MICHAELAnalytical Workstation Request a free trial

MICHAEL · Temporal analysis

Agency data gives you a window. MICHAEL counts it honestly.

A burglary reported Monday morning could have happened any time since Friday night, and most tools count it at the moment the report was typed. MICHAEL shows the week as 168 cells, counts a windowed incident nine honest ways, and puts every record, every call and every case event on one timeline you can zoom from months to minutes.

Case timeline / Sample call records79 CONTACTS · 5 LANES
The Case timeline window on the sample call records: five phone-number lanes over one week, calls drawn as arcs between lanes and texts as dotted lines, two case events marked as vertical lines labelled Burglary and Property pawned, a panel reading 79 contacts on these lanes, and the whole-span strip along the bottom.
79 contacts, and the arcs that stopThe sample call records that ship with the app: 79 contacts on five phone lanes over one week, each call an arc between the two numbers and each text a dotted line, with the burglary and the pawned property drawn as case events from a second loaded dataset. The pink number's arcs stop at the burglary line.

See the full analytical workstation

01 The week

Day by hour, across the whole week.

When the series actually operates is a question about 168 cells, not about a bar chart of hours. The activity heatmap answers it, and every other panel in the view reads the same records through the same filter.

  • 01168 cells. Weekday by hour, with the counts on a color ramp and the legend saying what the ends mean. Copy, Save and Add to briefing, like every chart in MICHAEL.
  • 02The clock and the calendar. Hour of day around a 24-hour dial, counts by month, by weekday and by hour, and hour of day drawn as one line per weekday so a Friday that behaves like a Saturday stands out.
  • 03The series over time. Count, sum or average by hour, day, week, month, quarter or year, as a line, bars or an area, on the date column MICHAEL detects or the one you choose.
  • 04Records with no time are counted where they belong. The three hour-of-day panels leave such records out and say how many they left out. Day of week, month and calendar views still count them, because those are known.
One filter, every viewTurn off weekends here, or set a time of day or a date range, and the map, the statistics, the network, narrative search and the reports all recalculate on the same records. There is no second filter to forget.
Temporal / Activity168 CELLS · DAY × HOUR
The Temporal view on the Los Angeles public sample: a weekday by hour heatmap of 168 cells with a legend running from 327 to 1,453 per cell, a radial 24-hour clock beneath it, and a bar chart of counts by month peaking in August.
150,000 incidents, 168 cellsThe Los Angeles public sample that ships with the app, 150,000 incidents: weekday by hour, the radial clock, and the month chart. Friday and Saturday afternoons run about 1,450 incidents a cell against about 330 in the quietest hours before dawn.
02 Time Span Analysis

Nine ways to count an incident that took eight hours.

An earliest and a latest time is what the report actually gives you. Time Span Analysis works with that window instead of pretending the report time is the crime time, and tells you which method made the chart.

  • WWeighted. Splits 1/N across every hour the window touches. An eight-hour burglary contributes 0.125 to each of its eight hours.
  • EEqual opportunity. Counts a full incident in every hour it touched, so the chart shows every hour the crime could have happened in.
  • AAoristic. Spreads each incident across its window weighted by the minute, over the 168 weekday hours, with a heat chart and a CSV. An end date with no time reads as the end of that day, so "from 22:00, discovered that day" is spread over 22:00 to midnight.
  • MMidpoint. The middle of each window, reported with the mean, the standard deviation and the 1σ and 2σ bands.
  • +5Exact start time, duration distribution, overlap heatmap, Gantt timeline and temporal coverage, which finds the genuinely quiet windows rather than the apparently quiet ones.
  • 06Two of the nine run on a single time. Most records systems export when a crime happened, not a window. Exact and Aoristic run on that one time; the seven that need an earliest and a latest are shown with the reason they are unavailable rather than hidden, so you know what the export is missing.
  • 07Every method explains itself. A one-line description of the method you are looking at, and a comparison of all nine saying which need a window. The chart you put in a brief names its method.
03 The timeline

Every record on one axis, from months down to minutes.

Timeline in Temporal, and Case timeline from the Network view's link table, opens a window with one lane per suspect, vehicle, beat or phone and every record placed at its moment.

  • 01Months to minutes. Scroll to zoom, drag to pan, and drag on the strip along the bottom to jump anywhere in the span. The whole window works from the keyboard, and a screen reader hears each change.
  • 02Call records draw themselves. The lanes start as the record holders and their common contacts. A call between two lanes is an arc, a call with anyone else a tick, and a text a dotted line.
  • 03Case events come straight from your own data. Choose a loaded dataset, the case's incident reports say, and one value of one column if you like, and each record is a line at its date. A timeline cannot drift from the records behind it.
  • 04Every mark is a record. Click one to select it and open it. Show these rows hands what is in view to every other view, and Save image and Add to briefing take the picture with you.
Case timeline / Zoomed to ten days50 CONTACTS IN VIEW
The same Case timeline zoomed to the first ten days of August: five lanes, ticks for calls and texts with numbers outside the lanes, two dotted text lines and one arc between the holders, the whole-span strip along the bottom with the viewing window dragged to the left end, and a panel reading 50 contacts on these lanes with the records in view listed newest first.
The same records, ten days wideThe same records, zoomed to the first ten days: the viewing window on the strip along the bottom has been dragged to the start of the span, and the panel on the right lists every contact in view, newest first. Scroll in again and the axis reads in minutes.
Timeline, Pair, Pattern of lifeSAMPLE CALL RECORDS · NO SOUND
04 Pair

Two numbers, and the silence after the last call.

