Metrics Guide · 20 min read
Text message communication metrics for evidence review.
Message totals, direction, timing, and daily change can make a long conversation easier to navigate. They describe the supplied record; context and human review determine what the numbers mean.
This page explains calculations and review concepts used to describe a two-person message record. It distinguishes what can be counted from the much harder question of why a communication pattern occurred.
What communication metrics cannot say
A direction imbalance does not prove neglect. A long response time does not prove avoidance. A dense period does not prove harassment. Each can reflect many conditions: work, sleep, shared logistics, a lost device, unequal responsibilities, an emergency, incomplete records, or a conversation that continued elsewhere. Metrics also inherit source limits. If an export omits a participant, date range, attachment, deleted message, or off-platform exchange, the calculation only describes what remains. State the source range and missing-data warnings beside the number rather than presenting it as a complete picture of the relationship.
Put context beside the measurement
A chart or count should tell the reviewer which run, date range, participant orientation, and timezone produced it. Textimony’s daily chart includes a tabular view of plotted aggregates. Candidate records and excerpts have their own paths into source context; the product should not imply that clicking every aggregate point opens a legal citation. Federal Rule of Evidence 106 is a fairness rule about related portions of statements, not a prescribed interface or fixed message window. The practical lesson is narrower: do not let one statistic or quotation stand in for the material context needed to understand it.
Metrics describe; they do not diagnose
Direction balance describes how many messages or characters appear in each confirmed participant lane. Response latency describes elapsed time under a stated pairing rule. Density describes activity under a stated window. Daily change describes the difference between neighboring calendar buckets. None establishes intent, credibility, fear, hostility, attachment style, parenting capacity, or a clinical condition. Use the measurement to find a period worth reading, then return to the messages and any relevant facts outside the export.
Denominators and scope keep rates honest
A candidate count has little meaning without its denominator and scope. State the total eligible messages, participant orientation, start and end dates, skipped or unknown rows, category definition, analyzer status, and whether the count comes from rules, models, or reviewer selections. A large raw count can arise inside an even larger history of ordinary logistics. A small count can still contain an important message. The denominator prevents one kind of misreading, but it does not decide relevance. The reviewer still needs the actual source text.
A reproducible calculation names its inputs
Direct counts should be reproducible from the same normalized messages, participant map, case timezone, date boundaries, and definition. If any of those inputs change, the number may also change. A chart is not independently reproducible when its input choices are hidden. Where a metric depends on an analytical choice—an inactivity window, a pairing rule, a minimum eligible message, or a candidate threshold—record that choice when the workflow exposes it. If the current report does not expose a parameter, do not claim that it does; describe the available run metadata and limitation instead.
Direction balance: who sent how much in this record?
A direction-balance ratio compares sent and received counts for two confirmed participant lanes. The numerator and denominator must be named, because message count, character count, and conversation turns measure different things. Unknown-direction rows should be visible rather than forced into one side. There is no universal healthy ratio and no fixed value that establishes harassment, withdrawal, or burden shifting. Some logistical threads are naturally one-sided; some records are incomplete. Compare periods under the same definition, then read the exchanges that produced the change. A ratio of 1.0 means equal values under the chosen formula. It says nothing by itself about relationship quality.
Response timing requires a pairing rule
Response latency is elapsed time between a message and a later reply under a defined pairing method. Consecutive messages from one person, unanswered messages, group records, missing timestamps, and conversations that continue by phone complicate that definition. Medians, ranges, or distributions can describe the observed timing more honestly than one average. A change may be worth reviewing, especially around a time-sensitive request, but it does not prove deliberate silence, emotional withdrawal, fear, or hyper-reactivity. State the case timezone and source timestamp precision. Define which message is paired with which reply. Keep unanswered and unknown-direction messages visible. Compare like periods under the same method. Read the message content before assigning significance to elapsed time.
Conversation density describes activity in a window
Conversation density is message or character volume divided by a stated interval. Hourly, daily, and episode-based windows answer different questions. Textimony’s current topography uses completed-run daily aggregates under the frozen case timezone; it should not be described as a universal hourly density analyzer. A dense period can reflect an argument, live event, emergency, travel planning, technical troubleshooting, or ordinary back-and-forth. No fixed number of messages proves flooding or identifies the moment a relationship entered crisis. Use density to select a period, then read it.
Topic persistence is a review concept, not a motive
Topic persistence describes how many consecutive turns remain on a defined subject under a stated coding method. Topic boundaries are interpretive: one message can address several subjects, and two reviewers may reasonably code a transition differently. Textimony does not claim that its current workflow calculates an exact, verified number of semantic loops or proves that a person refused to let a subject go. Search and candidate lanes can help a reviewer locate recurring language; the reviewer should document the topic definition and examples used. Turn persistence — Which turns were coded to one stated topic and why? — Calling the count obsession or fixation. Topic boundary — Was there a clear request to stop or change the subject, and what exceptions applied? — Treating every later reference as a violation. Topic shift — Where did the exchange move between logistics, history, money, or another subject? — Claiming a topic model established intent.
De-escalation language must be read with what follows
Words such as “I’m sorry,” “let’s pause,” or “can we return to the schedule?” may be de-escalation language. Their function depends on surrounding text. An apology can be sincere, strategic, partial, rejected, accepted, or followed by renewed disagreement; the message alone may not resolve that question. Configured rules can surface apology or repair-language candidates. Textimony does not certify good faith, co-parenting capacity, or whether another person accepted a repair. A reviewer can quote the language, describe the next messages, and leave motive to the appropriate factfinder. Document the words and sequence. Do not convert politeness into a character judgment or custody recommendation.
Repeated shapes in a timeline need manual comparison
Several periods may show a similar sequence: ordinary logistics, more messages, candidate language, and then less activity. A daily chart can help a reviewer notice those shapes, but visual similarity does not prove the same cause or a predictable relationship cycle. Textimony does not infer that a repeated shape was designed to maintain volatility. Compare the underlying messages, participants, dates, external events, and source completeness. If the periods differ, preserve those differences rather than forcing them into one story.
A no-reply period is an observation with many explanations
A no-reply period means the supplied record shows no qualifying reply during a stated interval. The conversation may have ended, moved to another channel, resumed outside the date range, or been affected by missing records. “Conversation collapse” should not be treated as a clinical or causal fact. When silence follows a candidate threat or financial message, chronology can be documented without claiming the first caused the second. Textimony cannot prove psychological impact or establish the elements required for a protective order. Those questions require the full facts and applicable law.
Published by
Textimony. Editorial status: Operational guide to descriptive message metrics. The interpretations are not psychological, clinical, or legal conclusions. Updated: 2026-07-23.
Sources
Federal Rule of Evidence 901 — Legal Information Institute, Cornell Law School; Federal Rule of Evidence 902 — Legal Information Institute, Cornell Law School; Federal Rule of Evidence 1006 — Legal Information Institute, Cornell Law School; Federal Rule of Evidence 106 — Legal Information Institute, Cornell Law School; Guidelines on Mobile Device Forensics — National Institute of Standards and Technology; SWGDE Best Practices for Mobile Device Evidence Collection and Preservation — Scientific Working Group on Digital Evidence