The problem with the SaaS industry is - most of the founders are either great Developers or Marketers or Trained in the field.
And the challenge here is that with AI & Innovation - the industry algorithms are changing rapidly - especially with giants like Google.
You can see how Gmail spam filters are constantly getting better and better every month - but can you remember when was the last time your email marketing tools invented the tech to fight those changes? 😶
Everyone is fighting on design, features, or pricing - but the age of AI needs a lot of innovation from the core. And that's what we're doing at BlogBing Group with growth-focused SaaS products.
The problem with cold emails is - we're still using the same email schedulers, the same dumb warmup algorithms, data mechanisms, and all. Let's understand how we're trying to reinvent each of the parts at GetMails - starting with Warmup today.
Let's Time Travel
Let's travel back to 2005-10.
- Emails were still early & best channel for leads.
- Companies like Gmail were focusing more on features & comfort
- Spamming or cold emails were still at an early stage.
- AI and Machine Intelligence was still very early kid requiring costly server resources - so not useful in production with scale ( yet ).
With time, slowly Google started focusing on spam & cold emails - and started building algorithms to fight the spam. And if you're Google - how would you do that?
- List out some spam words marketers are using in sales emails - and restrict them.
- Restrict emails sent by users - who are sending promotional emails only.
- Find out IPs that are sending a lot of sales emails, and restrict them.
- And few more metrics
Simple & accurate right? You can easily do this to block marketers & spammers. It worked perfectly. ❤️
Now companies around cold emails started building systems around those mechanisms.
Can you remember the spam word list, which you should not use in emails? Or the email warmup, where you send some dummy emails daily to reduce the promotional email ratio.
Or using clean IPs or trusted providers for better inbox delivery. That's how things were working and the business was great.
And as always in the SaaS industry, companies started replicating this technique with training courses and all - to build a business on top of it.
Put a few more features, reduce prices, and provide some new integrations, automation, or over-optimizations to innovate.
Now fast forward to 2022..
- The use of AI is super affordable, thanks to all the innovations in ML & Computing
- Google is now capable to understand the meaning of any text
- Google now understands relevancy & intent
- And we're all moving towards BlackBox Algorithms.
BlackBox Algorithms mean - there is no one writing If Else conditions or algorithms. Just feed a lot of data to the machine, and let it decide what works and what doesn't. Keep training and after a time, you've perfect algorithms. 🙌🏻
And none knows how these whole Algorithms work completely. Not even Google employees to be exact. It's all machine-trained data with a focus on a lot of factors.
The core is with a focus on NLP ( Natural Language Processing ) and user feedback. Sprinkle some external parameters - and our BlackBoxes are ready. 😅
Can you see the problem here?
We're still using those same warmup algorithms and tech built in 2010-15 to bypass spam filters - and it's still ongoing. Because it just works & there are enough buyers.
Let's Understand Email Warmup
Email warmup is simply focused on sending some normal emails from your account to some random user - to avoid promotion to real usage ratio.
So, user A will send emails to users B, C & D. Now maybe user C will reply to A. And the same way B will send to A, E, F, and so on.
The whole network of emails will talk to each other to create fake email conversations - to prove that the user is not just sending promotions - but a real user as well.
This is a really interesting concept. It was initially started by some freelancers manually, and later on, some companies started automating the same.
The goal is to - create fake conversations and balance your promotion ratio.
So, if you're sending 25 promotional emails a day - you send around 20-40 warmup emails daily - to make the account look legit.
I love this tech and proves how creative we marketers are. Hats off to the brain behind this email warmup idea.
Companies providing warmups have pools of thousands of email IDs in the network - to make it look natural. Also, some providers added GPT-3 to create real-looking emails in warmup - to make it even more real.
A lot of love ❤️
So, Where is the problem?
The problem is - like every one of us - Google also knows how warmup works. And it has been almost 7-10 years of warmup tech - so it'll be dumb to say that Google has completely ignored warmups, while heavily improving their spam detections.
It works - definitely. Warmup helps you improve deliverability. But now sending random emails to balance emails is a very small part of spam filters.
Remember that BlackBox? Welcome to the world of AI. 🤖
Gmail is now a lot more careful around your email content, whom you're sending to, your personal account activities, and many more parameters.
So just sending random emails to random email IDs might not be the best idea in 2022. Definitely, it works, but it's more like something is better than nothing.