Pair shows every contact between two numbers as a beam aimed at whoever was contacted, as wide as the call was long. The silence after the last one is usually the thing worth explaining.

  • 01Rate, first and last. Contacts per day above the beams, the first contact marked, who contacted whom, calls against texts, talk time, and the hour of day they talk.
  • 02Each case event is scored against the pair's own rate. How many contacts fell in the 48 hours before the event and how many after, as a multiple of what this pair does on an ordinary day.
  • 03Open it from anywhere a number appears. From a number's links in Network, or with the arrows beside a lane in the timeline.
Pair timeline9 CONTACTS · 7 IN THE 48 H BEFORE
The Pair timeline between two sample numbers: nine beams between two lanes, each as wide as the call was long, a per-day bar above them rising toward the burglary line, a shaded 48 hours before the event, and a panel reading 9 contacts, 39 minutes of talk time, 8 calls and 1 text, and 7 of 9 in the 48 hours before the burglary at 3.5 times this pair's daily rate, 0 after it.
Seven of nine in the 48 hours beforeTwo sample numbers: nine contacts and 39 minutes of talk between August 29 and September 6, seven of them in the 48 hours before the burglary, 3.5 times this pair's daily rate. None after it.
05 Pattern of life

One number, round the clock.

One number, or one suspect or vehicle: a 24-hour orbit where the angle is the time of day and each ring a weekday, every contact a dot in its own color for made and received, the contacts per hour around the edge, and the longest stretch of hours with nothing shaded.

  • 01Beside the orbit. Weekday by hour, every day of the span with the case events marked (it says when a number went quiet after one), and who it talks to most, each opening the pair timeline.
  • 02Cell sites, only when they are its own. A cell site in a call record belongs to the record holder's phone. For any other number MICHAEL says so, instead of listing towers that were never its own.
  • 03A record holder's route on the globe. With a tower list loaded, its own calls in time order, each run of calls at one tower a stop, drawn as a line through the stops with a point at every tower. Calls at a site the list does not have are counted and named, never placed.
Pattern of life36 CONTACTS · NOTHING AFTER THE BURGLARY
The Pattern of life window for one sample number: a 24-hour orbit with a ring per weekday and every contact as a dot, 36 contacts in the centre, a weekday by hour grid, a by-day chart reading active August 29 to September 6 with nothing after the burglary, 17 made and 19 received, the three numbers it talks to most, and a cell sites note saying the tower column belongs to the record holder.
36 contacts, nothing after the burglaryOne sample number: 36 contacts, 17 made and 19 received, active from August 29 to September 6 and nothing after the burglary. The cell sites panel says plainly that the tower column in these records belongs to the record holder, so it says nothing about where this phone was.
06 Built to be trusted

It places records in time.

Every chart on this page is a statement about records. The panels say so where it matters, and the brief you build from them should too.

  • 01A tower marks where the phone connected. The route on the globe says so, the cell sites panel refuses to list towers for a number that is not the record holder, and nothing here draws a person's location.
  • 02Every chart names its counting method. Seven of the nine Time Span methods exist because the report time is not the crime time. The chart names the method it used, and so should the brief, because two methods on the same records can put the peak in different hours.
  • 03A date with no time stays unknown. It is left out of every hour-of-day panel and counted. A chart that put those records at midnight would show a peak that nobody committed.
  • 04A gap in contact is a question for the detective. The pair timeline marks the gap after the last contact and scores the 48 hours before an event. What either means is for the detective, and the timeline does not guess for them.
Where to go nextTemporal analysis on the full MICHAEL page, the link analysis page for the network these phones belong to, crime series detection for the next-event forecast, or pricing. Questions go to info@sworn.ai or (801) 410-0123. A person answers both.
Request a trial

Try it on your own data.

Fourteen days, every feature, on one machine. Tell us who you are and we will email you your own download link. Nothing to activate. Your data stays on your machine and stays yours when the trial ends. No newsletter, no drip sequence.

Built for the people who do the work Sworn agencies, crime analysis units, public safety partners, and corporate investigation teams. Workstations are $2,500 a year, everything included. Academia and nonprofits: ask about license pricing.

Request a free trial

Or email info@sworn.ai. A person answers.