Jakub Olexa on NHI – Non-Human Interactions in email marketing

Jakub Olexa, the founder and CEO of Mailkit and Omnivery, is a leading expert on the topic of Non-Human Interactions (NHI) in email marketing.

In the ever-evolving email marketing landscape, staying informed about the latest trends and challenges is crucial for success. One topic that has recently gained attention yet remains largely misunderstood is that of non-human interactions (NHI). Often mistaken for bot activity, these interactions can have significant implications for email marketers. This blog post delves into what NHIs are, why they matter, and how marketers can adapt to this emerging reality.

Every email platform has a dashboard that indicates email campaign metrics, including open and click rates obtained in each email campaign.

Many email marketers measure their success (or failure) by referring to engagement metrics -especially email open and click rates.

However, open and click data have become increasingly distorted over the years and may fool marketers into thinking that their email platform dashboard’s “good” open and click rates indicate that their email marketing program is in good shape. The data (Non-human interactions) is inflated with false data that is far from reflecting reality.

Is it a bot (or not)?

We are witnessing a fascinating phenomenon changing the game rules regarding engagement data in email marketing: Non-Human

Interactions (NHI) that generate non-human opens and clicks. It’s not the recipients who engage with the emails; it’s “a bot”.

The NHI phenomenon dramatically affects how marketers need to measure and examine the success of their campaigns in terms of engagement metric data.

The phenomenon also interrupts marketing automation flows triggered by opens and clicks.

Non-Human Interactions (NHI)

In this episode, we are focusing on non-human interactions (NHI) in email marketing, a phenomenon that is reshaping how marketers understand opens and clicks. NHI are interactions that appear to come from the recipient but are performed by automated systems like security scans or privacy protection systems.

Apple’s Mail Privacy Protection (MPP), introduced in September 2021, is a major contributor to NHI. MPP loads emails in the background, fetching images and tracking pixels, resulting in inflated open rates. In Europe, around 60% of opens are coming from MPP.

The impact of NHI on e-commerce brands is significant. Not only are open rates bloated, but there are also non-human clicks, affecting automation, reputation, and list quality. Marketers cannot rely on opens and clicks for engagement signals, as they may be executing automation based on non-human interactions.

List hygiene and sunset policies based on engagement are also impacted, as marketers cannot clean their lists based on fake opens and clicks.

Engagement signals are crucial because mailbox providers use them to determine inbox placement, protecting users from unwanted emails.

Jakub Olexa, the founder and CEO of Mailkit and Omnivery, discussed solutions for ESPs and large senders to obtain more accurate open-and-click metrics. Omnivery has developed an API that detects non-human interactions, allowing ESPs to provide more accurate reporting to their clients.

The discussion highlighted the complexity of the issue and the need for marketers to adapt their strategies and tools to account for non-human interactions, ensuring they are sending relevant content to engaged recipients.

Link and sources

Mailkit: https://mailkit.com/

Omnivery: https://omnivery.com/

M3AAWG (Messaging, Malware, Mobile Anti-Abuse Working Group): https://www.m3aawg.org/

Questions and Answers

Q: What are non-human interactions (NHI) in email marketing?
A: Non-human interactions in email marketing refer to actions such as clicks or opens that appear to come from a human recipient but are performed by automated systems. These are often security scans or privacy protection systems, rather than malicious bots.

Q: Why is the term “non-human interactions” used instead of “bots”?
A: The term “non-human interactions” is used because these interactions are not necessarily negative or abusive, unlike the connotation often associated with “bots.” They are typically performed by systems for security or privacy purposes.

Q: Who coined the term “non-human interactions”?
A: The term was coined while developing a white paper at M3AAWG, a working group focused on messaging security and email abuse.

Q: How long have non-human interactions been a known phenomenon?
A: Non-human interactions have been around for about three years, and they have become increasingly visible. The topic was researched as far back as eight or nine years ago.

Q: What prompted the initial investigation into non-human interactions?
A: The investigation began when a small B2B customer noticed an unusually high open and click rate on their emails, which suggested non-human activity.

Q: What are some examples of systems causing non-human interactions?
A: Examples include security scans, corporate firewalls, antiviruses, and anti-spam systems that scan emails as they pass through corporate systems.

Q: Why is it important to understand non-human interactions in email marketing?
A: Understanding non-human interactions is crucial because they can affect email performance metrics, leading to misinterpretations of engagement levels and potentially impacting business decisions.

Q: What challenges are associated with detecting non-human interactions?
A: Detecting non-human interactions in real-time is challenging, especially in large datasets, because they can mimic human behavior but often show patterns that are not typical of human actions.

Q: What was the purpose of the white paper published by M3AAWG?
A: The white paper aimed to raise awareness of non-human interactions, highlighting them as both a technical and business problem in the email marketing industry.

