Editorial standards

Civic Informer's reporting is computed, not written. Every page, chart, and email edition is assembled by software from the day's verified statistics, using fixed rules and fixed language templates. This page defines that standard precisely, because "data-driven" is a claim that deserves a definition.

Every sentence is generated from data

The prose you read on a daily report ("Incidents were up 12% from last week, led by Downtown") is produced by templates: pre-written sentence structures whose blanks are filled by computed statistics. The template library was written by people, once, and reviewed like any other part of the product. The words change only when the numbers do.

This has a property no newsroom can offer: determinism. Given the same day's data, the platform produces the same report, every time. There is no editorial judgment call between the data and what you read, no framing decision made per-story, and no possibility of a sentence drifting from the statistic behind it.

Nothing is interpreted

Between an official record and a published sentence there are only the documented, mechanical steps in this reference:

• Numbers. Every figure is computed by the rules in How the stats are defined, directly from official records. Nothing is estimated or filled in.
• Categories. Classification is a fixed lookup table, published in full in the crosswalk. Nothing judges what an incident "sounds like."
• Words. Reports and email editions are assembled from the template library. Nothing rewrites, summarizes, or characterizes the data on the way to the page.

In public safety reporting, a subtly wrong detail is not a typo; it is misinformation about your neighborhood. Keeping the path from record to sentence mechanical is how we make that class of error structurally impossible, not just unlikely.

The newsletter is assembled, not written

The daily email brief is built by the same standard: a programmatic edition builder reads the day's statistics and assembles the edition from the template library. Every sentence in it traces to a computed statistic, and the same edition can be regenerated from the same data at any time.

Editorial protections are enforced by rules, not judgment calls made per edition. Incidents in sensitive categories, such as sex offenses and domestic violence, are never featured as notable incidents, and no names or personal identifiers ever appear.

Where humans are involved

People build and maintain the system; they do not shape individual stories:

• People write and review the classification rules, the category families, the sentence templates, and this reference.
• Unusual offense descriptions that do not match any rule are reviewed by a person and added to the public crosswalk, not guessed at.
• When a partner agency contributes a message or a post, it is published under that agency's own name and kept visually distinct from the data reporting. We do not edit agency statements, and we do not present them as our own.

Corrections

Because reporting is regenerated from source data daily, corrections flow through automatically: when a department revises or corrects a record, the platform reflects the revision on the next daily refresh. We do not manually edit published statistics; we fix rules, and the rules fix the numbers everywhere at once.