Oct 18, 2012

What Good Marketers Can Learn from V1aGr@ Spammers

The other day I wrote a post about subject line length and how it doesn’t influence open rates. But for most folks, open rate isn’t the issue—click rate is. And click rate arguably has something to do with subject line, and a lot to do with the content of your campaigns.

So the question is this: Once I’ve got someone looking at my campaign, what do the numbers say about how I should design my campaign?

The other day someone said to me, "You know who the best email marketers are?"

"Who!?" I asked with breathless excitement.

"The really bad spammers," he said. "Their emails have a clear call to action and a single link. They’ve refined their communication through years of experience."

It’s an interesting thought, isn’t it? To verify, we took a look at the bad users we’ve shut down over the years, thinking that surely there’d be a magic engagement number that all the good people fell above and all the bad people fell below. Much to our chagrin, the bad users we’d shut down regularly had better engagement than some of our good users. Uh oh. Maybe it has something to do with this campaign design discussion.

Are the spammers simple, lazy, or brilliant? Here’s a screenshot from my very own Gmail spam folder:

We have a clear subject line (although we know that ALL CAPS is detrimental to engagement). Then we get inside the email, and there’s no text, no pictures of some teary-eyed man suffering from ED. Just a single, Polish link. I know exactly what to do—click the link! Is this how email marketing should work? Is this what it takes to maximize clicks from those who open?

Let’s go to the numbers.

Specifically, let’s look at click rate as it relates to three values: word count, image count, and URL count. As part of MailChimp’s Email Genome Project, we’ve been saving off all the content we send, so for this study I was able to pull these summary values from 5.5 billion emails from earlier this year across our large user base (we’re at over two million now, and ticking up at 5+ thousand signups a day).

A Note about Methodology

We’re going to look at how word count, image count, and URL count affect clicks over opens in a campaign. I don’t care if someone who didn’t open doesn’t click, because they didn’t look at the content. To avoid small numbers in the denominator, I only pulled campaigns that had at least 1,000 opens.

Concerning images, I didn’t want to count small ones, so I placed some limits on what I qualified as an image. It had to be more than 49 pixels high and wide, and the image had to total more than 10,000 pixels. This was primarily meant to exclude pretty little borders, as well as teeny-tiny Twitter and Facebook icons and the like.

Also, there aren’t a lot of people who send 2003-word campaigns exactly, so I’ll be lumping url counts, image counts, and word counts into percentile buckets sometimes. Because I’m using percentiles, each bucket will have the same number of campaigns in it.

Let’s Start Simple

The first thing we should check out is just how these three values affect click rates individually. Let’s look at word count first:

I’m not seeing a whole lot there. As word counts increase, click rates go all over the place. There must be more to it than that.

Now let’s look at image counts:

Now that’s better. What do we see? Well, the first thing I notice is this dip to the left. Text-only campaigns (hello spammers!) do better than those with 1-9 images. The next thing I notice is that once we hit two images, it’s all uphill from there. So more images means more clicks? Perhaps more images means you’ve put more care into designing your campaign, or it might mean you’ve got more relevant content to send. Or maybe we’re all just visual learners, and we really like pictures.

What about URLs? They’re the thing that’s getting clicked anyway, so what happens if you put more of them in a campaign? Let’s see:

It’s almost all uphill! But that makes sense if you think about it. Let’s assume that most users are sending vaguely relevant things to their audience, such that each link has a non-zero probability of being clicked. In that case, if we assume that lots of URLs don’t scare people, then the clicks should go up as the URLs go up.

Let me clarify something here. If I padded my content with useless links, that’s probably not going to help me, because their probability of getting clicked is low to none. Instead, this result shows that people with lots of relevant content do better than people with little relevant content. This is indeed the type of result that makes services like Outbrain tick.

The spammers lose big here, because while they have a clear call to action, they’re only giving the reader one piece of content to engage with. I like to call this, the "whole bunch of Mums" problem.

Going Deeper: It’s All About Balance

I wasn’t happy with that word count result above. It didn’t stand to reason. If some dude sends me a campaign that’s got excerpts from Anna Karenina in there with a link every couple hundred words, then I’ve got to search for them pretty hard to click them. I might get turned off, especially if the long excerpts are from Levin’s farm scenes. Boring.

So let’s move beyond two-dimensional thinking and look at ratios of these values versus engagement. We’ll consider the value of "Words Per URL." WPU. Pronounced wuh-poo. How much copy should I put around the link I want someone to click to make my campaign successful?

The data above provides a much clearer picture than the first word count graph. We can see that 8-12 words per URL is a sweet spot. Get to the point. Be concise. Not spammer concise, as in zero words, but straightforward nonetheless. It’s really interesting to me that engagement falls off a cliff right around 100 words per URL.

What about images per URL? What’s the sweet spot there?

If we look at this data, we see the reverse of what we saw above. More images all the time is not good. We only put more images in campaigns that have more URLs. In other words, more content all around. If I’ve got fewer than 10 URLs in my campaign, then just a few images are plenty. Don’t overwhelm the content with pictures unless you’re selling posters or something.


People on the internet seem to like recipes, and while I hate being prescriptive, here it goes:

1. Provide your readers relevant content

2. Add more relevant content if you’ve got it

3. Accompany links with concise statements

4. Add images. Not too many, just enough for flavor.

5. Cook 45 minutes on 350. Cool before eating.



More Email Research: