Saturday, 11:40 p.m. A neighbor calls 9-1-1 and reports an armed fight on the street behind her house. The operator classifies the incident, the dispatcher assigns a patrol unit, and the unit arrives in nine minutes. It finds a party with loud music and two guests arguing on the sidewalk. Nobody is armed. The officers ask them to turn the volume down and leave.
Over the radio they report "nothing to note." In the system, the record still says armed fight.
On Monday, that street shows up on the weekly map as an armed-fight hot spot. On Tuesday, command adds patrols there.
Now the reverse case. A man calls because he sees someone lurking between houses. It is logged as a suspicious person. The patrol arrives and finds a forced window and someone walking out with a backpack. There is an arrest and a booking. The record still says suspicious person, and the booking lives in another system with nothing linking it to the call.
Two failures, one cause. The institution kept what someone perceived when they dialed and never wrote down, in the same record, what the police found when they arrived.
Mexico's Executive Secretariat of the National Public Security System (SESNSP) publishes national 9-1-1 call statistics every quarter. The report with data through September 30, 2026 opens with a warning worth reading slowly: calls "are not complaints before an authority." They are reports of emergency incidents based on how the caller perceives the event.
The January to September 2026 figures show the size of that perception:
- 41,103,146 calls received by the country's Emergency Call Centers.
- 29,534,368 improper calls, 71.9%. The report defines them as calls not routed to any agency because they are pranks, matters unrelated to emergencies, or unanswered or false calls. The largest category is the silent call, with 17,584,179.
- 11,568,778 valid calls, 28.1%: those routed to an agency because they relate to probable events that put people at risk or harm them. They grew 5.34% over the same period of 2025.
- 6,312,514 valid security calls, more than half of all valid calls.
That is more than six million records in nine months, classified from what someone heard over the phone. That classification is necessary to decide which agency to send and with what priority. The problem starts when it stays as the final truth about the event.
The Technical Standard for the Standardization of 9-1-1 Emergency Call Services, version 2.0, does not treat the call as the end of the process. It defines closure as the act of recording in the dispatch system "the result of the emergency response," provided by the staff of the institutions that responded. Only that record ends the follow-up on the case number.
The operating protocol turns that idea into concrete cases:
- If there is no emergency on arrival, the unit must notify the dispatcher, who records that the incident was a false response and closes the record.
- If there is an emergency, units add the additional information and report their arrival time so it is entered in the system.
- If the agency does not report what happened, the dispatcher must request that feedback and include it.
- For high-priority incidents, the supervisor closes the record only after verifying it has information from every agency involved.
The standard also proposes indicators that depend on that closure: the share of incidents classified correctly, which compares the type assigned by the operator with the closing type; dispatched incidents that were not emergencies; and records responded to that were never closed, which according to the standard itself compromise later analysis of the center's performance.
The national standard already recognizes two different pieces of data: what was classified when the call was answered and what was confirmed on arrival. The gap is in getting the field result back complete, on time and linked to the record that started it.
Mexico has no national public figure comparing the initial classification of records with their closure. Outside the country there are studies, and what they say about the data carries over.
In 2021, Rylan Simpson and Carlena Orosco published in PLOS ONE an analysis of 515,155 phone-initiated events at the Chandler Police Department in Arizona, between 2013 and 2019. They compared the type the dispatcher assigned with the type the officer used to clear the event.
The two matched in 85% of cases. Nearly one in seven events ended in a different category from the one it had when dispatched. The most revealing part is where the changes concentrated:
- Unknown trouble: 0%, because on arrival officers always learned what it was. Most were cleared as disturbances, assisting the public or suspicious circumstances.
- Assault: 51%.
- Domestic disturbance: 58%.
- Mental health: 65%.
The categories most sensitive for police operations are the ones that change most. And there is another finding: 80% of initial drug calls were cleared as drugs, but only 47% of those events produced a crime report. The call, the closure and the formal record are three pieces of data, and they do not always say the same thing.
The authors do not ask anyone to discard this data. They warn that misclassification can distort hot-spot analyses and recommend that agencies publish both the initial and the cleared type. The classification at answer time describes a request. What the police handled is only known when someone writes it down at closure.
The map. If territorial intelligence is fed with the initial classification, the party street gets marked as an armed-fight zone and the home burglary disappears inside "suspicious person." The map shows where people get scared, not where things happened. We explained in A heat map is not a patrol plan why a map needs to distinguish between complaints, calls and arrests. Closing the record is what makes that distinction possible with the municipality's own data.
