↓ Skip to main content

Milwaukee Police Phone Tracking Log Analysis

·2335 words·11 mins
Lysander Demos
Author
Lysander Demos
Lysander writes at the intersection of individual rights and collective sovereignty. Drawing from the philosophical traditions of the Social Contract and the democratic spirit of the ancient Dēmos (the people), he advocates for a society where every person’s inherent rights are held sacred above the accumulated privileges of institutions, corporations, or political classes.
Capture and Statutes - This article is part of a series.
Part : This Article

The Milwaukee Police Department produced eight files on October 19 in answer to a records request about cell-site simulators. The eighth is a spreadsheet, an export of a list the Special Investigations Division calls the “CS Tracking Log.” Each row is one request to track one phone. There are 1,955 rows, from January 2, 2020 to September 1, 2026.

The log arrived unredacted. It names the people tracked and the officers involved. A redacted copy is posted here: 08-cs-tracking-log-redacted.csv. Every figure below can be recomputed from that file unless I say otherwise.

What the redacted file keeps
#

The log has sixteen columns that carry information. The redacted file has one row for each row of the log, in order of the date the order was sent, with these columns:

  • order_month: the month the order was sent to the phone company. The day is removed.
  • status, carrier, and phone_owner_type: as the log gives them.
  • offense_category: one of fourteen categories I assigned from the log’s free-text “Crime Type.” The original text is removed.
  • under_18: whether the date of birth in the log puts the person under 18 on the day the order was sent. “Unknown” means the row has no usable date of birth.
  • days_to_provisioned, days_to_located, and days_to_cancellation: whole days from the order to each later date in the log. The dates themselves are removed.
  • css_operator_recorded, area_tracked_recorded, notes_recorded, and incident_number_recorded: yes or no, for whether the log has an entry in that column. The entries are removed.

Removed entirely: names, dates of birth, incident and case numbers, the internal project number, addresses, the notes, and the names of every officer.

The offense categories are mine. Where a row lists several offenses, it is counted once, under the first of these that applies: material witness, homicide, shooting, sexual assault or trafficking, kidnapping or child abuse or missing person, robbery, endangering safety, firearm offense, domestic violence, drugs, fleeing, property crime, warrant.

The count by year
#

Status in the log2020202120222023202420252026TotalShare
Completed - Located by CS Team1021171531481952461721,13358%
Terminated - No longer viable42327610479847649325%
Completed - Located w/o CS Team2216182533515221711%
CS Declined - See Notes3378122618774%
Phone Off18320000231%
Track In Progress00000010101%
CS Approved - No Order Received00100001
Awaiting Phone Company Action00000011
All1871712572853194073291,955

The count rose every year after 2021. There were 171 operations in 2021 and 407 in 2025. The 329 in 2026 cover eight months.

The busiest month in the log is January 2026, with 53. September 2026 has two rows, both dated the 1st, and is left off the chart.

How the operations ended
#

The log marks 1,133 operations, 58 percent, as located by the “CS Team,” and another 217 as located without it. Together that is 69 percent. A quarter were terminated as “no longer viable.”

The status “CS Declined - See Notes” appears 77 times. Only 25 of those rows have a note. In those 25 the reason is the phone or the carrier:

  • 11: the account was cancelled, suspended, or out of service.
  • 9: the order was written for the wrong carrier.
  • 2: the number had changed.
  • 2: the person was arrested before the phone company set up the track.
  • 1: the phone belonged to someone else.

None of the 25 notes records a refusal on legal grounds. The other 52 declined rows give no reason.

Carriers
#

Carrier2020202120222023202420252026Total
T-Mobile68105167177159146109931
AT&T1423454681140138487
Verizon15132341648481321
US Cellular81317181426096
Sprint - CDMA78131000092
Metro PCS10210004
Dish Network00000303
Sprint - IDEN (Nextel)01000001
Blank332218120

T-Mobile accounts for 931 operations, 48 percent. Sprint was the largest carrier in the log in 2020 and is gone from it after 2022. AT&T went from 14 operations in 2020 to 140 in 2025.

Offenses
#

Offense categoryOperationsShareLocated by CS TeamTerminated
Homicide (incl. attempted)50126%50%36%
Shooting39120%65%20%
Endangering safety, reckless injury, or battery21911%65%18%
Robbery or carjacking1819%60%22%
Firearm offense1638%59%25%
Sexual assault or trafficking1146%60%19%
Drugs734%44%38%
Fleeing or vehicle offense653%51%23%
Domestic violence, stalking, or threats643%64%17%
Kidnapping, child abuse, or missing person523%69%17%
Burglary, theft, fraud, or arson422%50%19%
Warrant, bail jumping, or harboring412%59%22%
Material witness412%59%27%
Other or unclear850%38%

Homicide and shooting investigations are 892 of the 1,955 operations, 46 percent.

By year:

Offense category2020202120222023202420252026
Homicide (incl. attempted)34529879859162
Shooting393947726410228
Endangering safety, reckless injury, or battery31101123324765
Robbery or carjacking27211425313825
Firearm offense12172515273433
Sexual assault or trafficking113611292529
Drugs910121311810
Fleeing or vehicle offense7314781313
Domestic violence, stalking, or threats422881525
Kidnapping, child abuse, or missing person355971310
Burglary, theft, fraud, or arson23572617
Warrant, bail jumping, or harboring54113765
Material witness2269796
Other or unclear1014101

The “Crime Type” column is free text, and officers’ wording changed over the years. The drop in “Shooting” and the rise in “Endangering safety” in 2026 may be a change in what officers typed. The log does not say.

