Jordie van Rijn wrote an article very inspiring to me, in which he aggregated different aspects and trends of what shaped email marketing automation in 2020 and what might have an impact in 2021.
I would like to add 3 more ideas which might impact email marketing in 2021.
- Phishing is black hat email marketing, so make sure everybody knows on which side you stand. You can do this by implementing fundamental standards like DKIM, DMARC, SPF and BIMI.
To protect users, emails from senders without basic trust factors will be deleted more and more without further notice even before reaching the users inbox.
- Listen, if you want to have a real conversation and a great CX. "Noreply"-addresses make you look like someone who only wants to talk about himself. Value response management as continuation of the conversation which you initially sparked with you email. No AI required but will help to manage high volumes.
- Watch out for AI’s using GPT-3 to generate email content, harder and harder to differentiate from something a human would have written if combined with a GAN
Deliverability is the term used to describe the probability that an email will reach the inbox of the recipient.
A lot of email marketers might think, that this is always the case. But it is not.
Indeed sending and recieving an email is a multi-step process. From the email marketing system sending the initial email, to the different email service providers and mail transfer agents relaying the message, until it finally reaches it’s destination inbox. And in all theses steps in between the email could be deleted as false positive SPAM or phishing message, without further notice.
Such a reject will never reported back to the sender (nope, bounce rates are something different) so that real SPAMers don’t get warnings that they have been busted.
As it can’t be measured, it seldom get’s managed, as Mr. Drucker said and so it’s still something, a lot of email marketers are neglecting.
Just like with SEO in the early years, optimizing the email deliverability rate can’t be measured by a simple KPI. But the impact can be huge, as only if an email reaches an inbox, it can be clicked.
More and more emails will be rejected, as phishing is still the most common way of infection in cybersecurity and email service providers want to protect their customers.
According to knowbe4
More than 90% of successful hacks and data breaches start with phishing scams. Phishing is a threat to every organization across the globe.
But what is phishing?
Phishing emails disguise as valid emails and try to convince the recipient to click on malicious links. The purpose is often to steal credentials or banking information, as well as installing malware to gain further access to the victims network.
Real-world phishing emails offer something worth clicking – or create a sense of urgency. The most dangerous do both.
And that’s exactly the same mechanics every white hat email marketer uses as well, when trying to create emails with high conversion rates. The difference is for sure the intent and the trustworthiness.
The decision, if an email gets delivered or deleted is made by trust.
If for example Gmail trusts you as a sender of good emails that most recipients want to have and read, measured by open rates and low amount of complaints, your email will reach the inboxes most of the times.
But if someone is new, has never send such a high amount of emails and may even get a lot of complaints, the sending email address and sending IP-address is burned.
So just like in real life, trust raises slowly over time by consistent and reluctant behaviour.
But did you ever received an email from someone you trust, just to realize that the senders address was spoofed and the email is a phishing email? Sometimes we get it to late, no matter how much we know about it, just like the police chief from Cologne recently. Nobody is safe to never get fooled by a well crafted phishing email, triggering the right topics at the right time.
IT-people usually answer such problems with IT-solutions.
That is why standards like DKIM, DMARC, SPF and most recently BIMI have been invented. If they are configured correctly, it is much harder for the bad guys to abuse trusted sender names and therefor the trust in real marketing emails will increase.
With tightened budgets due to COVID-19, in 2021 email marketers might be as well incentivised even more by conversion rates then sending volume. So having a closer look on domain alignment and deliverability factors will become relevant as well.
Communication is a 2-way process. Or should be at least.
In my perception, only very few organisations sending high volumes of marketing emails, really listen to their customers. Most of them are yelling their messages into the inboxes.
Complaints get managed by call centers, canned responses get prepared and the single complaining customer becomes a insignificant statistical number in a large blast of emails.
Jordie already wrote about Customer-centricity and Personalization. I’m more thinking of hyper-personalization when the email marketing not only measures behaviour and concludes to interests and dislikes, but really starts listening to the single recipient.
Utopia? Not really, as this is the same way chat-bots are working. The try to process natural language (NLP) to understand what the human counterpart is trying to say. They listen. And they try to give an individualized answer.
They don’t model an answer for a persona and give a generic reply to see if it matches the assumed interest. They listen first and answer then.
But what if you try to answer a newsletter from one of the big players in email marketing?
Lucky you, if it is not one unholy
No-reply can be translated into "listen to me but don’t bother me with your answer".
When some sender sends out blasts of millions of emails at once, it will be almost impossible to answer every reply, even if only 1% will request so. I get it.
This is exactly the scenario, where we need the AI-magic Jordie describes under the headline "The augmented marketer Human + Machine" in his article.
AI based content creation
But AI soon might help email marketers even more.
Today if a email marking system claims to have AI to predict e.g. subject lines, this is always based on large sets of emails that have been created and send before by humans. So there is sometimes a huge delay between buying such a system and benefiting from the first AI-created subject lines.
This gives us the feeling that the bots will never soon be able to write whole emails for us. And even then, the result has to be inspected by a human to avoid embarrassing nonsense in the inboxes of our recipients.
But there are 2 technologies that might change the game, as soon as they get combined in a product: GPT-3 and GAN’s.
GPT-3 by OpenAI is a language model that uses deep learning to produce human-like text. It has impressive capabilities, due to its immense amount of data available: 175 billion machine learning parameters. No other model has so much data available.
What it can do already?
There is a whole list of GPT-3 experiments curated on GitHub but in the context of this blog post, writing a whole article for the newspaper "the Guardian" really stands out.
Reading this article and having a closer look to the simple experiments conducted only a few month after the public beta of GPT-3, you don’t have to have much fantasy to imagine letting it write whole emails.
As the content creation process is still one of the most time consuming parts in email marketing, this is already a huge lever. If you combine it with the idea of hyper-personalization, it might be the spark needed to make the engines run.
Please take also into consideration, that it is almost always the text, the words in an email that make us click. The design and the images set the mood, the text transfers the message.
So even if an AI can pre-write an email, we would need a human editor to redact the texts. Don’t we?
What if a machine could learn what a good text is? What your preferred style of writing is? What your company sounds like?
Science fiction? Not at all.
Already 2014 Ian Goodfellow wrote a paper about "Generative Adversarial Networks" which – highly simplified – suggest exactly this: machines that train machines to create convincing results.
My favorite implementation is a website that produces images of people who don’t exist or cats that look most of the time very real to me.
Yes, this technology might not be perfect, as a lot of horses on the interweb might agree on, but imagine combining the capabilities of GPT-3 with the idea of GAN’s.
I would think, that this will reduce the time to write convincing emails drastically. Maybe even to a point, where no human interaction is required anymore.
And now listening to every single reply to your marketing emails and replying fully automatically with human-sounding, dedicated answers until the customer is either happy or the problem has to be managed by a real human, becomes something I expect to be available in the near future.
The technology exists already today and with more and more use cases and touch points with AI we as humans will also gain trust to use it more and more unattended in email marketing automation.
What do you think? Will this become reality soon or do you see other ideas more relevant? Let me know.
Photos by possessed photography, jesse ramirez, markus winkler, yi liu