Sep 17, 2012

This Just In: Subject Line Length Means Absolutely Nothing

There’s a lot of sage wisdom floating around on what constitutes a good subject line. On the helpful side, there are articles like this one in our Knowledge Base that emphasize clear communication, personalization, and word choice. The purpose of a subject line is, after all, to tell the reader what’s in the email.

But for every article that champions communication, there are another 10 that champion gimmickry to get your readers to open. Stuff like using unicode characters (HEARTS, STARS, HORSESHOES, CLOVERS, AND BALLOONS!!!!!!) or this article wherein the report from an ESP claims that the longer the subject line, the better your campaign is going to do. Is your audience some monolithic blob that mindlessly clicks long sentences? In the absence of good data, myths will abound, but we have lots of data in MailChimp’s Email Genome Project, so let’s take a look at subject line length and its effect on engagement.

High-Level Results

We pulled information on 12 billion email sends from earlier this year and analyzed subject line length versus open and click rate. The results are straightforward:

Chart showing that as your subject line gets longer, nothing happens.

As your subject line gets longer, nothing happens. Cheers.

A Tale of Three Users

OK, so we’ve established that there’s not a one-size-fits-all rule for subject length. More is not better. But that doesn’t mean that there isn’t a sweet spot for you. Let me show three quick examples from actual users.

Here’s a user whose campaigns with longer subjects were correlated with lower open rates:

And one with high open rates:

And one where it means not one single damn thing:

Most users examined in the study (and there were tens of thousands of them examined), fell into this third category: no statistical link. But you can easily check which user you are by just exporting the data from MailChimp. Might I suggest running an A/B split on a few campaigns where you vary subject line length and see what works for you?

The Mathematical Rant Part of the Blog Post

Granted, I’m just a mathematician and not an email-marketing expert, but I’ve got some theories as to what the data are showing us in the graphs above.

I’m sure you’ve read the phrase "correlation does not imply causation" before. Hell, there’s a Wikipedia article about it. Well, this whole subject-line discussion falls into that trap. Even if there existed a pretty graph showing a relationship between subject-line length and campaign performance, that doesn’t mean that by padding your subject with purple prose you’d suddenly be swimming in a sea of opens and clicks.

No, that relationship (like what is shown above for users X and Y) is likely caused by a third variable. For instance, what if I sent out a weekly campaign with the subject line "Christmas sale on processed meat logs full of fat and salt" and didn’t get any opens, so I changed it to "Sale on summer sausage." Well, "summer sausage" sounds a lot more appetizing than "meat log." I bet I’d get better performance. Would that be due to the lower character count? Only indirectly. Rather, it’s due to my astute change in diction. Hemingway would be proud.

But this rant aside, the truth is that we found, on the whole, not only no causation, but even no correlation. Over six billion of the emails we studied showed little or no correlation between performance and subject length (coefficient of determination less than 0.12).

Of those users who managed an r-squared above 0.5 for subject length versus open rate, 40% had some kind of improvement as their subject lines got longer, and 60% showed the exact opposite. That kind of split says to me exactly what I’ve been arguing: additional forces are at play in each user’s account that’s causing these relationships.

So, Where Do I Go From Here?

Your audience is chock-full of individuals with different reading habits, interests, and demographics. Maybe my audience is full of Apple fanboys and every one of them reads my newsletter on their iPhone. Well, then the subject line they see might need to be shorter for their small screen. Or maybe my newsletter is geared toward businessfolk who mostly run Outlook. In that case, maybe a longer subject is more acceptable. Within MailChimp, you can easily gather this user-agent data and many other juicy tidbits that flesh out what your audience looks like.

You could use Wavelength (just launched!) to tell what other things your audience reads. You can geolocate your readers. Find out if they’re night owls. Get social data and find influential readers. Point is, don’t just settle for what some ESP recommends (not even us). Do the math and find out what works for you.


See Also: