Inaccurate open tracking

Jason Thompson

Jason Thompson

Member
I've come to a conclusion that open tracking mostly works but not very accurately which is causing me some issues when remove inactive subscribers.

For example; i send a test email to my seed list and don't open the test email, yet when i view the stats in mailwizz it shows multiple opens, and when i search the IP that opens its a google proxy. So in this case when i segment my list into openers they are not actually real openers since majority of them are gmail IPs.

Is there some fix for this?
 
twisted1919 said:
Most likely no. We pushed a fix at some point which proven to be a mistake, so right now we'll take our time for it.
Click to expand...
Hi twisted

Can we please get an update for this ASAP?, without this update my list cleaning efforts (building an openers only list) won't make sense. This is extremely Important thing for me and many others.

I'm still using other paid autoresponder services along side Mailwizz and one the most important thing is the ability to clean list and keep only true openers because it really helps with overall long term deliverability.

I honestly think this should be our #1 priority in terms of updates!

Thanks
 
bz87 said:
I honestly think this should be our #1 priority in terms of updates!
Click to expand...
Here's the deal, opens from GMail's web interface are made only via the image proxy, so the request is not made by the user itself but by the proxy. If we don't track this behavior, then the open is not tracked at all. If we track it, it comes from the proxy not from the user. Do you see the problem?
 
twisted1919 said:
Here's the deal, opens from GMail's web interface are made only via the image proxy, so the request is not made by the user itself but by the proxy. If we don't track this behavior, then the open is not tracked at all. If we track it, it comes from the proxy not from the user. Do you see the problem?
Click to expand...
Yes, but that's not the only alternative and now what other ESPs are doing to overcome this. Mailchimp, Salesforce, Adobe Campaign, etc. have accurate open tracking despite Google's image proxy scenario. So there is a way other than simply blocking those IP addresses. It might require some research.
 
nadworks said:
Mailchimp, Salesforce, Adobe Campaign, etc. have accurate open tracking despite Google's image proxy scenario
Click to expand...
They have the exact problem, the difference is that the end customer doesn't get to see the stats as you can see them in MailWizz.
 
@Jason Thompson - Thanks for the link, we'll revisit this issue as soon as possible, as you said, this is indeed a sensitive topic and we don't want to get it wrong. If you ask me, it's better to get all the opens instead of no opens because of a mistake.
 
Hi @twisted1919 , do you have any ETA on this fix? It seems to be coming a big problem for me mostly causing deliverability issues. There must be some kind of fix out there that autoresponders are using. I think this topic is super important and needs addressing as i see above its not just me facing this issue.
 
Last edited:
Hi @twisted1919 , i've spent some time testing a lot with this google proxy image. Here is what i've came up with.

After sending test emails to my seed emails i noticed if the email lands in spam or promotion tab the google proxy does not open the email.

If the email lands in the inbox it immediately opens the email regardless of me opening the email.

So it goes like this if the email is not opened by the user.

Inbox: 1 Open (usually immediately)
Promo: Not open
Spam: Not open

If i open the email it immediately shows 2 opens from all folders. So im assuming 1 open is the original google image proxy scanning the email, and the second one is the real user agent, however i think the google image proxy masks the real user agent info so it shows 2 opens from google image proxy.

So the way my data shows if there is only 1 open from a google proxy image it means its not a real open. Its just gmail scanning for any malware or something like that.

So i'm wondering if its possible we can get the real user agent info thats masked behind the google image proxy agent, would that be possible?

Also maybe you could setup a condition like this:

(If google image proxy opens just 1 time then exclude that open from the stats)

Or

(If the google image proxy opens more than 1 time then include it as 1 unique open)
 
Jason Thompson said:
Hi @twisted1919 , i've spent some time testing a lot with this google proxy image. Here is what i've came up with.

After sending test emails to my seed emails i noticed if the email lands in spam or promotion tab the google proxy does not open the email.

If the email lands in the inbox it immediately opens the email regardless of me opening the email.

So it goes like this if the email is not opened by the user.

Inbox: 1 Open (usually immediately)
Promo: Not open
Spam: Not open

If i open the email it immediately shows 2 opens from all folders. So im assuming 1 open is the original google image proxy scanning the email, and the second one is the real user agent, however i think the google image proxy masks the real user agent info so it shows 2 opens from google image proxy.

So the way my data shows if there is only 1 open from a google proxy image it means its not a real open. Its just gmail scanning for any malware or something like that.

So i'm wondering if its possible we can get the real user agent info thats masked behind the google image proxy agent, would that be possible?

Also maybe you could setup a condition like this:

(If google image proxy opens just 1 time then exclude that open from the stats)

Or

(If the google image proxy opens more than 1 time then include it as 1 unique open)
Click to expand...
@twisted1919 I just came here to say the same thing.

