Success Insights

Please Note: Success Insights are only available in the ChurnZero Enterprise Edition. Please contact your ChurnZero Customer Success Manager for pricing and edition activation.

Introduction to Success Insights

Success Insights leverages machine learning to identify patterns among your customer base. First, you provide a data set of what a Churned Customer looks like. Then, we process the data set to start to identify patterns of behavior among your Churned customers. We then take your provided list of Active Accounts and start to compare their behavior to the patterns that were identified in your Churned Customer data set. This is where Success Insights start to identify which of your Active Customers are at risk of churn, and what specific behaviors (or lack thereof) are putting them at risk.

While not exactly magic, it's hard to provide specific details around the why of what is returned in machine learning results. If you want to learn more, we recommend checking out this amazing resource.

Creating a Success Insight

To create a Success Insight, click Health > Success Insights from the navigation bar on the left. Then click the Create button at the top.

Note: Your user account's Permission Sets must include the Success Insights permission to create a Dashboard.

When creating a new Success Insight you will be presented with the following screen.

Name- Name of your Success Insight
Description- Use this field to give more detail about the Success Insight
Start Date- Pick the field you will use for determining the "Start Date" of your customer. This must be a date type field
End Date- Pick the field you will use for determining the "End Date" of your customer. This must be a date type field
Audience- Choose which Accounts will be used in Analysis.
- Audience Quality- This tool looks at the Accounts in the Audience. For the best results, we recommend the following:
  - Total Accounts- At least 300 Accounts are to be included in the analysis
  - Inactive Accounts- Of your selected audience, at least 30% should be inactive
  - Accounts with Events- Ideally, all accounts being used for analysis should have event data. However, at least 75% of your Accounts should have event data in order to provide accurate results.
Date Range- The time frame in which we are analyzing the metrics. You have the option to select:
- Weekly-looking at time frames from the last 7 days, 7 to 14 days ago, etc
- Monthly- looking at time frames from the last 30 days, 30 to 60 days ago, etc.

Exclude Events- Choose which events you would like to exclude from the analysis. Use this if you feel like there are some events that are showing up as influencing Churn that you do not think are relevant.

Tips for Running an Accurate Success Insight

The general rule for machine learning is that the more quantity and completeness of data, the more accurate the results will be. Success Insights are looking at a very specific question, "Which Accounts are a high risk of Churn based off of their Events?". To get the most accurate results, we need the following information:

A healthy list of Churned Accounts to help identify patterns. Success Insights looks at the behavior of your Churned Customers to identify which of your Active Customers are at risk. The more data we have on Churned customers, the more accurate the models will be.
Accurate and complete Start Dates for Active and Inactive Accounts.
Accurate and complete End Dates for Inactive Accounts.
Robust Event data for both Inactive and Active Accounts. If you are newer to ChurnZero, consider importing historical event data.

Running A Success Insight

Once you click "Run", we will take all of the Account and Event information you have determined in your settings and begin to construct a model using Machine Learning. This is called Training the Model. This part of the process is looking at all of the behavior of your Churned Accounts and then recognizing patterns that it sees. Once the model is trained, it is then run against your "Active Accounts" Data set. This is what provides the information in the Success Insights Reporting section.

Training a model is where most of the heavy lifting of Success Insights is done. If you are creating a brand new success insight, after you click "Run", we begin to train the model. Depending on the number of Accounts and Events included, this could take some time to return results. This process will continue in the background, and you will be alerted via the Alerts Panel when the report is ready.

Models are re-run nightly and re-trained based on the cadence that you define in the "Date Range" setting (Monthly or Weekly). Setting the date range to Monthly will display results in 30 day increments and train the model every 60 days, while setting to Weekly displays results in 7 day increments and trains the model every 14 days.

Re-running the model is using the already constructed algorithm and analyzing any new Accounts that may have fallen into the segment.

Re-training the model is like a full data refresh; the model is completely rebuilt. Re-training the model periodically helps keep the model accurate as more data is collected within ChurnZero.

Success Insight Reporting

Account Risk Analysis

Here is where you can start to see the results of your Success Insight. Your Account Risk Analysis section is broken up into 3 sections:

High Risk- This is the list of Accounts that were identified as a high risk of Churn based on the data model.
Not High Risk- The list of Accounts that were not marked as High Risk. Keep in mind, this does not mean that these Accounts are "Low Risk". It just means that they were not flagged as High Risk when assessing them against the current model.
Could Not Assess- The list of Accounts that could not be assessed against the model. Usually, this means that they do not have enough events, or do not have valid start/end dates.

Once you click into one of the above sections, you will see a Summary of Accounts by ChurnScore. Clicking through the donut charts will take you to an Account Segment.

You can also see a list of Accounts included in each section. From this list, you can search for specific Accounts, customize your columns, as well as perform bulk Actions

High Risk Influence By Event Type

This section of the report will show you which event types are most influential in causing customer churn.

View - What are we displaying the data for? You can choose:
- All High Risk Accounts- This shows the influence by event type for all of your High Risk Accounts. This is good for looking for overall trends among your customer base to see which activities could lead to Churn
- Specific High Risk account- This shows which event types had the most influence for marking a specific account High Risk.
Rank- What is the importance of this metric in causing customer churn?
Event Type- Which event type
Calculation- Success insights analyze the event data in multiple ways in order to give you a better understanding of the underlying cause. These include:
- Event Instances- The number of times that event occurred within the selected time frame
- Event Quantities- The sum of the event quantities within a selected time frame
- Number of Unique Contacts- The number of contacts that performed the event within the selected time frame
- Active Days- The number of days that event occurred within the time frame
- License Utilization- The number of Contacts in the Account that performed that event in relation to their License Count
Date Range- The time frame in which we are analyzing the metric. The date ranges will be different depending on whether you selected "Monthly" or "Weekly" in your settings. However, you will always see the dates with respect to the "First" or Last".
- "First" always references the field that was selected for "Start Date". So first 30 Days means "first 30 days since the Start Date".
- What qualifies as "Last" changes depending on whether you are referring to Active or Inactive Accounts. Inactive Accounts refer to the field chosen as the "End Date" in your initial settings. Active Accounts will use today's date as a reference since they do not have an end date yet.
High Risk- The average metric value for an Account considered "High Risk."
Low Risk- The average metric value for an Account considered "Low Risk"
Account Value (only available when a specific account is selected)- This shows what the metric value was for the selected account, allowing you to see where it is in relation to High Risk and Low Risk Accounts

You will probably see an Event Type listed multiple times, and that is expected! That is because it could be listed with a different calculation (Number of Unique Contacts vs Event Instances), or a different Date Range (Last 30 Days vs First 30 Days).

NOTE: If a specific calculation isn't included in the report, the calculation is not considered a high-risk factor. (ex: If the Event Instances calculation isn't included, the number of instances of the event doesn't play a role in considering the account "high risk."

Box and Whisker Chart

This displays the distribution of the metric broken up by Active and Inactive Accounts. Start by selecting an Event Type metric from the table on the left. This will then update the chart on the right to the chosen metric.

On the Box and Whisker chart, we can see the distribution for the number of days the "Login" event was performed over the last 30 days, broken up by Active and Inactive Accounts.

You can also select a specific Account to see how their actions plot against the averages.

Looking at this data closely can help you draw conclusions about your customer behavior and what factors may lead to churn.

Note: There may be a discrepancy between what is listed as the maximum/minimum value, and what is shown on the chart. In order to keep the chart readable, we do not chart extreme outlier values.