Transcript

Title: Transcript – Tue, 15 Apr 2025 09:08:29 GMT
Date: Tue, 15 Apr 2025 09:08:29 GMT, Duration: [01:01:34.58]
[00:00:06.55] – Jakub Olexa
If you have an iPhone or iPad or plugged into your to your power overnight it below load every email that comes to your mailbox and fetch the images, including the tracking image. And as a result everyone who has an Apple device opens. That’s currently, you know, around 60% in Europe. In our stats, it’s 60% of all the opens are coming from Apple MPP. So everybody seems to be active. I mean they loved it for sure. It was a good day for a marketer who focuses on open rates, a bad day for those who measured the conversions, the conversion rates.

[00:00:51.22] – Sella Yoffe
Hello and welcome to episode number 12 of EmailGeeks.show, the podcast where almost once a month email geeks and lovers from around the world share their knowledge in email marketing, email deliverability and email marketing strategies. My name is Sella Yoffe and I’m an email deliverability consultant. I work with global email senders, startups, email and digital agencies, and with some email service providers on improving their email deliverability, email authentication and email strategy. And I’m the host of this podcast. Today, for the second time on the podcast, my guest is Jakub Olexa, the founder and the CEO of MailKit and Omnivery. I’m sure it won’t be his last appearance on the show. Today we are diving into non human interactions in email marketing, a phenomenon reshaping how marketeers understand opens and clicks. How are you today, Jakub?

[00:01:59.93] – Jakub Olexa
I’m good, how are you?

[00:02:01.31] – Sella Yoffe
Thank you for joining me for such an interesting topic.

[00:02:05.68] – Jakub Olexa
Yeah, yeah, it’s one of my favorite topics in the past couple of years.

[00:02:12.34] – Sella Yoffe
Many email marketeers are unaware of the topic of non human interactions. Nhi, how would you explain non human interactions in email marketing to email marketeers who haven’t heard about it?

[00:02:27.34] – Jakub Olexa
This term was coined when we started working on a white paper at M3AAWG, which is a working group for messaging where we deal with security and other aspects of email and abuse, etc. And you know, back then when we started, some of us started to notice something that you would call bot clicks or bot opens, but we didn’t necessarily want to call it bots because these are not bots per se. These are not abusive interactions or something bad. And the term bot is usually associated with something, something negative. These are non human interactions that are performed by automated systems for a good purpose. I would say in most cases, of course, they’re non human interactions that are bots. They’re negative. But in most cases these, what is often referred to as bot clicks and bot opens are actually non human interactions performed by security scams or privacy protection systems, etc. So to phrase it down into a simple sentence, those are interactions that appear to come from the recipient, from a human, but they are not. It wasn’t that intended recipient who did that click or that open, but some, some sort of a system.

[00:04:16.04] – Sella Yoffe
M3AAWG was talking about non human interactions way before Apple MPP was introduced.

[00:04:23.08] – Jakub Olexa
Yeah, it was a long time ago and it took about a year, maybe even more to get it published. We started way back because this is not a new phenomenon. Right. The non human interactions have been around for years. It’s just that their impact was getting more noticeable over time. And I started looking into non human interactions about eight or nine years ago when we noticed our customer, a tiny customer, a B2B customer who sent like 400 emails and all of them opened and all of them were clicked and multiple times. Right. And that is something, when you’re looking at massive lists with hundreds of thousands of recipients, that is not something that you would notice. If you have 400 out of those hundred thousand opening and clicking. Yeah, you’re not going to notice it, but you have, if you have a list that has a hundred percent open rate and 100% click rate, you start digging into that, why is it? And then you notice, oh, it’s not just that there was a click for every recipient, there was a click on every link for every recipient. That doesn’t look human to me. Right, right. And that’s the thing about non human interactions. They’re very, they are fairly easy to spot in B2B environment. And especially if you’re looking back at your data, you’ll be like, oh, this is obviously not a human. It’s kind of more tricky if you want to do it in real time and detect those non human interactions and flag them. But we started this research and we found out, okay, so these are not bad actors. These are security scams, these are corporate firewalls, these are antiviruses and devices, these are anti spam systems, corporations scanning every single email that passes through the system. So that got my attention. And then later on I worked on the white paper at M3AAWG with my colleagues from a couple of other large ESPs to get this published, to bring some attention to this, because it’s not just a technical problem, it’s also a business problem.

[00:07:26.77] – Sella Yoffe
I think that email marketeers became aware of non human interactions when Apple MPP was introduced. Is MPP considered a non human interaction?

[00:07:39.30] – Jakub Olexa
It is. MPP is definitely a non human interaction. The thing is, as I said, if you’re in B2B and you have a small list and everybody opens, everybody clicks or you have 90%. It’s unusual, it’s something that you know, notice. If you have a massive list and there’s a 5% impact on opens and clicks, it’s not something you would outright notice. It was when Apple rolled out their MPP when suddenly those non human opens or bot opens, maybe 50 or 60% of all the opens, that’s when people started noticing. And on top of that I think there was a big impact of COVID which started roughly four months later. And many marketeers were not aware of the impact of Apple MVP and attributed the, the massive increase in opens to people being home. You know, people are sitting home, they’re reading my emails, they love my emails. Our, our email campaigns went through the roof, thank you Covid. But the flag was it didn’t go through the roof. And you know, many marketers started, started to notice later on during the COVID period that yes, the open rates went up, but the conversions, the revenues did not copy that at all. Yes, there was increase in sales during COVID but it wasn’t 100% increase like it was with opens. A lot of this was kind of hidden through other impacts on, on our lives.