Could you please at least create an option on the "Filters" page to be able to segregate/show openers who opened more than 1x time?

Thanks
 
I'm no expert.

I've read that all email service providers like mailchimp, AC etc etc have problems registering real/true open rates.

Surely, better to measure engagement from concrete link clicks?
 
Jason Thompson said:
Inbox: 1 Open (usually immediately)
Promo: Not open
Spam: Not open
Click to expand...
Jason Thompson said:
So the way my data shows if there is only 1 open from a google proxy image it means its not a real open. Its just gmail scanning for any malware or something like that.
Click to expand...
This only works for inbox, but if you get the email in promotion and no open is tracked and then you open it for real, then you will skip the real open because you disregard those with a single open.

Jason Thompson said:
So i'm wondering if its possible we can get the real user agent info thats masked behind the google image proxy agent, would that be possible?
Click to expand...
We can't because the user agent is actually the google proxy user agent.

Jason Thompson said:
(If the google image proxy opens more than 1 time then include it as 1 unique open)
Click to expand...
Unique opens work exactly this way, no matter how many opens you get, it's going to show you just one.

wokka said:
I've read that all email service providers like mailchimp, AC etc etc have problems registering real/true open rates.
Click to expand...
That is also correct. As I stated numerous times, all the others don't show that many details and this is why most people don't notice this problem.

---

Here's how our tests look like:

1. We sent a campaign which landed in Promotions. We opened the email from the web interface and the email has been opened 4 times by gmail. In all cases, the IP addresses were Google's and none of them were ours. We then opened the email from our email clients on both mobile and desktop, and it recorded the right IP addresses.
2. We sent a campaign which landed in the Primary/Inbox. No open was triggered automatically. We opened the email from the web interface and there were 2 opens tracked, from Google's IP. We opened from our email clients, desktop/mobile and the right IP was tracked.
3. The number of opens done by google vary, for example, @laurentiu opens were recorded a different number of times than mines.

What we understand from above, as long as you open the email from your email client from desktop/mobile, then IP address registered is the one belonging to the person doing the open. So this is all good.
When you open the email from the web interface, no matter the tab from where you open it (we didn't test spam), then google will use the image proxy and trigger the open in behalf of the person doing the actual open, so the recorded IP address is Google's.

I don't really see a way around this, if the email is open from Google's web interface, there is no way you can get the actual user IP address, you will always get the one from Google, no matter what you do.
 
wokka said:
I'm no expert.

I've read that all email service providers like mailchimp, AC etc etc have problems registering real/true open rates.

Surely, better to measure engagement from concrete link clicks?
Click to expand...
I believe this is incorrect. Aweber, Getresponse, mailchimp and all other major email service providers are well established companies for many years, google image proxy only work on small companies or high risk mailers.
 
Jason Thompson said:
I believe this is incorrect. Aweber, Getresponse, mailchimp and all other major email service providers are well established companies for many years, google image proxy only work on small companies or high risk mailers.
Click to expand...
Nope. Nothing to do with that. The Gmail image proxy / caching is not selective. I used to work with one the UK's larges eCRM agency and the problem was well known across the industry.
However, I still think the leading ESPs have somehow sussed out a solution or at least a work-around. That's what I still believe is worth investigating for us.
In the meantime, MW should at least fix the device statistics and exclude the Google IP from those stats only, since that's the most glaring bs. When I present these numbers to my clients with 90% of their subscribers supposed to be opening on Mozilla on a Windows desktop, I'm getting laughed out of the village and lose credibility each time.
 
nadworks said:
MW should at least fix the device statistics and exclude the Google IP from those stats only, since that's the most glaring bs
Click to expand...
See my above case, where I legitimately opened the email and the open has been tracked just once having google's ip address. if we'd exclude that, we'd exclude a perfectly valid open, which is not acceptable.
 
@twisted1919
why not
implement Event Webhook because it will include the most amount of information on those engagement events.
Here is an example of how reports those open events to the webhook:

{
"email": "example@test.com",
"timestamp": 1513299569,
"smtp-id": "<14c5d75ce93.dfd.64b469@ismtpd-555>",
"event": "open",
"category": "cat facts",
"sg_event_id": "sg_event_id",
"sg_message_id": "sg_message_id",
"useragent": "Mozilla/4.0 (compatible; MSIE 6.1; Windows XP; .NET CLR 1.1.4322; .NET CLR 2.0.50727)",
"ip": "255.255.255.255"
}

This user agent and IP field should that of the recipient.

Could you explain why not above should not work to get accurate open data?
 
You must log in or register to reply here.
Back
Top