Can we again rebuild the warmup technology with a focus on today's internet? Let's try.
How GetMails is Rebuilding Warmup
We know that the current warmup works - so no meaning to cut it. But we can definitely improve it with the current requirement.
And add a second layer of warmup mainly focused on content-based detections.
So, we split the email warmup into two parts.
1) Account Level Warmup
2) Campaign Level Warmup
You can control both of them individually. Let's go deep
Account Level Warmup
Account level warmup is a bit similar to what we have till now, with some extra relevancy factors.
Here is how our relevancy algorithm works.
Instead of restricting to only Gsuite providers or premium accounts - we decided to bring all kinds of email accounts to our email pool. And provided scores to each of them based on multiple factors ( including email provider ).
Now combined all these IDs in our pool - and passed them to our next filter - Persona Pool.
We created 20 different personas for our pool - based on email activity & account campaigns. It helps us identify the industry or activity type for each email ID.
It can be B2B Marketing, B2B Local, B2C Local, and so on. Each email account is set in a specific bucket. We might change your email bucket with time, as we get more data for better accuracy. ❤️
And once we've identified the category for your account with your account score - it's time to start the warmup process.
Now we're creating a mixture for your account activity - based on these factors. Like a B2B Marketing guy will have more activity with B2B people, than some local fruit seller - and so on.
Also if you're contacting someone for the first time, there are chances that you'll contact them again in the next few days or even the next day - as conversations are mostly long-term via email.
And many many more factors to replicate human behavior.
Also, we've added Newsletter Subscription to add your email account to some newsletters - for some incoming promotional data. As our inboxes are not perfect in real life. So, we need to put some mess as well 😅
Now bring that same warmup tech here with all the relevancy and the account level warmup is ready.
We're still learning & improvising with the tech - but this is just a start to replicating real human behavior as much as we can.
Because we're not perfect with our email activities - so any algorithm should also not behave perfectly.
Now let's move towards our second part - Campaign Level Warmups.
Campaign Level Warmup
Do you know what is the most interesting part of BlackBox algorithms? They are trained on data. So, no human is defining exact standards or mechanisms.
Definitely, there are some defined parameters, but after that - it's all training. And here is the problem as well as the opportunity.
There are two kinds of Machine Learning models.
1) Defined Models
2) Continual Learning Models
Defined Models are trained once on a lot of data. And now they are ready to use for anyone.
Continual Learning Models are also trained - but they are still learning with every interaction.
Google & Gmail models are with continual learning. They constantly learn every time some user reports something as spam, mark as not spam, or provide feedback. 🧐
This is interesting - but can you see the problem?
If users are training them, users also have control to manipulate them. We've seen Google messing up with some facts in search engines regularly - when some SEO people are playing with their algorithms.
Can we do the same with Gmail Spam Filters? Let's try.
Here our goal is to - take any specific campaign - send it to accounts related to your target audience - and start marking as not spam.
Understand the difference here. We're not sending any random email - But your specific campaign email - and constantly training algorithms that this is not a spam email. ❤️
1, 2, 3, 4, 5, 50, 100.... and after you have provided enough not spam reports - the algorithm is trained not to mark that specific email content as spam. And we're set.
This might be a bit complicated concept to understand - but it's like the mass report of non-spam for any campaign - and once the algorithm has accepted the mistake - Move to real-life emails.
I'm not going into much detail here, as we're utilizing some complex parameters here to achieve the training output - but hope you got the basic concept here.
Welcome to Warmup 2.0
Just combine relevancy-focused Account Warmup with Campaign Warmup - and now you've one of the most powerful warmup systems ready for today's internet.
Definitely, this is not the end - we're in a cat-rat race and things will evolve over time as well. And so we might also jump with the next updates - as we get more data & results.
But this is the start of a new journey of email warmups.
We're launching our Warmup 2.0 by end of November 2022 or December 2022. Welcome to GetMails
What can be the best next article for GetMails?
- How we're changing the way email data works in cold outreach with help of Intent
- How we're training our AI Email Writer to help you get better output
Or maybe for our other products in Realtime SEO Data or Data-backed Hosting Business or Brand Monitoring technology. Let's see
Subscribe to stay updated. I'll be personally writing all the articles on this blog to help you understand how we're innovating in the SaaS industry. ❤️