[00:09:57.14] – Sella Yoffe
Apple MPP was introduced in September 21. Can you explain what Apple basically doing with that mail? Privacy protection?

[00:10:06.67] – Jakub Olexa
Yeah, so the, the idea of Apple MPP is that Apple will load the emails in the background when the device is not being used, preload everything, and by doing that kind of hide the IP address of the recipient and protect the privacy better. I would argue it just doesn’t work as, as intended because it also gives a very strong signal that the address actually exists and that the person behind that mailbox actually reads that mailbox and has an Apple device that loads those messages. So it’s something that bad actors can really abuse by testing to see is this person active or not. But that’s beside the point. The point is it loads the emails, it loads everything without the user actually doing anything. So if you have an iPhone or iPad or plugged into your power overnight, it will load every email that comes to your mailbox and fetch the images, including the tracking image. And as a result everyone who has an Apple device opens. Right. So that’s currently, you know, around 60% in, in Europe in our stats, it’s 60% of all the opens are coming from Apple MPP. So everybody seems to be active. Right. And it also meant that, you know, those who were seeing 25% open rates, you know, Covid started and Apple MPP kicked in and suddenly it was 50% open rate.

[00:12:07.49] – Sella Yoffe
Right.

[00:12:08.63] – Jakub Olexa
Which, I mean, they loved it for sure. It was a good day for, for a marketer who, who focuses on open rates, a bad day for those who measured the conversions, the conversion rates, roughly.

[00:12:25.01] – Sella Yoffe
Can you, you know, give us a sense of the impact? And I know it varies based on country and in my country or in the state it will be different.

[00:12:37.62] – Jakub Olexa
Yeah. Well, on average In Europe it’s 50%. Half of all the opens are coming from Apple MPP. In US that number will be higher because Apple devices are just more, more popular there, but could be 75%. So it’s, it’s a lot.

[00:13:06.42] – Sella Yoffe
Amazing. That brings me to the next question. Many ESPs have not updated their platform statistics or informed their clients about this change. Everything on their platform dashboard remains essentially the same.

[00:13:26.54] – Jakub Olexa
Well, everything is the same, but also everybody is very happy about high open rates. Right. So for many ESPs it became kind of a internal fight between the technical teams and the business teams because it is kind of hard to make the business case for. Why should we suppress the non human interactions? Right. Because it may scare off the customers when they see a sudden drop. At mailkit, we’ve been doing detection of non human interactions before Apple MPP kicked in. So when it kicked in it had no impact. Right. The customers didn’t see that spike in open rates, so we didn’t have to worry about them being unhappy about the drop. Now other ESPs who haven’t done that are in a tough position. And I mean it’s not that tough because there are ways how to present this to the customer. And we are also showing these are the interactions that are human and this would be the total number if non human interactions would be included. So you kind of see, okay, my open rate is 25%, but including non human interactions it would be 40%. Right. So there are ways, but it’s really about the business case for many ESPs. And when it comes to opens and when it comes to Apple mpp, it’s well documented by Apple how to identify Apple MPP open. So it’s not that difficult to implement. When it comes to clicks, that becomes really hard because you don’t have Apple MPP behind those and it’s not easy to undo, identify.

[00:15:35.01] – Sella Yoffe
Correct me if I’m wrong, but maybe it’s a question of a lack of standards. Let’s take two different ESPs. One is placing the tracking image pixel above the fold and the other at the bottom of the email, does it also impact the open rates statistics?

[00:15:54.08] – Jakub Olexa
Okay, so this is tricky. A the tracking pixel was never accurate, the open rate measurement was never accurate, and it was always a rough number indicative. Second, if you have the tracking pixel at the top of your HTML of your mail, if images are loaded on Gmail, that tracking pixel will load. Right? On the other hand, if your email is too long so Gmail decides to crop it, you won’t know about it because the tracking pixel loaded. If you have the tracking pixel at the bottom, you have the same likelihood of getting the information about the email being opened. But in addition, you will also get the indication if that email was cropped by Gmail. Because suddenly your open rates at Gmail will be super, super low. Right. So I don’t see an advantage in putting the pixel up top. I see the advantage of putting it at the bottom. Not to mention the reduced risks of altering the HTML and having something screw up the content.

[00:17:29.04] – Sella Yoffe
I wonder why there is no industry standard for placing the tracking pixel in an esp.

[00:17:36.65] – Jakub Olexa
Because this is not something that can be easily standardized. What would be the general benefit of standardizing this? Like you don’t have a standard that would generally cover how web ads, web remarketing ads work, Right? Okay. You can tie them. You know, every campaign can be tied to different attributes, different behavior, etc. So what would be the benefit of having this standard even within M3AAWG, when we already know that it will be inaccurate anyway? Right?

