Defining the prevention problem and work boundaries · Lesson 3

Where and when does harm concentrate?

8 min read

Lesson objectives

  • Apply Where and when does harm concentrate? to a fictional community case, documenting the decision and its limits in a reviewable work product.
Reading tools

A station baseline may be arithmetically sound yet hide where action is possible. A station-wide ratio cannot tell a place manager whether to examine storage design at the east or west entrance. Start a matrix with defined places as rows and defined time periods as columns. Place each event in one cell under an explicit rule. Choose boundaries the organization can act on, such as the east and west parking entrances, rather than a vague area called “near the station.” Give events without a reliable location an “unknown place” cell instead of assigning them to whichever site seems likely.

Consider a fictional four-week file. The east entrance has nine losses among 900 parking episodes; the west has six among 200. East has the larger raw count. Per 100 episodes, however, east has 9/900×100=1 and west has 6/200×100=3. If the immediate decision concerns loss exposure per parking episode, west warrants an early investigation despite fewer absolute events. Do not hide east's nine events: absolute harm still matters for resource planning, and intervention costs may differ. State the priority rule before calling either location a “hot spot.” The term could mean the most events, highest exposure-adjusted measure, or greatest harm severity; those are different decisions.

Partition time with the same care. Suppose five of west's six losses are assigned to 18:00–22:00 and only 50 of its 200 parking episodes occur then. The evening measure is 5/50×100=10 per 100 episodes. The rest of the day has 1/150×100, about 0.67. Such a large descriptive difference calls for a data-quality check. Did the log capture when a bicycle was lost or when the owner discovered the loss? Someone may park at midday and discover the bicycle missing in the evening. If the event time is unknown, label the field “discovery time” and do not present it as a proven time of occurrence.

A dark patch on a map can reflect heavier use, easier reporting, or a reporting desk nearby. Mapping organizes questions; it does not itself identify a cause. Inspect accessible routes, lighting, sight lines, the way bicycles can be secured, responsibility for the place, and service hours. Ask which conditions the facility can lawfully and practically change. Do not publish the precise location of a sensitive account that could identify someone at risk. Do not turn a cluster in a neighborhood into a claim about its residents. The unit of analysis is the opportunity and place context, not a user's origin or identity.

The problem analysis triangle, introduced later, connects place, target, likely offender, and existing controls; geographic concentration alone does not identify an offender. For now, write a testable hypothesis: “The higher west-evening measure may relate to reduced oversight of storage when the service desk closes.” Test this with authorized operational observation or a desk schedule alongside deidentified event data. Do not begin by filming users or requesting victims' names. If event timing is unknown for most entries, rewrite the hypothesis as a data need instead of treating uncertainty as a finding.

For every matrix cell, keep the event count, exposure denominator, method for assigning time and place, and proportion of records with missing fields. Then run a sensitivity check. If one west event with uncertain time moves from evening to daytime, evening becomes 4/50×100=8 and daytime becomes 2/150×100≈1.33. The ranking still points to the evening in this invented example, but the size of the gap changes. This check does not remove every bias. It reveals whether a decision hangs on one uncertain record. The specialist's output is a clear concentration table, stated limitations, and a small field question the team can investigate without stigma or privacy intrusion.

Add a layer showing how much is known about each location. A large map point can reflect more events, easier reporting or several places grouped under one address. Keep location quality separate from event volume. Record whether an address identifies one rack, a whole entrance or a broad area. The feasible prevention decision depends on that resolution.

Suppose ten events use the address “station,” while three identify the western rack. Dividing the ten equally between entrances creates false precision. Show a broad area with uncertain internal location and retain the three at their known position. Future reports can request better information safely, for example a choice from a list of general locations rather than a person's detailed travel route.

Check time boundaries in the same way. “Loss discovered at eight” differs from “event witnessed at eight.” The possible event window can extend from the last confirmed sighting to discovery. Preserve its start and end where available and display it as an interval. That distinction directs a site visit toward the period when belongings were exposed rather than only the time a telephone call arrived.

A cross-tabulation can reveal patterns hidden by totals. Divide records by entrance and broad time period, adding the available exposure for each cell. Watch small cells: one additional event can reverse their ranking. Show the cell count beside the measure and observe a suitable follow-up period before describing a stable pattern. Choose the detail according to participant safety and practical action.

Finish with a field question that can be checked. Replace “Which neighborhood is worst?” with “Can the closing employee see the western rack during handover?” This links a location to a changeable condition within a defined responsibility. Return to the map after the visit and add observations while distinguishing them from interpretation. The map then directs work rather than assigning judgments to residents.

Lesson summary

Someone proposes new lighting at the east entrance because it has nine losses versus west's six.