, is designed to keep you on Facebook. So if you’re running Sponsored Stories to get in a user’s newsfeed, those interactions aren’t going to make it into your Google Analytics.
Nobody has a great answer here yet, unfortunately, but there are 45.18 billion reasons, as of Friday afternoon, Aug. 3, at 4 pm ET.
– Dennis Yu, BlitzMetrics