[00:18:18.13] – Sella Yoffe
Yeah.

[00:18:18.69] – Jakub Olexa
So it’s like standardizing something that is broken. And it is broken. It has always been broken. It was not intended this way. Right. If the creators of email wanted to have a technology based information relayed about the status of the message being opened, they would incorporate that into the email standard itself. But because of how email was built and what was the purpose originally, nobody expected this to be used for marketing. Right. So there was no need for tracking the opens. And even today, when the industry is discussing having a standardized engagement, reporting from from the ISPs to ESPs that would allow us to see the engagement. Because right now there are very few ISPs providing this information. You have Mailru, which has certain dashboards that allow you to see how the message was placed and what were the opens and skim opens and deletes, etc. You have CESNUM cz which has similar data available over API. But is it comparable? We don’t know. You have Yahoo, who has commercial product, and then you have Google Post Master Tools which are presenting certain Information, Right. Completely different information. It’s not necessarily, oh, this many people have opened, this many people have clicked, et cetera. And even if there is a standard like that which is currently being worked on, a standardized engagement feedback loop, it’s going to be an aggregate, it’s going to be a total number. So if you send out 100,000 emails and you see, well, 20,000 people opened it, how does that help you? Right, yeah, it’s not going to be helpful for your automations because you cannot just, you know, randomly guess 20% of your list. Right, right. These were the ones who, who clicked or these were the ones who open. You will get. And that’s the intent of the new standard, to give you more reliable engagement metrics. So you know if, if your recipients are really engaged or you, you’re just believing a random number from, from one pixel, one square pixel at the bottom of an email.

[00:21:23.23] – Sella Yoffe
And this standard will rely on the mailbox providers.

[00:21:28.93] – Jakub Olexa
Yes, that is something that the mailbox providers are working on to provide unified data. So what you get from Google, what you get from Yahoo, or what you get from Intel Nines or Apple icloud that it’s comparable. So you can tell, okay, I had an engagement of X here, X here and X here. It’s comparable numbers, but interesting. We are far from having that. The tracking image, whether it’s top, bottom, multiple is the best we can do to make sure that it has as wide compatibility as possible. Yes, with AMP you could do other things, but again, the support for AMP is fairly limited. You could incorporate schemas or many other technologies that would give you some sort of information. But would it be accurate enough? Would it work for 90% of your recipients? Like 20 years ago when we started we were talking about, hey, by the way, the accuracy of the open rates is low because people have to click to load images. Right. Remember that 20 years ago it was people had to like to load images and 20% of people were using email clients that didn’t support HTML at all. So it was a completely different world. And open rates were extremely inaccurate back then. Right. Marketers got used to the fact that it became more accurate thanks to Google image caching. Right. So that made it more accurate because images were loaded by default for everyone. They didn’t have to confirm, yes, I want to load images. Right. Like everybody adopted that. So it got more accurate way more than it used to be. It doesn’t mean it became accurate enough.

[00:23:53.69] – Sella Yoffe
Right. I think that you mentioned the Gmail. Gmail is also doing some kind of a pre fetching of images, proxying it.

[00:24:04.75] – Jakub Olexa
Yeah, so that’s another thing. Google, Yahoo, Yandex, you name it, pretty much everyone these days does image caching, right? So instead of downloading the image every single time, they will just cache the images and store it in their system. So when you open an email that I opened an hour ago, they don’t have to go and download that image again and again and again for every single person because that’s a lot of bandwidth. Yeah, now that caching is very efficient. But at the same time if they cache an image that is unique to a single recipient, it does not affect the open rates really it would affect the metrics, how many times that single email was opened. But because the tracking image has a unique URL for every recipient, it’s not affected by the caching in terms of measuring open rates or in 99% of cases, or 95% of cases. Because there are cases when this loading is not done when the user opens an email, but when Google wants to see that email, do some security scans.

[00:25:41.63] – Sella Yoffe
What’s the impact of non human interactions on e commerce brands?

[00:25:46.49] – Jakub Olexa
It’s massive. Not only because the open rates are bloated, bloated you know, two times. But also because the non human interactions there are clicks. Right. And today when e commerce is all about automations and you know, do this if that it not only affects the numbers that you see in your reports but it affects your cost because suddenly your executing automations based on a bot, if we call it that, based on an automatic system somewhere doing a scan of your email. Now not only that, it can and it will have a impact on your reputation because ultimately you’re sending emails, signals based on signals that are not true. And if you ask whether, whether these scans also scan double opt in emails, of course they do. And if your double opt in only relies on that single click then it will subscribe your email. Right? So now you can have malicious scribes that are confirmed. With double opt in you will be sending to factually inactive users and you will send them a whole scenario of emails based on their clicks. So instead of sending them one click and realizing that that’s a dead user, not opening, not clicking, suddenly you have all these interactions. So you will send a follow up email and then the other follow up and then the follow up. So you’ll send a lot of emails. And lastly spam traps are scanning just as much as any other email. So you cannot simply say well this user is an active user, it’s because he’s opening, he’s clicking. It might as well be a spam trap. So if you’re unable to detect non human interactions, then the overall quality of your list is going down the longer you keep following these false false events.

[00:28:36.11] – Sella Yoffe
Amazing. I’m saying to my clients, don’t rely on opens, you might rely on clicks. And now you’re saying it’s accurate. Yeah, more accurate.

[00:28:48.16] – Jakub Olexa
Let’s say today on average it’s around somewhere between 7 and 10% of clicks are non human in B2C.

[00:29:00.97] – Sella Yoffe
B2C?

[00:29:02.15] – Jakub Olexa
Yeah.

[00:29:03.56] – Sella Yoffe
How is that?

[00:29:04.67] – Jakub Olexa
Here’s the thing. The business addresses like Google Workspace or Outlook365 do these scans, right? Right. No question about that. They will do the scans and Microsoft is more known for this simply because they’re kind of aggressive when they’re doing the scans and more easier to spot.

[00:29:37.89] – Sella Yoffe
Right.

[00:29:39.49] – Jakub Olexa
But in the business address world, you also have multiple of these scans at the time because many of the customers don’t rely on Outlook365 Security but have a proof point in front of it. So then you could easily have two scans on every link. That doesn’t mean that regular Outlook and regular Gmail don’t do these scans. They do. They do them just as much. It’s just they not every email they use more clustering. Outlook 365 is a business address. If you have a company that has 100 email addresses, that’s a very small sample to do a cluster, it’s more efficient rather than clustering those messages, it’s probably more efficient in most cases for them to just scan every message. It’s the same for a proof point, right? Proofpoint doesn’t have time to waste on clustering and figuring out whether the message sent a minute ago is similar to the one that is sent right now.

[00:31:00.39] – Sella Yoffe
By clustering you mean that they are sampling.

[00:31:04.15] – Jakub Olexa
So clustering is when you have a message that has certain attributes, they can cluster all the messages that are similar.

[00:31:14.53] – Sella Yoffe
From the same sender.

[00:31:15.82] – Jakub Olexa
So it comes from a similar sender. It doesn’t really have to be the same sender because a lot spammers often spammers use thousands of sending addresses and domains, etc. But in general the content is the same. So they can cluster them together and see, oh, this is all the same spam or this is all the same promotional. Good stuff.

[00:31:44.72] – Sella Yoffe
Right?

[00:31:45.70] – Jakub Olexa
That clustering is using hundreds, maybe even thousands of attributes of a message to cluster them together and then see, okay, this is my group and this new message that comes in belongs to that group. Let’s not scan all those million messages. Let’s pick samples randomly and use those. And that’s what they’re doing. So if you’re sending a newsletter to Gmail and you’re hitting million addresses, Google doesn’t have to scan all the million addresses or millions copies of that message. They can just look, okay, so this is a cluster. Let’s look at samples, let’s look at thousand, maybe 10,000 random samples. See where they’re pointing, what is it? What’s behind that? What’s the risk factor here? And then based on that, they can decide, okay, this will go to Spot, this will go to promotions, this will go to Inbox, this will go to spam for sella, Inbox for Jakub. They’re very smart at this.

[00:33:08.80] – Sella Yoffe
Right. I think that the sender guidelines, I’m not just talking about the new sender guidelines was to send emails to people. Who wants to get those emails? Non human interactions make it impossible, almost impossible to the common sender to obey those guidelines. What the common sender using the common email platform can do?

[00:33:38.57] – Jakub Olexa
Oh, that’s, it’s tricky for sure. But a, the guidelines are very general. So they’re also telling you, okay, make sure your frequency is not too high. Right. Make sure you’re sending content that your recipients want that is relevant. So that will be different if you’re, if you’re selling food delivery. Right?

[00:34:04.11] – Sella Yoffe
Right.

[00:34:04.57] – Jakub Olexa
And if you’re selling cars, I certainly, I’m, I’m certainly not buying cars every day. Right. And not every day, but I need food every day.

[00:34:14.09] – Sella Yoffe
Yeah, Right.

[00:34:14.69] – Jakub Olexa
So the frequency has to be different. Now does it mean that the car maker should send me or my car dealer should send me an email only once every two years? Probably not, because I would forget about him. So that that dealer or that brand needs to think about what’s the balance between annoying my customers and getting a value to them. Right. Which also dictates the content. Now if I’m sending every two weeks, I have a good content and I’m seeing a bunch of opens but no clicks or the clicks are going down. I can take those clicks as an indicator. Right.

[00:35:06.71] – Sella Yoffe
What about newsletters?

[00:35:08.78] – Jakub Olexa
But it’s still tricky. Yeah, it’s still going to be tricky because if that recipient has Apple MPP and my ESP is not suppressing mpps, then I will keep sending to that recipient and be annoying as hell. Right. So marketers need to consider, does my platform have tools for suppression of at least Apple mpp? If not, how can I suppress it myself? There are Ways how markers can in many platforms import a list of certain suppression rules. So if the click originated at Apple MPP gateway IP don’t count that open, it’s an Apple MPP open. There are many ways how marketers can do rudimentary suppressions. In my opinion, it’s not the marketer’s job. They’re paying the platform for the tools so they don’t have to do it themselves. So I would argue if your platform doesn’t give you that option, you should probably switch your platform. Find another one.

[00:36:39.11] – Sella Yoffe
Do you have a knowledge about the percentage at least of ESPs that actually did some filtering around MPP? I think it’s a fraction.

[00:36:51.55] – Jakub Olexa
It is a fraction. I know that this year a lot of ESPs started to worry about this. It became something that they started to work on, started to build some tools, some detections, et cetera. Whether they rolled it out to their customers or not is a different question. How did they roll it out? I know that some are still presenting the overall number, but in fact if they know that it’s a non human open, it’s coming from Apple mpp, they will not include it in the automation triggers, et cetera. So it’s really a question of asking your esp, how do you deal with this? What is your situation when it comes to non human interactions? How do you handle those?

[00:37:50.51] – Sella Yoffe
For those who not switching an ESP right now, but still wants to navigate those bloated opens and bloated metrics, what should they do?

[00:38:02.34] – Jakub Olexa
Really, really good marketers with very high engagement will get to 30 plus real opens. But that guidance is not the best because it’s not about the overall number. Right. The deliverability is again is not overall. You cannot say you have good deliverability or bad deliverability in general because that’s generalizing. The biggest value in my opinion of open rates is when you’re looking at open rates over time, perhaps destination. So Gmail for example, have 30% open rate at Gmail. You cannot expect that you will have 30% at Yahoo, it will be less because their caching and behavior of their user base is different. The same goes for GMX because those systems plays into different folders, users behave differently at these platforms, et cetera. So you cannot say well my deliverability is bad at Sesnome because my open rate is 15% and Gmail is 30%. Unless you understand that CESNO has a different logic of placement of messages so that 15% in general, half of what you see at Gmail is actually a good number, that is what you should be aiming for. And this goes for every single mailbox provider. It’s like Google, Yahoo and Outlook. Each will show you a slightly different number because of many reasons. But if you see a sudden change in those numbers over time, that’s the indication of certain deliverability issues. That’s what you should be focused on for anything else. Open rate, you know, is a random number. You shouldn’t really care, right? And in fact, you shouldn’t really care about click rates as well. Right? Because those are more accurate. Way more accurate. But again, unless your business model brings you money from clicks, you need to focus on conversions, on actual sales. So yes, markers are used to looking at open rates and looking at click rates, but those rates don’t mean that much. It’s the actual events that matter. And if it was a human, so if a human clicks and you can say it was seller who clicked the link in the email, you can use that information, you can trigger some automation, offer discount, do whatever, right? But if you’re just looking at the overall number, it’s just a number.

[00:41:31.23] – Sella Yoffe
Sometimes we open emails not based on the actual emails. We see the email, I’m clicking on my browser and I’m on their website. So this is also a good signal.

[00:41:44.36] – Jakub Olexa
Yeah, well, I would argue it’s not okay, simply because, so attribution in general is a nightmare. But if you cannot reliably link that action to an email, then you cannot just assume that it was because of that email. And I agree that often you get an email, you don’t open it, you delete it right away. But that brand is so high on your mind that you will just, you know, go to your bookmarks or enter it into your browser. But you cannot say that it was because of that email because if you saw a Facebook ad, it’s a weak signal. It, in my opinion, it’s a pointless signal. Because it could be whether it was an email or it was that you saw a Facebook Facebook ad or you know, a YouTube video that mentioned that brand, or you saw billboard on the street, or you’re watching TV and you’re seeing their, their commercial, right? Like it’s a coin toss. It could have been any of that. So assuming that my email recipient is active just because he’s active on my website, but he never clicks, never opens my emails, and I will still keep sending him emails because I think, well, he’s active on my website, right? That’s just waste of money on those emails. And I agree emails are cheap, but if you have millions and millions of emails, it starts to add up and you’re just hurting your reputation in the long run.

[00:43:34.76] – Sella Yoffe
How do non human interactions impact list hygiene and sunset policies based on engagement?

[00:43:43.36] – Jakub Olexa
You cannot do list hygiene based on fake opens and fake clicks. You cannot. That’s, that’s the whole thing. What is engagement?

[00:43:56.73] – Sella Yoffe
Right?

[00:43:57.32] – Jakub Olexa
Engagement is not some click or some open. It’s the actual human click or human open. So if you’re, if you’re cleaning your list based on people who did not click and did not open, let’s remove those. Many do that, many do that. It’s better than nothing. But it doesn’t work, right, that you’re already including all the fake engagement in this process. So the.

[00:44:39.76] – Sella Yoffe
Do you think that marketeers needs to ignore engagement altogether? Because if they’re not using platforms that can make sense of out of the engagement and it’s a sophisticated question because most marketeers no. Don’t know what’s behind their platform. You know, a.

[00:45:02.19] – Jakub Olexa
They should ask. But there is engagement that you can reliably measure. Right? But it’s difficult to do that and marketers are not used to doing that. And that is connect your e commerce data, your actual sales.

[00:45:23.38] – Sella Yoffe
Okay.

[00:45:24.09] – Jakub Olexa
Your actual visits and where they came from. Realize that you can say, okay, this came from an email and it had two minutes worth of interaction with my website. Because those bots that will click your email when they’re scanning the link, they will visit your website and in many cases they will show up as a visitor in, in your Analytics, Google Analytics, et cetera. Because they will load the scripts and analyze the page, but they will not follow links on that page unless they identify something dangerous.

[00:46:08.78] – Sella Yoffe
Okay.

[00:46:10.82] – Jakub Olexa
So marketers need to put more focus on connecting the dots, connecting their commercial data. These are my sales, these are the people that I’m selling to that actually made a purchase in past six months and I can attribute that in any way to a direct or indirect way from an email, then that’s relevant. That’s a relevant engagement. That’s the best engagement, right? And once you have these, then take the rest and try to find a way to suppress those who are non human by the basic means. Right? If you end up with half of your list, that’s it.

[00:47:06.63] – Sella Yoffe
Mailkit is the best in class email and automation platform and Omnivery is a premium transactional SMTP service in the cloud. Mailkit and Omnivery lead the market in terms of innovations and deliverability. You highlighted the internal dilemma that many ESPs face regarding Apple, MPP and NHI. Many chose to do nothing. But you have a olution that allows ESPs and large senders to obtain much more accurate open and click metrics. Using your solution, can you share more?

[00:47:46.09] – Jakub Olexa
So the problem of non human interactions is not just a technical problem or a business problem, but when you go deeper, you find out it’s a complex technical problem. Apple MPP is the simple problem. It’s well published, well documented by Apple. That’s a problem that affects your opens and your open metrics and if you have campaigns that are tied to Opens, then obviously it will affect your automations. But the bigger problem, and the one that affects B2B just as well as B2C and affects the business even more is the non human clicks, which is, you know, where you actually usually make your money. You know, I haven’t heard of anyone who made money by, you know, looking at an email, but clicking on an email that usually brings the money. And these non human clicks occur usually when security scanners scan the emails. Now in B2B, as I mentioned, that could be 100% easily. It’s usually a very high number, right? But in B2C, because of much more advanced techniques and the mailbox providers being able to cluster similar messages together so they can be much more efficient and instead of checking every single message from a campaign and every single link in every single message, they would do a cluster and randomly pick, look at this message, we’ll look at that message, scan the links in those messages to see where they’re pointing to whether it’s a phishing attack or something that would be nefarious. So that’s the point of those non human clicks. The problem is that these are very difficult to detect. They are usually much easier to detect in the B2B space because of obvious patterns in B2C which is where marketers are mostly focused on, it’s really difficult. Again, it’s fairly easy to be retroactive. Go back, look at your reports and you see as a human you see, oh, this is weird. You look at one single recipient and you see, oh, he opened that message and then he clicked five, five links and each click came from a different country, obviously not a human. Right? But doing this at scale at the time when this occurs is much more difficult than it seems to. And it’s usually not as easy as these highly visible cases where you see opens from different countries.

[00:51:12.55] – Sella Yoffe
Right?

[00:51:13.92] – Jakub Olexa
It’s usually much more obfuscated because these are security scans. They want to be able to perform these security scans, so they don’t want to be seen and most importantly, they want to make sure that attackers won’t be able to change their destination URL in real time based on whether it’s a scan by a anti spam system or it’s a click by a human. It’s really hard to detect. But because we’ve been doing this for eight years, we’ve got pretty decent at this. And we, you know, this year we decided to release the engine that we use for this detection as an API product. So third parties like ESPs that need to perform these scans and identify the information for their reporting can pass the data to our API and get the data back with the information whether it was human or not. That’s the whole point, being able to provide more accurate information in the reporting. And we have a couple of ESPs already on board and taking advantage of this bot detection API.

[00:52:57.76] – Sella Yoffe
Let’s say we agree on the business terms, you’re well known of your vetting process. Is it for everyone?

[00:53:07.44] – Jakub Olexa
It’s not for everyone. There are definitely parties that we would not allow access to this API because we believe that they could abuse the information. We also have technical limitations embedded in the API so it cannot be abused for real time redirection, etc. But yeah, good. ESPs are definitely welcome. Even it could be used by brands, large brands. It doesn’t have to be an esp, but we are very careful about who gets to use this API because it not only has a lot of our data, but it also leverages data from our partners who require certain assurances from our side.

[00:54:09.50] – Sella Yoffe
If an ESP cannot distinguish between non human interactions and human interactions, but can identify user behavior on the website, can that signal be used as an engagement signal?

[00:54:24.75] – Jakub Olexa
If you record an activity of a user that came from YouTube or AdWords or Facebook or whatever and made a purchase on your website, but you have no report of him clicking a link or opening an email, then you cannot simply say, oh, it’s an active customer. It is an active user for sure. But he has no interest in email, right? You’re not losing that customer because you stopped sending an email to him. He will come through another channel if he wants to get your emails. He can always resubscribe, but don’t be afraid to remove inactives. And it really comes down to being aggressive on your list. And you have the tools today that allow you to automate all of this, including if there is a click, call our API, ask if it was human and have that in your automation. Yes, you can have that. And that’s amazing. Make A better decision. But if you as a marketer just sit there watching your automation because you’ve done it, you build your automations and don’t adapt to the changes of the environment, you probably don’t deserve the salary. Right, because anyone can sit on a chair and just watch how the automation does the job. Yeah, but the environment, the email environment, the environment of web, online ads, social networks, everything is changing. You cannot expect it to work the same all the time.

[00:56:21.17] – Sella Yoffe
I’m curious why engagement signals are so important. Because email users can always mark emails as spam or, or unsubscribe with the hit of a button.

[00:56:33.65] – Jakub Olexa
Yeah, but you’re, you’re assuming that everybody is a nice marketer.

[00:56:41.73] – Sella Yoffe
I hear it.

[00:56:43.32] – Jakub Olexa
So spammers don’t honor unsubscribes. They don’t care.

[00:56:49.25] – Sella Yoffe
Okay.

[00:56:50.30] – Jakub Olexa
A lot of brands take their time and I mean they take a lot of time to unsubscribe. And there were brands where I, where I had to actually contact the, the, the esp, my friends at that company and tell them put me on a suppression list because they’re ignoring my unsubscribes.

[00:57:11.90] – Sella Yoffe
Amazing.

[00:57:13.17] – Jakub Olexa
And those were no small brands. But I’m, I’m not saying that this, this is what goes on everywhere. But there are big brands who are trying to, you know, find a way around things. There are managers of marketers who just don’t want to see those unsubscribes. And they’re actively pushing the marketers who would like to follow best practices to send again and again and take advantage of whatever open window there is after the unsubscribe. It’s one of the reasons why Google said unsubscribe needs to be honored within a certain time frame. And then there are the bad actors who don’t care. They will not even include a unsubscribe list. But even then, even if it was a good sender who honors the unsubscribes and everything is fine, what if that sender sends million messages every single day to their million recipients? Why should all those people unsubscribe, do the action rather than just ignore it?

[00:58:37.98] – Sella Yoffe
Okay.

[00:58:40.05] – Jakub Olexa
And let Google calculate the engagement as insufficient and say, well, nobody wants these emails. Let’s put them into spam. And that’s what Google is doing. Google is protecting you as a Google user from. The same goes for Yahoo. Yahoo doesn’t care about the sender, about the brand. They’re not there to cater to the brand. They’re there to cater to their customers, which are the owners of the mailboxes and they want their best for them. So marketers need to look at it from. From that perspective. Google is not doing it simply because they would want to save some change on hardware on spam filtering. Because from their perspective, it would be a change. Right. It would be peanut peanuts. Even if. If it. If it’s millions of dollars, it’s still peanuts.

[00:59:43.09] – Sella Yoffe
Right?

[00:59:43.88] – Jakub Olexa
Right. Their concern is Sella being happy about what he gets into his inbox and that he doesn’t have to click report spam or unsubscribe on every single message that he gets every day. Imagine that. You know you would be getting hundreds of emails per day that you would be unsubscribing. Yes. That would be the last time you get that email. But you would have to manually unsubscribe 100 messages per day.

[01:00:16.65] – Sella Yoffe
That’s what was before 20 years ago or something. Yes.

[01:00:20.75] – Jakub Olexa
And we were not happy about that. Nobody was happy about that.

[01:00:25.05] – Sella Yoffe
Yeah.

[01:00:26.40] – Jakub Olexa
So engage with is a critical signal for the reputational systems to calculate and tell whether that mail should go to inbox or spam.

[01:00:42.32] – Sella Yoffe
I think we covered it.

[01:00:43.65] – Jakub Olexa
I guess.

[01:00:44.11] – Sella Yoffe
So something that I forgot or do you want to add something?

[01:00:49.00] – Jakub Olexa
Oh, I always have something to add. But we would be here for another hour and a half and you know, I loved it.

[01:00:57.90] – Sella Yoffe
You said I I will be bored, but I told you no. I always love to learn from you and you’re always full of surprises.

[01:01:08.65] – Jakub Olexa
Yeah, well, always happy to chat.

[01:01:11.36] – Sella Yoffe
I will count it and hope it will be sooner than later.

[01:01:15.75] – Jakub Olexa
Yeah, sure.

[01:01:17.75] – Sella Yoffe
Amazing.

[01:01:19.36] – Jakub Olexa
Thank you for having me. It was a pleasure chatting with you.

[01:01:23.67] – Sella Yoffe
Thank you very much, Jakub.

Scroll to Top