Skip to main content

Limitations & accuracy

What Civic Informer's figures can and cannot tell you, stated plainly. Known limits of the data, the methods, and the comparisons.

Honest reporting includes being honest about limits. Everything below is a known, deliberate boundary of the platform. None of it is hidden in the numbers; this page exists so you do not have to discover it yourself.

Records, not ground truth

Civic Informer reports what departments record. Crime that is never reported to police does not appear here, and national research consistently finds that a substantial share of crime goes unreported. Our figures describe recorded public safety activity, not the total amount of crime that occurred.

The reverse also applies: a call for service is a request for help, not a confirmed event. A "burglary" call may turn out to be a false alarm. This is why the platform keeps calls, incidents, and arrests distinct (see Where the data comes from).

These are not official crime statistics

Civic Informer is not an FBI UCR or NIBRS submission, and our figures will not match a department's official annual crime report. The differences are structural, not errors:

  • Our violent designation is broader than the FBI's four-offense definition (documented in How we categorize).
  • Official statistics are compiled on annual cycles with their own revision and audit processes; ours are daily and reflect the source as of each run.
  • Some cities display their department's own category system rather than a national one, by design.

For year-over-year official comparisons between cities, the FBI's published statistics are the right tool. For what is happening in your city and neighborhood right now, that is what this platform is for.

Arrests are a lower bound

Arrest details are extracted from incident narrative text. Narratives that do not follow the expected format can cause an arrest to be missed, so arrest counts should be read as "at least this many," never as an exact total.

Geocoding has edges

Neighborhood assignment comes from geocoding block-level addresses against the city's own published boundary files. Two limits follow. Locations near a neighborhood boundary can occasionally be assigned to the wrong side of it, and records that cannot be geocoded at all (including records whose location the department withholds as confidential) are excluded from maps and neighborhood breakdowns rather than guessed at. Citywide totals are not affected; a record without a location still counts toward the city.

Small samples are floored, windowed, or withheld

Daily counts in a single neighborhood, or daily emergency response volumes, are small numbers, and small numbers are noisy. The platform's protections are described with each statistic: quiet neighborhoods are held to a minimum baseline in the Activity Index, response-time figures are computed over multi-week windows, and comparative claims are withheld entirely below a minimum sample. Where a figure would be statistically meaningless, we do not show it.

Comparisons have boundaries

  • Across cities. Departments differ in what they publish, how they describe it, and how quickly. Category definitions on this platform are consistent, but the underlying recording practices are not, so cross-city comparisons of raw volume should be made cautiously.
  • The Activity Index. Each score is measured against its own local baseline, so scores are not comparable across surfaces or cities by design (see How the stats are defined).

Sources publish on their own schedule

Some agencies publish in near real time; others release once a day. The platform updates daily against whatever the source has released, so a city's figures can trail events by up to a day depending on its source. Dates shown are publication dates, as documented in How the stats are defined.

Revisions flow through

Departments amend records: reclassifying an offense, correcting a location, unfounding a report. Because the platform regenerates from source data daily, those revisions are reflected on the next run. A figure you saw last week can therefore differ slightly from the same figure today; the newer one reflects the department's current records.

Found an issue?

If you believe a specific figure is wrong, we want to know. Every record keeps its original case number precisely so that claims can be checked against the department's own records. Reach us through the about page.