30 Jun 2016
How machine learning can improve email marketing ROI
Machine learning is entering into a very exciting period in its history.
A Chatbot has passed the Turing Test, and learning machines are now frequently entering into arenas once thought to be the exclusive domain of the human brain, led, step by step, towards becoming involved in almost every aspect of our lives.
Indeed, it seems that algorithms, learning machines and artificial intelligence have flown under our collective radar and into our wallets on a massive scale.
Our world, currently populated by several generations raised on media and literature rife with dire warnings about insidious AI, learning at geometric rates and eventually turning on us, is seeing an outcome largely disconnected from the concerns and fears we’ve been spoon fed for so long.
Instead of the nuclear launch codes, the learning machines have, thus far, been given access mostly to the photos we upload to Facebook, the films we watch and our shopping habits. Data, which while mildly creepy, feels pretty harmless.
Because it is.
Instead of our planet, the machines want money, because that’s what we’ve told them to want.
And they are getting it.
As of 2009, High Frequency Trading (HFT) algorithms accounted for 73% of all US equity trading volume.
And that’s just one example of many.
The truth is, anywhere there is a big enough data set, there is money to be made through machine learning optimisation.
This is just as true in email marketing as it is anywhere. Maybe more so.
As of 2015, over 112.5 billion business related emails were sent worldwide every day.
If even 10% of these represent marketing emails (which is a conservative estimate), that means that over 11 billion marketing emails are arriving in email inboxes in every corner of the globe on a daily basis.
Some get opened, some don’t.
Some generate revenue, some don’t.
It is the very definition of “Big Data”.
Data so big, in fact, that computers are really the only practical option for sorting through it all.
And sorting through it all is what machine learning is all about.
There are patterns in all human behaviour and our responses to marketing emails are no different.
Identify and capitalise on those patterns and you’ve got a recipe for email marketing gold.
How machine learning can improve your email marketing ROI
What is the goal of an email marketing campaign?
It’s fair to say that the answer to this question can be different things for different companies.
Maybe one wants the email to drive website traffic and page views, while another wants the emails themselves to directly drive sales, while yet another is looking for newsletter subscriptions.
In any case, the underlying goal is always the same: to drive revenue.
If an email marketing campaign does not increase the revenue of the company conducting the campaign, neither directly nor indirectly, the campaign is a failure.
On this, hopefully we can all agree.
Luckily for email marketers, this rarely happens.
Email marketing return on investment has historically been and continues to be fantastic. At least when measured against other online marketing channels.
But can’t it always be better?
Of course it can.
And machines can help.
Testing emails is generally accepted industry-wide as best practice.
Even if it wasn’t, anyone looking logically at email marketing can see that testing out different approaches, styles, and tactics and focusing on those which accomplish the desired goals of the campaign makes sense.
The rub is, human responses to such things are never static. What worked last week might not work today. And what works today might not work tomorrow.
This is a given.
And, this is where machine learning comes in.
A long-term email marketing strategy that is monitored on an on going basis to identify the changes in what makes said strategy successful or unsuccessful is the smartest approach.
Which is a tall order if one tries to accomplish it manually.
Luckily, there are already machine learning programs which do it for you.
The truth is, that as far as email marketing goes, the goalposts have moved.
At this point the only question is: do we get with the times now, or play catch up later?