Response time. The standard distinguishes reception, dispatch, arrival and resolution times, and sets as the benchmark for urban areas an average of 13 minutes from when the call is answered until the unit arrives. If arrival time is not recorded, the indicator is calculated with whatever exists: the time someone remembered to radio in, or nothing. A municipality reporting response times without reliable arrival times is reporting an estimate.
Accountability. When a complaint comes in, the institution needs to rebuild the chain: what was classified, which unit was dispatched, when it arrived, what it found and what it did. If the record says one thing and the booking another, the reconstruction depends on the memory of whoever was on shift.
In practice, a closure useful for municipal operations contains at least this:
- The initial and final classification, as separate fields. One never overwrites the other. The difference between them is information.
- The reason for reclassification. "We arrived and it was a party" is data. A category change without explanation is a correction nobody can review.
- Arrival and closing times, recorded when they happen, not reconstructed at the end of the shift.
- What the unit found, in a brief narrative written on scene.
- The operational outcome: no emergency, false response, mediation, referral to another agency, transport, arrest or booking.
- The link to any resulting record. If there was an arrest, the call record and the booking event must each be findable from the other.
- Who closed it and with what information. Especially in high-priority incidents involving several agencies.
None of this requires sophisticated technology. It requires the officer to record what they found without depending on someone in the dispatch room typing it from the radio.
Tribuna treats what happens in the field as an event with an origin, a location and an outcome, not as a note someone passes over the radio.
In Tribuna, events are recorded by arrest, call or patrol round, with a georeferenced location. Having the origin written down makes it possible, during analysis, to separate what started as a call from what the police found on their own.
Recording happens on scene. Officers capture from Android tablets, with voice narration transcribed on the device, and the app works without signal: it stores everything on the tablet and syncs when the connection returns, with a sync center that shows the status of each record. INTELLO AI helps draft the narrative of events from the event itself and check consistency before submitting. It suggests; the decision about what gets recorded belongs to the person.
When the event ends in an arrest, what the officer captures reaches booking, the medical examiner and the judge on the same case file, without being retyped. The event that began with a call and the booking it produced stay in the same history.
That record feeds the intelligence map, with layers for incidents, heat and neighborhoods and concentration by hour, built from what was recorded in the field. And every case file keeps its full timeline, with a log of who read and who changed each piece of data.
It is worth being precise about scope. Tribuna does not replace the emergency call center or its dispatch system. It can integrate with existing systems, and which sources are connected is defined during implementation with each institution. What Tribuna puts in order is that what the municipal police found and decided is written down, with time, place and the person responsible.
Before investing in another dashboard, the public security department should answer with data:
- What share of the records dispatched to your agency last month has an arrival time recorded?
- How many records remain open after the unit left the scene?
- Does your system keep the initial and final classification separately, or does the second overwrite the first?
- How many incidents changed category at closure, and in which types are those changes concentrated?
- Can you go from a booking to the call that started it, and from the call to the booking, without asking anyone?
- Do your maps and reports use the call classification or the closing one?
If several answers are "we don't know," your planning rests on what people said over the phone, not on what the police found.
A call is not an incident. It is the first warning of something nobody has seen yet, described by a person who is afraid, in a hurry or working with partial information. The incident exists when a unit arrives, looks and writes down what it found.
Closing the record with that result protects planning, which stops adding patrols to streets because of loud parties and stops losing burglaries inside a generic category. And it protects the institution when it has to explain what it did with each call.
If your institution is reviewing how it records what happens between dispatch and closure, explore Tribuna, see how Agora organizes the municipality's citizen services, or request a demo.
- Executive Secretariat of the National Public Security System (SESNSP), Estadística nacional de llamadas de emergencia al número único 9-1-1, data through September 30, 2026, October 2026
- National Information Center, SESNSP, Norma Técnica para la Estandarización de los Servicios de Llamadas de Emergencia a través del Número Único Armonizado 9-1-1, version 2.0, sections 3.20, 3.90 to 3.95, 6.2.3, 6.2.4, 7.3.6 and 8
- Rylan Simpson and Carlena Orosco, "Re-assessing measurement error in police calls for service: Classifications of events by dispatchers and officers", PLOS ONE 16(12), December 8, 2021