All-AI List management system : Chapter 4: Business objective

How do you manage a list of users? That is what I specialize in. I extract the most value possible from a given list of leads and/or customers. Many, many factors go into optimizing the return on such a list, and a good portion of these factors can / should be improved by AI. In fact, I am here to say that, at long last, it is now possible to set up an entirely automated list management system, providing a personalized experience to every single person on the list, thanks to a series of layers of decisions made by AI.

Throughout this series, I break down each of these layers, and explain how to best leverage today’s AI technologies to maximize the engagement / revenue you can get from your list of users, without ever discarding tried and true human-driven tactics. The winning formula for list management will always be a combination of AI and human-made decisions. What will change is the delineations of tasks assigned to AI and those reserved for human judgment.

Today’s layer:? Business objective

Once you know what is the correct contact frequency for each user, the time has come to determine what you are going to talk to them about.

This is where operates the reversal of the fundamental way humans have been marketing to each other since the dawn of time. This is the polarity change I have been announcing and demanding for years. This is what AI allows that no other human-made system ever could.

We are now - at long last - moving away from the structure “product-finds-audience” to a completely individualized “human-finds-product” system. Every single user coming in has a different likelihood of converting to different products or features of a given brand. Some call this likelihood ‘propensity”, especially if they work with machine learning. Others can refer to it as the correct sequence of conversions for a given individual, if their thought remains company-centric. You could also be customer-centric and think of this as providing each user with what they most need, or what they could best use next in their own unique personal experience.

What’s propensity?

A platform like Iterable (or rather, Iterable. How many platforms are “like Iterable”?) allows marketers to create with one click a propensity score for any event flowing into it from a database or customer data platform. Marketers are best positioned to know which “conversion events” are the most important to go after, and their data science team can help them by providing them with upstream events that are strong predictors of these downstream “conversion events”.?

Therefore, any marketer with an Iterable instance connected to their database or CDP can easily create new user properties of the type: “propensity to convert to event X” for any number of key conversion events which they are free to determine, and have the value for these properties (i.e., their likelihood to convert to each event) update daily. Then, marketers are free to set up automations which trigger when the propensity for a certain event passes above or below a certain threshold. It is also possible to segment in or out groups of users within a certain propensity range when making business decisions in a journey or a blast campaign.

So how do you make the most of all this power??

A cascade of decisions

You know best what are your most important business objectives. You must know what your most important conversion events are, and I hope that you have a team that can help you zero in on crucial upstream events which are predictors of downstream events, but that is not necessary for this plan. At the very least, you must be able to boil your goals down to 1-3 absolutely key events which you must order from most important to least.?

So, once your user has exited the workflow which controls the proper delay between emails, it’s time to send them something. And before we split test the communication itself, we need to decide what the topic of the communication will be.?

This is where a cascade of decisions needs to be built. It is the linchpin of this entire system. The exact way in which you prioritize your business goals will be the main source of testing and optimizing but, in case you're unsure of how to build this, here's a good starting point.

A solid way to kick things off

First, pass your user through a filter that will check if they have a high propensity to convert to your most important business goal (make sure they are eligible to do so), and if they do have a high propensity, send them to the journey dedicated to that business goal. If they don't have a high propensity to convert to that first business goal, move on to the same question for your second most important business goal, and keep going until you’ve run out of propensity scores for key business goals. If a business goal can’t be turned into an event and therefore cannot receive a propensity score, then do what you would have done otherwise and manually configure your best human guess for what would constitute a good user profile (or persona) to target for that business goal.?

This series of filters will help users get into the journey targeting the business goal they are most likely to convert to. Now, the majority of users likely won’t be at a “high” propensity to convert to any business goal. You’ll need to decide what to do next with those remaining users.

Two ideas for lower propensity users

If you have a sophisticated enough platform like Iterable, and you have a good data team, or well-defined personas, then you can probably make a good decision with respect to which business goal you would like to promote to which kind of users. You may have noted that, at this stage, we are back to the crude segmentation by hand that I’ve been decrying for years. But at least, we’ve resorted to this only once the AI option has run out. These lower propensity users will still get a more personalized send cadence, channel and time of day optimization, and will get the correctly themed message if and when their propensity to convert to a key business goal reaches “high”. So this is still a vast improvement over the previous all-manual list management system.

Another possibility is for you to continue to add filter nodes for progressively regressing propensity scores in succession, and work your way down until everyone has gotten the topic they are most likely to convert to. For your 3 business goals, you’d have a succession of 3 nodes that check for 90% propensity, then 3 nodes for 80%, and so on.?But there is a caveat to this technique, I discuss it further down.

Don’t forget the branding

You can begin the business objective splitting journey (the journey which is the object of this post) with a quick percentage split node and send, say, 15% of users down a separate path called “branding”, where users will receive not a conversion communication, but instead a “cool new fact about your brand” communication. But a lot of the branding can be time-sensitive and require timely batch & blast communications. Plus, there are other ways to include evergreen branding content in each separate business goal journey. However, sending a certain percentage of users in a branding journey right from the start is a good way to control for your branding-to-conversion weighting in your lifecycle communications, so I thought I’d highlight this option.

The last resort

If everything else fails, send the remaining low propensity users to a business goal splitter node, where they are randomly assigned to a business goal. Below 60% propensity score for a business goal, you are pretty much grasping at straws anyway. Propensity models aren’t as good to predict the entire gamut of likelihoods, from most probable to most unlikely. They usually specialize in a pretty precise range that is usually something like top 15%.

Don’t forget, platforms like Iterable allow you to weight each split, so you can have that randomizer send 60% of the time to your most important goal, 30% of the time to goal #2, and 10% of the time to goal #3.

Conclusion / Foreshadowing

And that is how you let AI assist you in assigning each user to their best topic of communication.

Next, we will look into how to best manage the way in which users are contacted with respect to their optimally chosen business goal.

NB: I truthfully do not get a quote if you sign up with Iterable. I’m just a fan.