Microscope: Predictive Analytics

Cortex makes multiple predictions about each individual user on a daily basis. Cortex uses each one of these predictions to determine user behavior, likelihood to spend, and the best time to contact the user. In this article we'll go through each of the predictions that can be viewed in the Microscope tool under the Predictive Analytics section. 

Future Value

Future value is a prediction of how much a user will spend in the next three months. The dollar value displayed for each user is a prediction that is influenced by a number of different factors. A few of these factors are explained below:

1. JS Actions: These actions are the specific tracking events that Cortex collects on every user. Each action indicates a level of intent. As actions become more frequent and the duration of the time spent on the website increase, so does the probability that a user will make a purchase. This helps Cortex determine the likelihood of a purchase in the next 3 months.
2. Item Views: A view of a specific item also relays a specific price associated with the item. If a user is consistently looking at one $30 item, they have a higher likelihood to make a purchase of $30. If they are consistently viewing multiple $30 items this can increase the potential future value.
3. Order History: If a user normally spends $60 on every order Cortex, will use this information to estimate the likelihood of the user making a purchase of this dollar value again.
4. Buying Cycles: The order history can uncover specific buying cycles for the user. If a user makes a purchase every 2 months, Cortex can pick up on this behavior and will factor this in to the CFV calculation.

All of the above items can influence the CFV to move up or down, based on positive or negative inputs. Cortex will calculate and weight the possibility based on past trends and user data to come to a conclusion on the likely spend of each individual user.

It's important to remember that the value is only a prediction. This means that Cortex expects a user to spend the predicted dollar amount, but it may be incorrect at the exact value of the spend. Cortex will learn when it has made an error and adjust predictions in the future to become more accurate over time. The important thing to understand is that the users with a higher CFV are more likely to make a purchase and are a good segment of users to target with messaging designed to sell products.

Lifetime Value

Lifetime value is a much simpler value to understand. This is simply the total amount that a user has spent with your business plus the expected future value spend as outlined above. The example below walks you through this calculation: Suppose a customer made three purchases:

Purchase 1: $532.25

Purchase 2: $324.10

Purchase 3: $410.65

The traditional lifetime value for this customer would be:

$532.25 + $324.10 + $410.65 = $1267.00

The lifetime value in Microscope incorporates the customer's Future Value. If this customer had a Future Value of $250.25 then the Lifetime Value would be:

$1267.00 + $250.25 = $1517.25

This value allows you to treat higher value customers with better service, better offers and continue to gain more value from them. Remember it costs 5x more to acquire a new customer than to retain an existing customer. Treat these customers well and watch your bottom line grow along with their Lifetime Value.

Lifecycle State

The lifecycle state is a representation of how your customer is currently behaving. The lifecycle state is only predicted for customers, meaning that they have made one or more purchase. If the prediction is shown as N/A then this means the user has not made a purchase. The Lifecycle states that are predicted by Cortex are:

1. N/A: This identifies users who have never made a purchase. Work on targeting these specific user through segmentation to get them to convert.
2. Ready to Buy: These users are predicted to have the highest affinity of all your customers to make a purchase. Similar to Future Value, Cortex identifies these users by the collection of their online actions. The more active a customer is the more likely they are to make a purchase. The Ready To Buy customers are actively browsing your website and spending more time on site than the other two groups. Cortex will also pick up on a customer's buying behavior. If a customer consistently makes a purchase within a given interval Cortex will identify this interval and classify them as Ready to Buy when that interval approaches.
3. At Risk: These customers are trending down with their onsite actions. They were once Ready To Buy customers but have consistently shown a decrease in his/her onsite actions. With these customers it's important to try to reengage them with your business, through content or incentives. These customers are still able to be saved and have a higher likelihood to buy than Churned customers.
4. Churned: These customers have fallen off even further in their onsite actions. They have a very low likelihood of purchase. You can try to reengage through content or incentives, but the likelihood of them making a purchase is extremely low. Keep them around by engaging in content, so that they can be reactivated to Ready to Buy when it's time for their next purchase.

On average a Ready To Buy customer is 10x more likely to make a purchase than a Churned customer. Talk to your CSM to understand the specific statistics for your site. Keep this in mind when thinking about which groups to target with what types of messaging.

Best Time

Cortex predicts the best time to contact a user using a combination of your customer's onsite behavior and interaction with email. Cortex will predict the best hour to contact a user based on this combination with a slightly increased weight towards open. The goal is to identify the best combination of open and onsite engagement, so that the email can have an optimum effect on creating a conversion.

Average Order Value (AOV)

This is a simple calculation that is performed for each of your users. The example below outlines the calculation in detail:

User 1

Purchased 3 items on 9/1 for a total of $125.13

Purchased 1 item on 9/15 for a total of $25.35

Purchased 4 items on 9/22 for a total of $85.25

The AOV is the average of these three orders

($125.13 + $25.35 + $85.25)/3 = $78.58

This calculation will help you understand what your customers spend on average. With this information you can structure you initial incentives around this value. Either based on the AOV on the first purchase, or segment on users who spend a certain dollar amount on each order. This will help you differentiate your high value customers in different ways, and gives you a data point to increase through different incentive strategies.

That's the basics of the Predictive Analytics for Microscope! You can check out the descriptions on the other sections below:

1. Customer Profile and Demographics
2. Product Recommendations and Last Products Purchased
3. Email History
4. Activity and Latest User Actions