Call Waiting: Follow the Investigation →
When you call for help, the wait feels like one stretch of time. Portland measures it in pieces. A phone must be answered. Someone must work out what is happening. A call may enter a queue for police. An officer must be assigned, then arrive.
The newest public police data gives us a clearer view of two of those pieces. In September, the average time from entering the police queue to the first officer's arrival was shorter than in August for calls classified as high priority. Almost all of that monthly difference was in the queue, before dispatch.
That is a useful change to see. It also sharpens the next question: what made the queue shorter, and who still waited?
What Changed in September
The Portland Police Bureau's October 6 export contains 19,376 September dispatched-call records. Of those, 4,789 are classified as high priority. Response duration is available for 4,606; the remaining 183 have no reported duration.
Among the measured high-priority responses, the average fell from 22.83 minutes in August to 20.55 minutes in September, a decrease of about 10%. July's average was 20.61 minutes.
The middle of the distribution changed little: the median went from 10.20 to 9.95 minutes. The longer waits changed more. The 90th percentile fell from 48.56 to 41.88 minutes. In ordinary language, about nine in ten measured responses were at or below that threshold, while roughly one in ten took longer.
About 14.9% of September's measured high-priority responses exceeded 30 minutes, compared with 16.6% in August. A lower monthly average can coexist with a difficult wait for an individual person.
“High priority” here includes priority numbers 1, 2 and 3. It is a broad reporting category, not a count limited to immediately life-threatening emergencies. Looking at those three priority numbers separately, each had a lower mean in September than August. That does not mean every type of call or every part of the day improved.
Where the Time Accumulated
On the same records with complete timing information, the mean wait before dispatch fell from 14.61 to 12.34 minutes. The mean interval from dispatch to arrival was nearly unchanged: 8.22 minutes in August and 8.21 in September.
This places the monthly change before dispatch in the published timeline. It does not tell us whether it came from staffing, fewer simultaneous incidents, different duties, changes in the calls themselves, or something else.
These measurements begin when a call enters the police dispatch queue. They do not begin when someone dials 911. The field labeled travel time starts at dispatch, so it should not be read as driving time alone. The export also leaves some intervals unreported under PPB's publication rules. Missing durations are unknown, not zero. PPB's definitions explain the boundaries of these measures.
A Citywide Average Can Hide a Different Evening
For high-priority calls with a recorded queue-entry time and response duration, the midnight-through-5:59-a.m. mean fell from 22.4 to 14.1 minutes. The evening mean, from 6 p.m. through 11:59 p.m., stayed near 26 minutes: 25.6 in August and 25.8 in September.
That contrast is a reporting lead, not an explanation. These are different sets of calls. The time-of-day comparison excludes measured responses without timestamps, including entire sensitive categories in some months. It cannot describe everyone's experience.
A further check gave the two months the same mix of priority and final call category, using categories with enough measured calls in both months. The September mean remained lower, by 2.74 minutes in that comparison. It retained about 95% of measured high-priority responses in each month. This reduces one source of misleading comparison. It does not account for officer availability, overlapping incidents, severity within categories or unreported waits, and it does not establish a cause.
The Other Clocks Still Matter
BOEC's latest director's report located in this review is labeled September but reports August call-answering performance. It reports an average answer wait of 23 seconds. Its answering measures exclude calls disconnected by the caller. Those figures describe a different population and stage; adding them to the police average would not reconstruct a resident's journey. BOEC September report, page 2.
PPB's staffing report, dated September 23, lists 318 officers assigned to patrol and estimates 57 working citywide at a given time. It explicitly distinguishes that estimate from the number immediately available for a new call. We need time-specific availability and workload records to test how capacity relates to the queue. PPB staffing report.
What We Need to Learn Next
The next reporting step is to ask for the records behind the queue: when calls arrived, when priorities changed, what units were available, and what other work was competing for their time. The evening pattern gives us a concrete comparison to investigate. The newly published file also revises older location fields, so neighborhood findings need a clear explanation of how those fields were assigned.
An October 7 check identified a possible record for the September 12 NE Sandy collision: a high-priority collision entered the police queue at 7:02 p.m., listed at NE 20th and Sandy. The witness account places the collision near the 2200 block. The public row records 10 minutes 15 seconds in queue and 9 minutes 17 seconds from dispatch to arrival—19 minutes 32 seconds combined. The location difference and anonymized identifier mean this is a candidate, not a confirmed match. Those durations must not yet be presented as the wait for the collision I witnessed. PPB has been asked to help identify the incident and clarify the data. This analysis does not identify an I-5 closure effect, prove intentional withdrawal of service, or demonstrate that an NG9-1-1 product would reduce these waits.
The purpose of CALL WAITING is to make the handoffs visible enough to ask useful questions. September's figures move that work forward: the recorded queue shortened overall, evening waits did not follow the same pattern, and the reasons remain open to investigation.
Source and Method: Analysis of the PPB 2026 dispatched-calls export, downloaded October 6, 2026 at approximately 3:52 p.m. Pacific and retrieved again October 7 at 1:52 p.m. Pacific. The two files are byte-for-byte identical. These are published dispatched calls, including 911 and non-emergency requests, not every attempt to obtain help. Figures use available response durations; stage comparisons require complete timing records. The analysis retained all published rows, excluded missing durations from duration summaries, checked duplicate identifiers and stage arithmetic, and compared the new release with the October 4 snapshot. The source file contains 174,304 January–September records, including 19,376 September records. No duplicate public identifiers were found; all 166,121 complete timing records reconcile queue plus dispatch-to-arrival with total response. Source SHA-256: d896d29fafe1982299500f98d334b4846b7d089806b3f0c80d090b63ecb79865.
Disclosure: I develop Groundtruth 911's Validation Workbench. This investigation does not test or claim that product's effectiveness. No operational data was sent to an AI service for a product demonstration.