Layer 1
18-category keyword ontology
Categories derived from family law case review. Each category weighted against the document corpus and tuned to surface communication patterns that matter for custody matters.
Free white paperPatent pending
From Messages to Patterns
Introducing Content-Derived Dynamic Temporal Clustering: the patent-pending technology transforming custody evidence preparation.
The problemYour clients have thousands of messages.
Your clients have thousands of messages.
You have hours to prepare.
High-conflict custody cases involve thousands of messages. The evidence that strengthens a case can be buried in logistics about soccer practice and pickup times. Keyword search misses context. At typical attorney review pace, manual review of a multi-thousand-message corpus can run into the low five figures.
We built something better. And it’s patent pending.
What you’ll learn
The full methodology, the case study, and the practical limits.
- Why keyword search fails to find the messages that matter most
- How our 18-category keyword ontology and patent-pending clustering algorithm identify communication patterns automatically
- The research foundation: Gottman’s Four Horsemen and contemporary coercive-control frameworks
- A case study: 6,495 messages, 23 actionable incidents, surfaced in seconds
- How attorneys leverage proprietary technology their opposing counsel doesn’t have
- The deterministic, on-device incident-finder design: no model drift, no cloud round-trip for detection
The methodology
Three layers. One human in the loop.
CustodyComms surfaces messages for attorney review. It does not diagnose abuse or render legal conclusions. The attorney remains the qualified person under FRE 902(14); the software structures the record.
Layer 2
Gottman’s Four Horsemen
Communication-pattern framework from decades of family-systems research. We surface candidate matches for attorney review; we do not diagnose.
Layer 3
Coercive-control signals
Drawn from contemporary coercive-control literature. Same posture: the algorithm surfaces messages; the attorney determines what the evidence means.
Case study
In one documented case study: 6,495 messages, 23 actionable incidents, surfaced in seconds.
Drawn from the case study chapter of the white paper. Individual results vary by matter. The algorithm surfaces; the attorney decides.
6,495
Messages
23
Incidents
Download the white paper
Get the full 11-page guide.
We email a PDF link to your work address. One email, one paper, no surprise nurture sequence.
Patent-pending technology gives you a competitive edge. Most opposing counsel still review messages by hand. Read the paper, then run the tool on your next high-conflict matter.
Sources
- 1. Case study documented in the white paper. 6,495 messages drawn from a single high-conflict custody matter. ↩
- 2. Estimated review-time reduction on the surfaced incidents. Derived from typical attorney message-review pace (~50-100 messages/hour for substantive review). Individual results vary by matter; figures reflect one documented case study, not typical results. ↩
- 3. Measured on the case-study dataset using the production algorithm. Final review remains a human task. ↩
A note on “patent pending.” CustodyComms has a patent application on file with the United States Patent and Trademark Office covering Content-Derived Dynamic Temporal Clustering. The phrase “patent pending” refers to that filed application; it does not refer to an issued patent. CustodyComms is software for organizing custody-related communications and surfacing patterns for attorney review. It does not diagnose abuse, render legal conclusions, or determine admissibility — those remain the responsibility of the qualified person handling the matter.
