What a Staffing Study Can and Cannot Tell You

Someone above you has asked for a number. How many officers do we need. There is a defensible way to produce it, and it is not the one most cities reach for.

• The three benchmarks in common use — officers per thousand residents, peer-agency comparison, and your own authorized strength — are not staffing analysis. Federal guidance says so directly.
• A workload study is the defensible method, and it is arithmetic you can follow line by line.
• The study hands you a target. Whether you ever reach it depends on hiring and retention numbers the study does not contain.

Three numbers that are not analysis

The Department of Justice’s community policing office published a plain assessment of staffing benchmarks in August 2024. It names three methods agencies lean on and explains why each one fails.¹

Officers per thousand residents. The appeal is that it takes ten seconds. In the authors’ words, “while the per capita method is easy to calculate and apply, it does not account for how officers spend their time nor for community conditions and expectations.” A ratio cannot tell you where to put a shift.

Peer comparison. The trouble is upstream of the math: two cities of the same size are rarely the same workload. Call volume, geography, event load, how much the agency self-initiates, what the community expects answered in person — all of it moves independently of population.

Your own authorized strength. This is the one that gets quoted in budget hearings, and it is the weakest of the three, because it usually descends from the other two. The DOJ assessment is blunt: “it can become a false ruler for gauging staffing needs and deficiencies.” An authorized number set in 2011 and carried forward is a historical artifact, not a requirement.

What the same guidance recommends is narrow and unambiguous: “a workload-based analysis is the only approach to systematically analyze and determine staffing needs based on demand.”

What a workload study actually computes

Strip the consulting language and a patrol staffing study is four inputs and one constraint.

1. Demand. A year of calls for service out of CAD, separated into citizen-initiated and officer-initiated work, then spread across hours of the day and days of the week.
2. Time per call. How long a response actually occupies an officer, including the units that back it and the report that follows.
3. Availability. How much of a scheduled officer’s year is actually on a beat, after leave, training, court, light duty and sick time.
4. A ceiling on committed time. How much of the shift you are willing to spend answering calls before there is nothing left for anything else.

That last one is where judgement enters, and the most-used model is explicit about it. The ICMA Center for Public Safety Management’s Rule of 60 sets two thresholds: roughly 60% of sworn officers assigned to patrol, and patrol workload held to no more than 60% of available patrol time — leaving 40% of the shift uncommitted for proactive work, investigation and the next emergency. The same model assumes about 25% of assigned officers are unavailable on any given day.²

Here is what those inputs look like filled in. A consultant’s staffing plan for one large East Coast city, published by the department itself, ran the arithmetic on 290,943 calls for service in a single year at an average of 45 minutes per response, against a target of 60% of time on calls and 40% on proactive work, and a shift relief factor of 2.3 — meaning 2.3 officers must be assigned for one position to be covered around the clock. It concluded with a specific recommendation: 805 police officers, 115 sergeants and 27 lieutenants for patrol.³

Some agencies now call this performance-based staffing: allocating against measured workload and service outcomes rather than against headcount history. The name matters less than the discipline. Every number above came out of the agency’s own systems.

A study is a snapshot, and it expires

Accreditation standards treat workload assessment as recurring work, not a one-off. The staffing plan cited above points to the accreditation standard requiring a documented workload assessment at least every three years;³ one university police department’s published allocation policy sets its own cycle at “at least once every four years…for each organizational component,” and notes leaders may run them more often when the budget demands it.⁴

Three or four years is a long time in call volume. Annexation, a new entertainment district, a diversion program that moves mental-health calls off patrol, a corridor that starts generating crashes — any one of those invalidates part of the model. The study is accurate on the day it is delivered. What it describes keeps moving.

The number the study does not give you

A workload study tells you the agency needs, say, 805 patrol officers. It does not tell you whether 805 is reachable. That answer lives in the hiring and retention data, and it is usually worse than the staffing conversation assumes.

One state’s standards and training commission surveys every agency it certifies. Across 180 agencies employing 6,013 certified officers, the agencies that ran a recruitment in 2024 reported this:⁵

Agencies in the same report estimated that, during probation, around 9% of new hires leave voluntarily and 8% are terminated.⁵ Hired is not staffed. Nothing in a workload model accounts for that.

The gap between authorized and available is where the plan dies. A mid-Atlantic county’s legislative oversight office reported its police department carrying a 15% vacancy rate, with patrol operating at “only 74% of its staffing level.”⁶ A large Southwestern city’s auditor found the department’s vacancy rate had more than doubled to 18% between 2020 and 2024, with over 41% of vacant positions open more than a year, and a hiring process that “could take between 6 to 12 months to complete.”⁷

And this is not a market problem any more. In a survey of 214 agencies employing 151,335 officers, 2023 hiring was “nearly 30 percent more” than 2020 — while January 2024 sworn staffing was still 4.9% below January 2020, with resignations running 28.6% above 2019.⁸ Agencies are hiring harder and staying short. The constraint moved.

Two numbers, not one

Commission the study. Then put a second set of numbers beside it — the ones that decide whether the first set is reachable.

Five things to run yourself, none of which require a consultant:

1. Compute your own relief factor. Total scheduled hours minus every hour of leave, training, court and light duty, divided by the hours a post needs covered. Do not accept a default. It is the single assumption that moves the answer most.
2. Separate withdrawals from rejections in your hiring data. Candidates you turned down are a standards outcome. Candidates who left are a process outcome, and only one of those is in your control.
3. Time-stamp every stage of the hiring process, automatically. If a recruiter types the dates in by hand, you will not have them when you need them.
4. Track field training completion as a staffing number, not a training number. A hire who does not finish is a vacancy that already cost you a salary.
5. Forecast attrition against the study’s target, not against last year’s budget. If projected separations exceed projected graduates, the gap widens no matter what the study recommended.

If you are about to commission one

Four things to put in the scope, each of which costs nothing to ask for and is painful to add later.

• Ask for the model, not just the report. A PDF is obsolete in a year. A spreadsheet you can re-run with next year’s call data is a tool you own.
• Require your own relief factor, calculated from your leave records, with the assumptions written down.
• Require officer-initiated activity to be counted, not just dispatched calls. Leave it out and you will understate workload and overstate available time.
• Ask what the model assumes about vacancy. Most assume full staffing. If yours does, the recommendation describes a department you do not have.