June 18, 2019

The 7 Most Commonly Misunderstood Marketing Metrics

Dan Kerpen

Marketing has created a dizzying array of abbreviations and somewhat esoteric terms. Unfortunately, many of the simplest sounding terms also have technical definitions that aren’t aligned with common-sense readings of the words involved. So, we’ve created this guide to the common questions you get about social metrics and how to answer them—soon, when you hear these kinds of utterances, you’ll know just what to say to get some clarity.

“How can my average session duration only be 0:01 in so many cases in Google Analytics? And my bounce rate is nearly everybody? That’s all junk, isn’t it?”

Average Session Duration is maybe the most widely misunderstood metric of all time. Bounce Rate competes for this distinction, though. And they are closely related.

In Google Analytics, whose mighty little snippets sit on somewhere between 55-85 percent of websites, a session always begins with the first tracked event or pageview. It ends when a user-defined period of time (default = 30:00) elapses without an event or pageview occurring. At that point, the time between the first and last recorded interactions is calculated as the session duration. Spelled out in more detail:

  • When only one pageview or event is tracked in a session, the session length is the duration between the first and last (i.e. the same) interaction. In Google Analytics, that is rounded up to 0:01, even if in fact the user spent time reading or watching content on your page.
  • The above also exactly defines a “bounce.” For many situations, a “bounce” is expected behavior. Imagine, for instance, clicking a link to a news story or a blog. Do you normally click around a lot, or do you consume the content you came for and then leave?
  • The more on-page events you track, the less likely you are to have a great deal of bounces. This can include tracking button clicks or scrolling, or even a “pingback.” That means authoritative benchmarks are hard to come by, because every configuration is different.

As always, your best bet in benchmarking is to measure against yourself, and try to get continually better over time. As long as you are consistent and clear on definitions, you should be able to move in the right direction.

“Google says there are only 6 conversions there, and Facebook says 12. So Facebook must be lying, right? Or they aren’t counting the same conversion!”

This is frustrating, but there is no meaningful sense in which anyone is trying to deceive you. The systems don’t play nice, and there are good reasons for them to diverge in almost all cases. Here are a few things to keep in mind as you sort through this:

  • By default, Google Analytics measures via a “last click” attribution model. Both of those words are crucially important, because (a) “last” means the final source gets 100 percent of the credit in typical cases, and (b) “click” means it only sees click interactions—not views.
  • By default, Facebook counts all conversions that occur within 28 days of a click. It also counts all conversions within 24 hours of viewing an ad.
  • Additionally, whereas Google Analytics very rarely tracks users accurately across devices, Facebook does so with tremendous accuracy. They tout this in most explanations of people-based measurement.

So, which is right? Really, both. They just measure different things. And which you should lean into depends on a lot of factors, like how much cross-channel advertising you’re running and what a typical path to conversion is for your business. In any case, this can happen even when the way you’ve defined a conversion is exactly the same.

“What’s my potential reach? That’s the same as audience size and projected reach, right? I need to know exactly how many people we reached and will reach!”

The way we think about it, there are actually four different metrics described here:

  • Potential Reach: This is the number of people an ad set could reach, adjusted for all targeting, placement choices, and the number of people recently using the chosen networks.
  • Audience Size: We use this term to describe the widest version of the above-mentioned parameters: opted into all placements, with a broad version of the target.
  • Projected Reach: This is a function of Audience Size and—most importantly—budget. By far the easiest way to make this calculation is to look at Reach in similar campaigns.
  • Reach: This is an estimated number provided by Facebook, describing actual unique persons reached. It’s the only version of this metric that is in the past tense.

The main reason this matters is that it’s entirely too easy to talk past one another when Reach comes up. Asking a couple quick clarifying questions helps get everyone on the same page.

“What’s this ad recall number? ‘Cost Per Estimated Ad Recall Lift (People)’? Why is it so wishy-washy?”

