Investigating the Impact of Gmail Email Image Caching

Recently, Gmail rolled out to production, the caching of all emailed images on Gmail servers.

Many email marketing companies, including Constant Contact, have been wondering what impact this would have on the reporting of open rates for its customers.

The following is what we found during our investigation of the issue.

First, a little background…

Email marketing providers use an invisible image, or tracking pixel, in the email they send on behalf of customers. When someone opens the email, the image loads and is recorded at Constant Contact to let us know the email was opened.

Even if the tracking pixel doesn’t display, we can still detect that the email was opened if any link is clicked. The link’s URL has to begin with “http://” rather than with the “https://” secure form.

Normally, we receive information every time the same subscriber opens the same email, which we report in our customer’s All Opens report.

Here’s how the Gmail image caching changes things

With this Gmail image caching change, the first time the email is opened by the email subscriber, the images in the email (including the tracking pixel) are displayed in the email. The tracking pixel is requested from the servers of the email marketing provider and this event is recorded at the email marketing provider which can be reported to the customer.

In subsequent openings of the same email by the same subscriber, Gmail displays the cached tracking pixel. This implies that for the subsequent openings, the tracking pixel is not requested from the servers of the email marketing provider. Therefore the email marketing provider does not have a way to gather any further tracking information, until the caching expires. The cache has a Time-To-Live (TTL).

Like any other email marketing provider, Constant Contact is impacted by this change.

Gmail image caching impacts our customers in the following ways

Our customers track Unique Opens and All Opens.

The Unique Opens metric is NOT impacted by the Gmail caching change. If anything, since Gmail is also now auto displaying all email images, the accuracy of unique opens may increase. (This is something we will cover in a separate topic after some testing and research.)

The All Opens metric IS impacted because all opens takes into account unique opens and all subsequent opens. With the tracking pixel cached, all subsequent opens are not reported back to Constant Contact servers until the cache expires (after TTL).

Here’s an example:

Email ‘Email Campaign X’ Sent at 12:50pm on 12/01/2013 to 5 recipients.

Recipient 1 – Opens on 12/01/2013 (at 12:55pm, 1:00pm, 5:00pm), and 12/04/2013 (2:00pm)

Recipient 1 – Opens on 12/01/2013 (at 1:00pm, 2:00pm), and 12/02/2013 (11:00am)

Recipient 3 – Never opened the email

Recipient 4 – Never opened the email

Recipient 5 – Opens on 12/01/2013 (at 12:58pm)

Assuming TTL is 48 hours. With this assumption, for Recipient 1, the Cache expires at 12:55pm on 12/03/2013.

Reporting for Email Campaign X

Time (Report is viewed) Unique Opens before caching change
(Count)
Unique Opens after caching change (Count) All Opens before caching change (Count) All Opens after caching change (Count)
12/01/2013 at
12:57 pm
1 1 1 1
12/01/2013 at
12:59 pm
2 2 2 2
12/01/2013 at
1:05 pm
3 3 4 3
12/01/2013 at
2:15 pm
3 3 5 3
12/01/2013 at
6:00 pm
3 3 6 3
12/04/2013 at
3:00 pm
3 3 7 4

What was our approach to testing the impact of the Gmail image cache?

First, we wanted to identify the impact on unique opens and all opens. We also wanted to be comprehensive in covering all four major email providers and all major email client viewing tools (browsers, mobile devices, etc.). Lastly, we wanted to figure out the TTL (~ expiration) for Gmail’s cache.

For comprehensive testing of the impact, we worked through the following scenarios:

  • Email providers (Gmail, AOL, Yahoo, Hotmail)
  • Email viewing tools (Browsers – IE (IE 9.X, 10.x), Firefox, Safari, Chrome; Desktop Clients – Apple Mail, Outlook; Mobile Native Apps – iOS/iPhone, iOS/iPad, Android Phone/Tablet; Mobile Browsers – Safari, Opera, IE, Android)
  • Opens at the following intervals – Right away (0-5 min), 1 hour, 8 hour, 24 hours, 72 hours
  • Opens – unique opens and all opens

Thus we tested 14 x 4 x 5 = 280 Email checks and 280 x 2 = 560 Metric checks (opens – unique opens and all opens). Phew!!!

