Estimated Revenue

To display the estimates in RiskMatch, from the Show me segment of Page Settings: use the dropdown and click Estimated Amounts. Once Estimated Amounts is the chosen view, all insights being displayed will be based off the predicted model.

Since Written revenue is a value displayed across several RiskMatch dashboards and reports, the Written view on RiskMatch is used to view premium revenue and fee upon policy inception, which is useful in understanding where business is coming from before it is billed.

Because the value is pulled from the AMS, which may not be a mandatory field (or is frequently left empty), Estimated Revenue was designed to improve the insights business leaders need to project their revenue stream over time.

To best illustrate this issue, the RiskMatch team has reviewed in-force policies across different AMS systems at the policy level to determine levels of Premium versus Revenue entered in the system(s). Optimally, when Premium and Revenue is entered (far right graph group), leadership is provided with best insights.

The RiskMatch team reviewed the current Written decision tree (see left table below) based on the AMS data depicting 50 percent of AMS360 policies with missing revenue, and created a Partial Billed backup based on policy billings that may, or may not be complete. While this may be the case, this data set will provide a better estimation than a zero-dollar value.

While the Written decision tree is broken out across All AMS (TAM, EPIC, etc) and AMS360, the Estimated decision tree (see tables on the right, below) was developed based on the learning experience of the Written tree and, as such, encompasses All AMS.

For the Estimated Revenue decision tree, when a broker enters both the premium and revenue values, these values are used by RiskMatch to report on in-force policies.

When Premium is entered and Revenue is missing, the entered Premium value is used in conjunction with the Estimated Revenue developed by the RiskMatch team.

Finally, when both values are missing, the logic returns information based on the Partial Billed backup, designed in the Written instance.

The graphic below is designed to depict how a very basic Written model was used as a foundation for the current Learned model, which will deliver insights that leaders can rely upon.

Using the Estimated decision tree on in-force policies, RiskMatch looks at multiple data points or sources to derive the Estimated Revenue displayed when run. The chart shown below is the same as the original Estimated tree from above for All AMS, indicating the premium and revenue sources, from each of the three scenarios to develop the predictive model.

For greater detail regarding RiskMatch sources, please view the How-To video for Estimated Revenue:

RiskMatch has built a specific Data Quality Report to help with both Written and Estimated amounts. From the RiskMatcn Feature Menu, choose Analytics and Data Quality Report to access the In Force Policy Financials DQ Report.

By filtering the Business Division and External System, the report will display areas in the organization where data may be lacking, or where the quality of the data entered could be improved.

Types of policies this report targets:

• Policies with very large AMS-entered premium or revenue:

  • Policy flagged when the AMS-entered premium or revenue is in the top 0.1percent, when entered premium > $700,000 or entered revenue > $50,000

• Policies where the partial-billed premium is less than the AMS-entered premium:

  • Policy flagged when billed premium is < one percent of the AMS-entered premium.

• PL & CL policies with large AMS-entered revenue rates:

  • Policy flagged when the AMS-entered Revenue Rate >= 50 percent.

• Older policies that do not have any bills:

  • Policy flagged when the policy is at least one year old and has $0 in partially billed premium (legacy billings are considered).

Goal:

This change seeks to sharpen existing revenue estimations by reworking the model’s priority order for the bonds-and-benefits insurance segments.

Background:

The existing Estimated process followed four stages, ranked as follows:

1. AMS Entered Values

2. Predicted by ML

3. Predicted via Partial Billing Rev Rate

4. Actual Billed Fallback.

We recognized that the estimated values for the "Predicted by ML" category in the bonds-and-benefits space were less accurate than values contained within the commercial or personal lines spaces.

Change:

As a result of this change, we have restricted the bonds- and-benefits insurance segments from training in the model, and we changed the Estimated selection order for these as follows:

1. AMS Entered Values

2. Predicted via Partial Billing Rev Rate

3. Predicted by ML

4. Actual Billed Fallback.

Analysis:

This alteration has greatly improved the Estimated Revenue results. We know this because we can test our Estimated predictions several months later, when the billing cycle completes. Bond and benefits segments will now have a greater emphasis on “Predicted via Partial Billing Rev Rate,” which requires an AMS-entered premium and at least one partial billing. This step multiplies the AMS enter premium by the revenue rate of the first billing. We have found that the first billing is a highly accurate representation of how the policy will look. If there is no AMS-entered premium, we use the actual billed values as a fallback for estimated revenue and premium. In this fallback case, we would have no AMS-entered premium by which to multiply the partial billing revenue rate.