Measuring inbox deliverability and image blocking using clicks and opens

Latest posts | Feed | | By Mark Brownlow

email symbolWhen you send an email, it has three possible fates:

1. It never reaches the recipient's email account
2. It reaches their account, but lands in the junk or spam folder
3. It arrives in the recipient's inbox

How can you tell what percentage of your emails actually make it to the inbox?

You can't.

Or can you?

We can certainly get a good feeling for inbox deliverability using the techniques outlined in this earlier post: intuition, domain segmentation, inbox monitoring services, proxy monitoring measures and deliverability testing tools.

But an actual number for inbox deliverability? Tricky...

Ask a different question


Let's change the question slightly: what proportion of your emails are seen by subscribers?

If someone sees your email, then it must have reached their inbox (notwithstanding those few who diligently look through their spam folder).

Theoretically, the open/render* rate tells you how many people "saw" (however fleetingly) your email. When your email appears in front of the user, it triggers a tracking pixel and records an open.

So maybe we can use the open/render rate number as a proxy for inbox deliverability?

Unfortunately, image blocking rather cripples that idea.

We can never know the real open rate, because we don't know what proportion of the email clients or webmail accounts used by subscribers block our images and stop an open from recording.

And some people will only see the text version of our email. No pixel is displayed, no open is recorded.

So we're stuck again. Almost...

Use clicks and opens to measure image blocking and text-only displays


One technique used by e-newsletter publishers in particular is to infer image blocking and text use from clickthrough statistics. That in turn lets you calculate an adjusted "real" open rate.

You need three numbers to get there:
  • Unique number of opens (number of delivered emails that caused a tracking image to display at least once)
  • Unique number of clicks from opens (number of delivered emails that generated at least one click and caused a tracking image to display at least once)
  • Orphan clicks (number of delivered emails that generated at least one click, but no tracking image was displayed)
Those orphan clicks are people seeing your email as text only or with blocked images, but still clicking on one or more links.

Let's say your unique opens was 20000, unique clicks from opens was 1000 and your orphan clicks 150.

What can we extrapolate from those three numbers?

We know that 20000 recorded opens produces 1000 clicks, a ratio of 20:1. So among those recipients for whom an open was recorded, 5% click.

Let's apply this ratio to the orphan clicks...

To get 150 clicks, our ratio tells us that 3000 recipients must have "seen" the email, either as a text version or with blocked images.

This tells us that a total of 23,000 recipients actually "saw" our email (however briefly). If our campaign report tells us 25,000 emails were delivered then this implies that 92% of our emails likely made it to an inbox.

We have a number for inbox deliverability!

But we also have another number: about 13% of our subscriber base sees text-only or blocked image versions of our emails. Another interesting metric.

Wait though. Before we get too excited, there are problems with this technique.

Problem: click rates are not the same for all email versions


The biggest issue is the assumption that people are just as likely to click on a text or blocked-image email as they are on an email showing all the images.

You can argue about text versus HTML and all that, but if images are making any kind of positive contribution to the email experience, then this assumption is false.

(If images aren't making a positive contribution, you might ask why they're in the email.)

In most cases, you'll underestimate both inbox deliverability and the extent of image blocking and text-only displays.

The better you account for image blocking in your copy and design - for example by using alt attributes - the more accurate the assumption and the resultant figures.

Problem: not every sighting is an open


If an email lands in an inbox, it won't always cause an open to trigger.

Consider webmail services, for example. Their interfaces let you mark emails to delete without opening or previewing anything: an email reached the inbox but never displayed. And our adjusted open rate calculation cannot account for such cases.

So, again, you underestimate true inbox deliverability when using open rate as a proxy.

Problem: where do you get the numbers?


Another issue is finding those three numbers.

It's not uncommon for ESPs to automatically record an open for every click, regardless of whether the tracking pixel displayed for the email in question. After all, if someone clicked, then they must have opened the email.

So it may be impossible to split your clicks between opened and non-opened emails.

So, we have some new metrics here, but they're far from perfect. Like any stats, they should be treated with caution. Don't base decisions on any individual number alone, but incorporate that number into your wider understanding of just what's happening to your emails.

*If you want to read up on open rate methodology and the arguments around names and definitions for the metric, here's an open rate overview, and here are two links on terminology and definition issues.

More on statistics and deliverability | Tags: , , , ,

[This post brought to you by Campaigner Email Marketing]
Permalink | April 02, 2009 | 0 comment(s) - add yours!
Get posts like this: as an RSS feed | biweekly email | via Twitter
Sign-up for the Email Marketing Reports NEWSLETTER
Twice a month, free, packed with email marketing advice and all the posts from this blog.
Email:      First Name:     
    More info and sample

0 Comments:

Post a Comment