As alluded to above, a lot of Facebook’s metrics are estimated. That isn’t a bad thing. It means it’s calculated in part by sampling and modeling. No need to fear; Facebook has excellent viewability partners, and plenty of evidence that its ads move and motivate viewers and clickers alike.

Recall is actually a pretty amazing thing to be able to optimize for. It’s calculated by tracking and storing how long a user usually spends on a post, and optimizing for interactions in which users spend statistically significantly longer periods on the post. Longer time spent correlates with higher ad recall. Understandably, this metric is estimated in several ways, since no actual lift study is done in most cases.

We often simplify this number by calling it “cost per lift” or even “CP Lift,” to keep things brief. If you need to see more about how the sausage is made, lift studies start at just $30K now. And if that’s too rich for your blood, this is probably the wrong objective for you! The “Video Views” objective is probably a better fit.

“We want clicks, so we can get hits!”

There is a definition of “clicks” that persists in part from a shameful era of social media advertising history. And while that’s mostly faded, be careful of snake oil purveyors. As measured within almost every social media platform, “clicks” aren’t always “link clicks.” The column named “clicks” may contain a wide variety of unimportant actions. Always check the definition. Here are the overly broad versions across a few key platforms:

  • Facebook: Now called Clicks (All), this counts link clicks; clicks to the profile or profile picture; clicks to view reactions, comment sections, or the list of users who shared; clicks to expand media; and clicks to directly “like” the page.
  • Twitter: Clicks includes clicks on links, hashtags, image or video cards, the follow button, the favorite or comment button, and clicks on your username or profile picture.
  • LinkedIn: Clicks includes clicks on your ad content, company name, or logo, but not clicks on your likes, comments, or shares.

And hits? That’s a term of much historical confusion. Technically, it describes file download requests made to your web server. Our official advice: Just think about what you really want, and talk about that instead! Generally, pageviews or sessions are a better soft metric.

As for what we mean by “clicks,” unless otherwise specified we always mean link clicks. Nowadays, it’s just easier that way!

“Video views, video plays, video watches—can we just agree to use one phrase from now on?”

That’d be nice. We long for the day when everyone agrees about these words. In the meantime, here’s the skinny from the Facebook side of things:

  • Video Views refers to the number of times your video played for at least n seconds, or for nearly its total length if it’s shorter than n seconds. (On any given impression of a video, they’ll exclude any time spent replaying the video. But if they show the video to that user again and that user watches to n seconds, that’s a video view and one user can have multiple views.)
  • Video Watches is used to describe various points in your video’s playback pattern. So video watches at 25 percent and 50 percent count the number of times a user reaches the 25 percent and 50 percent points, respectively, even if the user skipped to that point.
  • Video Plays counts the number of times your video starts to play, excluding replays. There is no publicly known minimum threshold for when a video has started. It’s likely very fast.

Since Video Views are what the network actually optimizes for, we almost always recommend sticking with that one. Exceptions crop up when measuring view duration in depth is important. Video Watches are very good for that.

“Engagements are likes, comments, and shares, right?”

Although likes, comments, and shares are types of engagements (and what people usually think of when they hear the “E word”), so are all kinds of clicks (see above) and other actions like voting in a poll or browsing through an album post. On Facebook, even lingering on a video for three seconds counts as an engagement.

To be clear and to ensure historical records make sense, we always recommend that videos be promoted for something other than engagement. For further clarity, it’s sometimes a good idea to count key engagements and measure them separately from all engagements. And, more often than not, brands should take a hard look at optimizing for lift or video views when they’re seeking upper-funnel activity.

Keep up with the latest marketing metrics and social media updates. Subscribe to our weekly newsletter.

Tags: Data & Analytics, Facebook, LinkedIn, Paid Media, Tools, Twitter

Stay in the know.

Social smarts delivered to your inbox every week.

Want to Work With Us?

Let’s Talk