Research · 2 min read
The Crime Map Has a Blind Spot: Last Week Is Not Finished Yet
Occurrence-date totals for recent periods are structurally incomplete and should not be compared with settled historical periods.
You open the crime map Monday morning and your block looks quiet. Quiet because nothing happened — or because the records haven't arrived yet? In Seattle's public data, fewer than half of property-offense records show up within 24 hours, and one in ten takes more than twelve days. Last week's map is not finished.
Our analysis of 116,787 records with a valid, nonnegative timestamp interval, reports published july 1, 2023–june 30, 2026.
How long should a recent crime view remain marked provisional?
Among 116,787 Seattle property-offense records published from July 2023 through June 2026 with valid nonnegative intervals, only 44.2% appeared within 24 hours of the recorded offense start and 16.4% took more than seven days. Recent occurrence-date totals should remain provisional.
The numbers behind the answer
Selected measures only. Denominators and interpretation stay attached so the headline cannot stand alone.
44.2%
Published within 24 hours
51,580 of 116,787 valid nonnegative publication intervals were no longer than one day.
16.4%
Published after 7 days
19,185 accepted records took longer than one week to appear.
292.5 h
90th-percentile lag
Nine in ten valid records appeared within about 12.2 days; one in ten took longer.
A recent occurrence-date map is a partial view of the period
A dashboard can be refreshed this morning and still be wrong about last week. "Refreshed" means the data was re-downloaded — not that it describes everything that happened. In this extract, 44.2% of records were published within 24 hours of the recorded offense start. 67.7% made it within 72 hours. 83.6% within a week.
Which means the newest stretch of any occurrence-date chart is structurally incomplete. That reassuring dip at the right edge of the graph? Part of it is records that haven't landed yet. The recent period isn't a smaller version of history. It's a different, unfinished product.
Safe comparison
Compare settled periods with settled periods, and mark recent occurrence-date totals provisional.
The middle is measured in hours; the slow tail is measured in weeks
The typical record isn't the problem. The median lag is 32.7 hours — annoying, but workable. The problem is the tail. The 90th percentile is 292.5 hours, about twelve days. The 95th is 622.6 hours: nearly four weeks.
If you suspect the exact-midnight timestamps inflate this, they do — a little. Excluding them, the median drops to 28.4 hours, the 90th percentile to 250.2 hours, and the share taking more than a week falls from 16.4% to 14.5%. Better. Still slow.
Evidence visual
Share of valid records published by elapsed time
This is the observed lag among records present in the extract, not a direct estimate of how complete a live feed is at each age.
A good dashboard separates data age from event age
You don't have to stop using recent maps. You have to label them. A good dashboard runs two clocks: when the data was refreshed, and how mature each period's totals are. Anything younger than a stated cutoff gets marked provisional, full stop.
And compare like with like: this month at 30 days old against last month at the same age — never against its final, settled total. One caveat for the truly careful: a single extract can only measure records that eventually arrived. Counting what never showed up takes saved historical snapshots. The checklist below turns all of this into policy.
- Display both occurrence date and data-refresh date.
- Mark recent totals provisional.
- Use matching maturity windows for comparisons.
- Keep negative and invalid intervals visible as quality checks.
- Retain historical snapshots before claiming completeness.
Questions property teams ask
Does a live crime map show everything that happened yesterday?
Not necessarily. In this extract, only 44.2% of valid records were published within 24 hours of the recorded offense start.
How long should I wait before comparing recent totals?
There is no universal cutoff, but the observed curve shows that 83.6% appeared within seven days and 95.6% within 30 days. The dashboard should show its maturity rule explicitly.
Are late records wrong?
No. Publication delay and record accuracy are different questions. A record can be accurate and still appear after an operational or administrative lag.
Can this study calculate exact feed completeness by day?
No. That requires saved historical snapshots of the dataset. A single current extract only measures lag among records that are now present.
Our methods, limits, and sources
How we calculated this
This is original 911 Sentinel research — we gathered the records, ran every calculation below, and published the aggregate dataset.
We subtracted each accepted record's offense-start timestamp from its report timestamp and summarized 116,787 valid nonnegative intervals ourselves.
- We parsed offense-start and report timestamps for the study window.
- We excluded 92 negative intervals and reported them as a quality check.
- We calculated cumulative shares and the median, P90, and P95 lag.
- We repeated the summary measures after excluding exact-midnight offense starts.
What this analysis cannot establish
- The extract contains records that were published; it cannot reveal records still absent at extraction time.
- Historical as-of snapshots are required to estimate live-feed completeness directly.
- Offense start times may be estimated or may describe broad occurrence windows.
- The result describes Seattle's publication process during this period and may change.
Sources
The raw records come from the sources below; the study design, analysis, charts, and conclusions are our own.
Related questions and practical guides
New coverage — intake open
Read recent maps with the right clock
Use the coverage map as a screening view, then account for data age and inspect the property conditions that a public record cannot show.