Whose phones
#

The log has a column called “Phone Owner Type.”

Phone owner type2020202120222023202420252026Total
Suspect1741582402723104013171,872
Victim1111101262355
Other227010416
Blank000124512

Victims. Fifty-five operations tracked a victim’s phone. Thirty-four of them are homicide investigations. Thirty-five of the 55 were terminated, and 13 were located by the CS Team.

Witnesses. Forty-one operations are labeled as material-witness matters. The log marks the phone’s owner as “Suspect” in 35 of them and “Other” in six. Twenty-four of the 41 were located by the CS Team.

Minors. A date of birth can be read in 1,780 rows. In 110 of them the person was under 18 on the day the order was sent: 106 marked as suspects, three as victims, and one as other. The count by year is 5, 11, 10, 16, 22, 27, and 19. Thirty-two of the 110 are homicide investigations, 23 are shootings, and 17 are robberies.

Ages. The median age in those 1,780 rows is 28.

Age when the order was sentRowsShare
Under 181106%
18 to 2455731%
25 to 3463536%
35 to 4432218%
45 and over1569%

The age bands are computed from the unredacted log. The redacted file carries only the under-18 flag.

How long it takes
#

The log has four dates: when the order was sent to the phone company, when the company set up the track (“provisioned”), when the phone was located, and when the cancellation was faxed. Only about a quarter of the rows record a time of day, so I count whole days.

Setup. Of 1,640 rows with a usable provisioning date, 1,334 were provisioned the same day the order was sent, 81 percent. Within one day it is 95 percent.

Locating. Of 1,109 rows with a usable located date, the median is two days.

Days from order to locatedPhonesCumulative share
Same day13412%
136645%
213457%
39766%
4 to 723187%
8 to 148494%
15 to 304899%
31 to 6014100%
Over 601100%

The bands after day three are wider than the first four, so the bar for “4-7” covers four days.

The timing is steady across years and carriers.

YearPhones locatedMedian daysWithin 1 dayWithin 7 days
202073245%88%
2021112241%92%
2022150248%88%
2023142242%80%
2024195246%88%
2025245154%89%
2026192235%84%
CarrierPhones locatedMedian daysWithin 1 dayWithin 7 days
T-Mobile525244%87%
AT&T292248%90%
Verizon184244%84%
US Cellular56334%77%
Sprint - CDMA41249%88%

Ending. The court order runs sixty days. Of 1,682 rows with a usable cancellation date, the median is four days after the order, and 73 percent were cancelled within a week. Ninety-three ran 31 to 60 days. Four ran longer than 60.

When orders go out. Ninety-five percent of orders were sent Monday through Friday. Saturday has 65 and Sunday 29. Of the 476 rows with a time of day, 351 fall between 7 a.m. and 5 p.m.

The “Track by CSS Operator” column
#

One column is titled “Track by CSS Operator.” Where it is filled, it holds an officer’s name. It is empty in every row before January 2, 2024.

202420252026
Operations319407329
Operator named188225164
Share59%55%50%

From 2024 on, the column is filled in 577 of 1,055 rows. Of those 577, the log marks 559 as located by the CS Team, 11 as terminated, and seven as located without the team. Read the other way, 559 of the 613 operations located by the CS Team since 2024 name an operator.

Thirteen officers appear in the column. One of them is named in 281 of the 577 rows.

Three of the 577 rows are victims’ phones. Thirty-seven are people under 18.

The log does not define the column or the abbreviation, and “CSS” appears in none of the other seven records. I read it as “cell-site simulator,” and I read a name in the column as the record that the simulator was used for that operation. The production does not say so.

A second column, “Area Tracked,” is filled in 48 rows, all with orders sent from June through August 2026. Each entry is a street address, a block, or an intersection, and four begin “Located at” or “Located in.” Every one of the 48 rows names a person and a carrier and has a located date, and 47 name an operator. The column records where a known phone was tracked to. None of the 48 describes turning the device on at a place to learn who was there.

Who asks
#

The “Requestor” column holds a name. After evening out spelling, there are 259 distinct requesters.

  • The ten most frequent account for 538 operations, 28 percent.
  • The 25 most frequent account for 948, 48 percent.
  • 82 appear once.

Fourteen people created the log’s entries. One of them created 1,044, 53 percent. In 570 rows the requester’s surname matches one of those fourteen. That match is by surname only, so it is an estimate.

Repeats
#

An incident number appears in 1,869 rows. In 198 cases the same number is on more than one row, which means more than one phone tracked in the same investigation. The largest is twelve rows on one incident number.

About 130 names appear on more than one row. Names are typed by hand, so that count is approximate.

About 80 rows give a circuit court felony case number in place of an incident number.

What the log does not record
#

  • Whether an operation ran on a warrant or on the emergency exception. There is no column for it.
  • Whether a cell-site simulator was used, before 2024. From 2024 on, I read the “Track by CSS Operator” column as that record.
  • What was collected from other phones, and whether it was deleted.
  • What happened after the phone was located.

Oddities in the data
#

  • “Phone Off” is used as a status 23 times, and not after 2022.
  • Nine intervals come out negative, where a later step is dated before the order. They are blank in the redacted file.
  • One operation shows 366 days from order to located, which looks like a mistyped year. It is left as recorded.
  • Carrier is blank in 20 rows and phone owner type in 12.
  • 202 rows have notes, 74 of them from 2020.

Sources

A note on method. The redacted file and every count here were produced by a script run against the log as the department sent it. This piece was assembled with AI assistance. Any error is the author’s. Corrections are welcome at contact@peopleoverprivilege.com.

Capture and Statutes - This article is part of a series.
Part : This Article