There is no doubt that artificial intelligence and machine learning has come a long way. It can now amongst other things be used to create a better user experience and to simplify the customer’s buyer’s journey. Here is how you can take advantage of machine learning in both the marketing and sales hub of HubSpot’s software.
Back in the eighties when I studied at the university, I was in a small group spending quite a lot of time trying to understand the then current state of artificial intelligence. A.I. was mainly – at that time – restricted by far less powerful computers that were available, compared to the computers or even smartphones we use today.
Let us fast-forward 30 years, and some things in the A.I.-field seems very much unchanged regarding the discussions of what A.I. is capable of offering you. However, it is also apparent, that you now see more and more real products and features emerge on the market, that in one way or the other, has embedded some A.I. approach.
One subset of A.I. that has started to make an impact on our daily lives is machine learning.
What is machine learning?
It is when machines learn. It is as simple as that. More precisely machine learning is a subset of the total field of A.I. The basic idea is that you by using statistical techniques can give computers the ability to "learn" based on the data you present to them.
Instead of precisely telling the computer what it is supposed to do (which is what a standard computer program is doing), it is given an "environment," where it can improve the performance on a specific task over time. Pattern recognition is often the base of machine learning, combined with algorithms that can learn from data and make predictions based on data.
One way to look at machine learning is this: It can give you access to hidden insights that you are not able to discover yourself – at least not as quickly as the computer can.
Practically, you base most machine learning on you sending considerable large datasets to the computer and then let the computer find patterns in the dataset. The computer "learns" by getting an understanding of what patterns constitute success and what patterns constitute a failure.
Machine learning inside the Marketing Hub
A HubSpot portal is very much the source of precisely such an extremely large dataset: your contacts, their lifecycle stage, their timeline actions, company info, your content, social messages and so on.
It is a great thing that HubSpot is actively implementing more and more A.I.-inspired features into the portal and this also goes for the use of machine learning techniques inside the platform.
Let’s have a look at how A.I and machine learning can be used in the HubSpot platform today and how it can solve some of the most challenging marketing and sales problems.
The new Predictive Lead Scoring
Lead scoring has always been a problematic marketing resource to establish and take advantage of in your inbound strategy. However, it is correspondingly important.
With the new Predictive lead scoring version 2 in the platform, the system is now via machine learning looking at your dataset (contact interactions, website visits, read emails and so on) and not some arbitrary outside-in look. This will over time make the predictive lead scoring much more precise and useful, for example in workflows.
Content Strategy (auto-clustering and audit)
With this machine learning-based feature, you can create new topics automatically based on your existing HubSpot pages and blog posts. This feature is a great way to get some help from the system when you want to start working with topic clusters (which is highly recommended, by the way).
This feature helps you optimize and organize your current content. Through machine learning your platform finds the broader topics and also identify failures and success.
Blog Post Recommendations
Wouldn't it be great, if your marketing system automatically could promote the next recommended blog read for your visitors? Well, you can now in HubSpot Marketing Enterprise, where recommendations can be shown based on what your leads have already been reading.
As with any machine learning model, the suggested reads will be better and better over time automatically.
When setting up a new lead flow: do you have a good hunch of what pop-up type would be the best in the specific situation?
Where should it appear
How many seconds should you wait for it to show?
HubSpot has a machine learning based solution (in early beta at the moment) to this challenge. You can choose to let a lead flow make its own best choices. The system will then test out the different lead flow types automatically and find the best possible solution.
This method is far more advanced than a standard A/B split testing, as in an A/B-test you risk sending much traffic to a lead flow that is not the best-fit solution depending on the single contact.
The HubSpot system will test the different types and then quickly start the continuous optimization process. This optimization is not only a general optimization. It also takes other elements into account like the contacts in your CRM and properties like Lifecycle Stage, Lead Status and which device the user is on. The model is much more complex than this, so just to let you know that this is just a simple example.
Usually, A/B tests stop, but this continuous improvement process will be going on "forever." It gets smarter all the time and will send traffic to the "best" version. Over time, you will see this type of continuous improvement processes in more and more existing features of the HubSpot platform.
Machine learning inside the Sales Hub
Sales activities also quite naturally lean to be able to gain help from machine learning. As a sales rep you are often performing quite similar tasks day in and day out; for example, sending emails, having calls and producing documents. So, it would be great to have some assistance by your side, that can give you professional guidance.
Here are a few examples of existing usage of machine learning in the Sales/CRM part of HubSpot.
Sending Time Recommendation
You have just written the perfect email answer to an important lead. So, when is this lead most likely to open your emails? The HubSpot system can give you recommendations on the perfect send time by using machine learning.
HubSpot has plans to move this functionality to your marketing emails as well, which could help to improve the open-rate efficiency of those emails automatically.
If you use HubSpot's own calling feature inside the CRM, you can now request to have a call transcript made based on the recorded call. There are no humans involved, as a machine learning algorithm does this transcription.
The feature will also be able to detect parts of the text and recommend task suggestions.
Well, this is just a little too good to be true – so what’s the catch?
Yes, there is a catch. Most of these machine learning based features I have described here will only be available in Enterprise platforms, not Starter and Pro.
Why is this? As you might remember from the beginning of this article, machine learning works best when it can find patterns in massive data sets, which is much more likely to find in an Enterprise platform.
HubSpot has just started unveiling these first examples of using machine learning techniques. The HubSpot platform is h a great place for this, as you have both your marketing automation platform and your contact database inside one system. That combination can be a killer solution for these type of tools to solve some challenging marketing and sales problems.
According to HubSpot, we can expect machine learning to spread out in the marketing and sales platform in terms of:
Website platform pages
HubSpot is also working on figuring out how machine learning can help you when creating content: What is working – and what is not? For example, HubSpot can help you choose the right images and suggest what kind of social media posts you should produce.
The machine learning technology is now usable in marketing and sales, and it is clear that HubSpot will extend the possibilities and benefits to a great extent in the coming time. That pattern is clearly recognized.