How Open Rate Really Works
There's a lot of disagreement about what the thing called "open rate" actually measures and how it works. Obviously, at some basic level we all understand it purports to quantify the number of recipients who actually open our email. But what about text messages? How about the preview pane? What about a user who checks her email offline?
Well, I'm here to set the record straight and put these questions (and more) to bed -- once and for all.
Method to the Madness: Counting Opens
Gauging the number of opens your email receives is achieved by embedding an image into your HTML message. The image isn't actually distributed in the message, but is rather referenced from within the message, as with all HTML documents. Each time someone opens that email, he loads up the HTML document and renders it in his email client.
As the email is rendered, the recipient's email program must request a copy of that image from your server and attempt to load it in the message. Count the number of times that image is requested and -- voilà! -- an open rate.
If you send a message to 10,000 recipients and log 5,000 requests for your embedded image, you can safely say you've got an (approximate) open rate of 50 percent. Strictly speaking, the open rate isn't the actual number of people who opened the email but the number who requested the embedded image.
Preview Panes and Other Banes
Armed with that understanding of open rate's basic mechanism, you should have no problem with the preview pane. If someone reads your message in her preview pane and the preview pane renders HTML (e.g., Microsoft Outlook 2000) -- boom -- she'll request your embedded image and register an open.
If the reader moves down through his inbox and passes your message without actually reading it, does that record an open? No. If the message isn't loaded in the preview pane, it won't give you an open. Even if the user stops momentarily on the message, long enough so it starts to load in the preview pane, but he moves on before anything is requested from your server, you'll get no open. Only if he stops long enough to load the entire message (and your embedded image) will an open be recorded.
Consider putting the embedded image you're counting at the end of the message. That way, it'll be the last thing rendered. It will count as an open only for people who've had the message open for a few seconds at least. I call that a "true open."
Here's a rundown of other banes:
- Text messages. Obviously, text emails can't contain HTML or images. Thus you cannot measure open rates for them. If you want to measure the open rate, HTML is your only choice.
- Dial-up/offline readers. If your recipient isn't online when she opens your email, she won't be able to pull down the image you're counting. No open rate there, either. (She won't be able to pull down any images, for that matter. If your only call to action is a graphic "Order Now!" button instead of a text link, you have bigger things to worry about.)
- Multipart messages. If you send multipart messages (text and HTML included in the same email), only the HTML part of the message can measure opens. This introduces a bit more ambiguity to the numbers. If you send a multipart message to 10,000 recipients, you assume most people will decode the HTML message, and some small percentage will decode the text message. If you receive 5,000 opens, it means of the people who could read HTML, 5,000 opened. You don't know how many text readers opened. Could be 100 percent, could be 10 percent. Your open rate will underreport text opens. Given the small number of text-only readers, I suspect this has a statistically insignificant effect on open rate.
- Firewalls and mystery opens. At last week's ClickZ Email Strategies Conference, a few speakers mentioned some corporate firewalls can do things such as strip all the images from HTML email (to conserve bandwidth). Obviously, this would eliminate any opens from those recipients. On the other hand, some proxy servers can precache all the images in HTML email for all recipients companywide (again, conserving bandwidth). This could register false opens for people who simply hit delete. The impact of these two scenarios is probably pretty small. Because one would push numbers up while the other would push them down, I suspect they cancel each other out.
- JavaScript. To protect against viruses, the preview pane generally does not execute JavaScript. Make sure any images you reference are done using plain, old vanilla HTML. Most email delivery systems have this issue covered, but watch out for using banner advertising code (often JavaScript-indulgent) in the preview pane to count opens/impressions.
Don't Worry, Be Happy
With so many variables and unknowns, it's tempting to throw up one's hands and kiss open rate goodbye. Don't do it! Yes, it's true a small number of opens will be false or duplicates. Yes, it's true there are a handful of cases where opens are underreported or fail to be reported at all. As long as you treat the number as a ballpark figure, not some unerringly precise scientific standard, you should be in good shape. Think of it as "(approximate) open rate."
Remember: Probably the most important thing you can learn from open rate is how your behavior influences it. Most other variables (firewalls, dial-up, etc.) remain the same from mailing to mailing. What willchange is your subject line, your subscriber base, and all the things you can do to help increase your open rate. Watching the pluses and minuses weekly is where the action is. Growth is probably more important than baseline. (Approximate) open rate is just perfect for that.
Got a question? Think I'm full of it? Let me know -- send me email!
Subscribe to Newsletters Subscribe to RSS Feeds 
Stumble It!
Digg this!
Add to del.icio.us
Article Archives by Edward Grossman
Shopping for an E-Mail Vendor, Part 2: Your RFP Is a Recipe for Success - Mar 21, 2003
Shopping for an E-Mail Vendor: Part 1 - Feb 21, 2003
Bounced E-Mail? Deal With It! - Jan 17, 2003
The Myth of Multipart - Dec 20, 2002
More article archives
We want to know what you think about
Edward Grossman’s column, "How Open Rate Really Works"
Rant. Rave. Voice your opinion.
Archive