Why did we do it this way?

We knew that Gmail made the image caching change but we also wanted to make sure that other major email services had not made the same change. There are also so many tools to check emails these days. We wanted to see which email viewing client tools were affected by the change — desktop apps, web, mobile web and mobile native apps. We also wanted to know what the TTL for the cache was. This approach allowed our tests to be comprehensive and cover majority of the use cases.

These are the steps we took:

  1. Send out a campaign for each browser/email provider combination. For example, Internet Explorer (IE 10.x, IE9.x), and Gmail, AOL, Yahoo, Hotmail.
  2. After the email campaign has been sent, track the opens reported in Constant Contact to make sure they reported properly.
    • track the opens in the Unique Opens report
    • track the opens in the All Opens report
  3. For tracking the opens, as the contact, open the email:
    • 0-5 min after the email has been sent (the count should be reported as the Unique Opens report and also in All Opens report). Mark in the table if this open was reported in the Unique Opens report and All Opens report
    • One hour after the email has been sent (the count should be reported in the All Opens report because this is the second time the email has been opened; the count in the Unique Opens report should stay the same). Please mark in the table, if this open was reported in the Unique Opens report and All Opens report
    • Eight hours after the email has been sent (the count should be reported in the All Opens Report because this is the third time the email has been opened; the count in the Unique Opens report should stay the same). Please mark in the table, if this open was reported in the Unique Opens report and All Opens report
    • Twenty-four hours after the email has been sent (the count should be reported in the All Opens report because this is the fourth time the email has been opened; the count in the Unique Opens report should stay the same). Please mark in the table, if this open was reported in the Unique Opens report and All Opens report
    • Seventy-two hours after the email has been sent (the count should be reported in the All Opens report because this is the fifth time the email has been opened; the count in the Unique Opens report should stay the same). Please mark in the table, if this open was reported in the Unique Opens report and All Opens report
    • For item #3 above, if your email client asks you to download images make sure you download images.

What did we find?

Here are the impressions of the findings:

  • Unique opens in all contexts tested (52 combinations) is consistently reported as expected.
  • In all non-Gmail email providers tested {AOL, Yahoo, Hotmail} each subsequent open is consistently reported in the All Opens report.
  • Gmail consistently caches images for between 30 and 48 hours in all desktop and mobile web browsers and Android native applications; however, desktop applications tested {Windows MS Outlook, Apple OS Apple Mail} and mobile native iOS Apple Mail do not.
  • Images downloaded by the same subscriber during the cache time will not count towards the all opens count. An image downloaded after the initial cache time expires will register another open for the same user in the all opens count.

TTL: 30 hours to 48 hours

Unique Opens: Unaffected

All Opens: Affected only for Gmail email provider, browsers (web and mobile) and Android Native Email Apps

While we tested quite a few combinations of devices, email domains and timings, we can’t test all of them. However, we have covered all common and important ways people read emails, giving us highly accurate results.

For Constant Contact customers, unique opens will consistently be recorded as always, and all opens for Gmail recipients will not increase during the first 24 hour period following the initial open when in email browsers (desktop or mobile) or Android native email applications.

Other email marketing providers are also seeing similar impacts on their reporting.

This is simply a change all marketers are experiencing

The good news is that Gmail has announced they are using this image caching technology to protect their users so they can allow images to always be on by default. We are currently validating this with our research and testing and will update here soon but we strongly believe this change will actually give you more accurate open rates. More on this to come!

Comments

  1. Raj says:

    Great Share… keep posting such informative articles…

  2. NBB says:

    I’m curious what results Constant Contact has found from reliability of images opening. I am getting feedback from Gmail users that the images in our emails now don’t load, or only half of them load. We are not doing click tracking on these images — they are simply embedded and hosted on our own website. If you right click on a broken embedded image and do “Open in New Tab”, you’ll get a “500 Server Error” for googleusercontent.com.

    This is driving me nuts! It is not a Constant Contact problem, but I really don’t know what to do. I’ve been active in several related Google Product forum threads, but it feels pointless. What a total f$#kup on Google’s part.

  3. stacey crawford says:

    Nice discussion , I learned a lot from the information . Does anyone know if my company could obtain a sample CA I&A mileage Form document to type on ?

Leave a Comment