Issue No. 11 | June 16, 2026

For warranty administrators who manage loss ratios, reinsurance, and contract performance.

This Week: Your Agent Book Has A Distribution You Aren’t Seeing

THE CURVE: 7

Seven lifecycle states. That is how the Dark Sky Data Retention Analysis Tool classifies every account in a dataset:

• New
• Retained
• Returning
• Lost
• Inactive
• No Longer a Customer
• Not Yet a Customer

Retained accounts are further classified as:

• Retained (Incr)
• Retained (Flat)
• Retained (Decr)

based on volume movement relative to the prior period and a configurable threshold.

The tool was built for private equity diligence to understand whether a portfolio company’s customer base is expanding, stabilizing, or quietly deteriorating beneath the topline. The same logic describes what a warranty administrator manages every month across a book of agents and dealerships, whether it is measured that way or not. The configuration parameters — inactivity window and minimum activity threshold — are adjustable instantly. For an administrator book where agents may go quiet for 60 days before resuming submissions, setting the right inactivity window is the difference between accurate classification and systematically overstating churn.

Your agents are accounts. Your dealerships are accounts. Each one has a current classification. Few administrators measure that distribution systematically.

THE ADMINISTRATIVE ANGLE

An administrator managing 80 active agents saw $42M in total submissions last month. Here is what that number does not tell you: one of your top five agents dropped from $600k to $200k in annual premium over 18 months. Three others have gone from regular contributors to occasional. Six haven’t submitted in four months but are still on the roster. The blended total looked fine. The book is not.

Volume totals tell you what the book produced. They do not tell you which relationships are healthy.

Account-level classification makes that movement visible. Every agent tracks as Retained (Incr), Retained (Flat), Retained (Decr), Inactive, or Lost across every period. You are no longer reading a total — you are reading a distribution: how many relationships are expanding, stable, contracting, or at risk before the volume is actually gone.

Two configuration parameters make the classification accurate for an administrator book. The inactivity window determines how many consecutive zero-submission periods pass before an agent moves from Inactive to Lost — because an agent quiet for 60 days before resuming is not lost. The minimum activity threshold ensures a single small submission from an otherwise-dormant agent does not reset their status to Retained. Both parameters adjust instantly and recalculate the full output across every agent and every period in seconds.

FROM THE BLOG

Retention Analysis Is Live: Customer Persistence Analytics for Private Equity

This week’s post announces the launch of Dark Sky Data’s Retention Analysis Tool. Beyond lifecycle classification and standard retention metrics, the post details why configuration flexibility — adjustable inactivity windows, minimum transaction thresholds, and parent/subsidiary aggregation — makes the tool materially faster and more accurate than custom Excel models for analysts and operators running account-level diligence under time pressure.

THE RESERVE QUESTION

If you measured the distribution of relationships across your book today, how many would be expanding, stable, contracting, inactive, or lost? And how does that distribution compare to where it was 18 months ago? And if the distribution has shifted, when did it start — and would your current reporting have shown you that?

Until next Tuesday,

If you want to walk through how account-level retention analysis can be applied to an agent or dealership book, reply here and we’ll schedule 20 minutes. We read every reply.