The Score Des Moines Jail Enigma Inside The Metrics Management And Mysteries Of A Critical Facility

Understanding Facility Metrics When the Numbers Don't Add Up

Des Moines runs its jail operation under a scoring system that most people outside the department never see clearly. I spent about four years working the administrative side of inmate population tracking, and the first thing you learn is that the raw numbers tell one story while the actual floor reality tells another. The discrepancy isn't usually fraud or incompetence. It's structural. What we called the Score was essentially a composite index. It pulled from booking throughput, bed utilization rates, disciplinary incident counts, medical call responses, and staff-to-inmate ratios. City managers used it to determine funding allocations and staffing decisions. The problem is that some of those data points come from different systems that don't talk to each other, and the timestamps often don't align.

The Score Des Moines Jail Enigma Inside The Metrics Management And Mysteries Of A Critical Facility

The enigma part isn't mysterious at all once you've actually watched the evening shift changeover. At 6 PM, when the night unit takes over, there's a thirty-minute window where bookings from the previous shift are recorded but the new occupancy count hasn't been entered yet. During that gap, the Score drops artificially. If you're reviewing the data on a Tuesday afternoon and seeing a dip around 6 PM on a Monday, that's not a real decrease in population. It's a timing artifact. I learned this the hard way in 2019. My supervisor asked me to pull a report explaining why our compliance rating had dropped seventeen points between October 14 and October 15. I sat down with the raw data and noticed something nobody else caught. We'd switched to a new electronic health record system for medical calls on October 12. The old system logged timestamps with military time precision. The new one rounded to the nearest five minutes. Suddenly, every response time metric looked worse because the denominator had shifted. We weren't slower. We were just measured differently. You have to understand how the Score actually works before you can work with it. It's not one number. It's weighted. Bed utilization carries about thirty percent. Incident reports carry twenty. Staffing ratios carry fifteen. Booking processing speed carries fifteen. The remaining twenty is split between medical metrics, classification accuracy, and reoffense tracking. Each component has its own data source. Some are pulled automatically. Some require manual entry by shift supervisors.

Manual entry is where things get interesting. A lieutenant on the day shift might classify a disturbance as a "verbal altercation" while the same event on the night shift could get logged as "physical aggression." The difference matters for the Score. It matters for state reporting. It matters for budget conversations that happen six months later in rooms you weren't in. Here's a counter-intuitive thing most beginners miss. When the Score looks good, it often means the facility is under-reporting rather than over-performing. I've seen this pattern across three different county systems. Facilities with consistently perfect metrics tend to have higher turnover in their administrative staff. The people who know how to code incidents in the most favorable way stay longer. The people who flag problems get reassigned or leave. By the time you're reading the annual report, the numbers reflect institutional behavior, not operational reality. On the flip side, a facility with a fluctuating Score is often the one telling the truth. Someone is writing things down honestly, which means some incidents are being captured that would otherwise be invisible. The Score dips, the managers get uncomfortable, and then something actually gets fixed. That's the system working as designed, even if it looks like failure on paper.

I ran into a specific edge case during a state audit in early 2021. The auditor wanted to verify our average length of stay for pretrial detainees. The official metric showed 14.2 days. My calculation showed 18.7 days. Both were technically correct. The official number excluded anyone who had been transferred to another facility or released on electronic monitoring before their court date. My number included them. The auditor accepted my methodology, but the published Score still used the official calculation. This is the kind of thing that creates entirely different policy recommendations depending on which denominator you choose. The workaround I developed was simple but undocumented. I created a parallel tracking sheet that flagged three categories: standard Score entries, adjusted entries that reflected actual facility presence, and edge cases that required notes. When someone asked for the Score, I gave them the official number. When someone asked what was actually happening inside, I had the adjusted data ready. It took me about twenty extra minutes per week to maintain. Twenty minutes that probably saved me from several embarrassing moments in meetings. There's also the issue of weekend effects. Booking volume drops on Saturday and Sunday, but classification reviews don't. That means the system can flag an inmate as "under review" for up to seventy-two hours over a weekend. During that time, the inmate counts as reviewed in the monthly statistics but not in the dailyScore calculation. The gap creates seasonal distortion that shows up clearly if you plot the data month by month. September and October always look better than June and July, partly because of actual operational differences and partly because of how the scoring algorithm handles mid-year staffing changes.

I should be clear about what this system does not do well. The Score doesn't capture quality of care. An inmate could receive excellent medical attention and still register as a negative outcome if their case stays open too long. It doesn't measure staff morale or institutional climate. It doesn't account for external factors like changes in arrest patterns or court backlogs that push population up independently of facility management decisions. These blind spots are built into the design, not bugs that someone forgot to fix. If you're trying to improve your Score, the conventional advice is to focus on reducing incident reports and speeding up booking times. That approach usually backfires. When you actively suppress incident reporting, the Score looks better for a quarter and then you have a real problem that surfaces during an inspection. The faster you book people without proper classification, the more re-offenses you see in the ninety-day tracking window. The Score punishes you eventually, just with a delay. A more sustainable strategy involves fixing the data pipeline first. I found that spending one week cleaning up how incidents were coded across all three shifts produced better long-term Score improvement than any operational tweak. The coding standardization reduced our inter-shift variability by about forty percent. That alone moved the Score up six points without changing a single operational practice. The system rewarded accurate reporting, not hidden problems.

There are also configuration settings that most facilities don't adjust. The default threshold for what counts as a "missed medical call" is set at ten minutes. In a facility that's regularly holding eighty to one hundred percent capacity, that threshold creates false positives during shift changes. Lowering it to seven minutes during handoff periods and keeping it at ten otherwise eliminated about fifteen percent of our incident flags. The adjustment required a manual override in the administrative panel, which meant documenting the rationale. That documentation became useful during audits. For anyone trying to access or download the actual Score framework or calculation methodology, the Des Moines jail system doesn't publish it as a standalone document. The closest thing available is the annual performance report, which includes aggregate numbers but not the weightings or algorithms. If you work inside the system, you can request the scoring configuration from the compliance office. They typically provide it within five business days. External researchers have filed public records requests that have been granted in some cases and denied in others, depending on whether the data is considered part of an ongoing audit. The most reliable way to understand what the Score measures is to request quarterly trend data for each component separately. When you see them broken out, the relationships become clearer. Bed utilization and incident reports usually move in opposite directions. Staffing ratios and booking speed tend to correlate. Medical metrics are the least predictable because they depend on population health demographics that shift with arrests for different offense types.

I've found that the best time to review your Score data is mid-month, not end-of-month. End-of-month data gets compressed because everyone is rushing to close out reports. Mid-month gives you a complete week plus the tail end of the previous cycle. The patterns are more stable. You can spot anomalies before they become policy. One more practical note that probably won't help anyone outside the system but mattered to me. The Score calculator has a known issue with duplicate inmate entries during mass bookings. When twelve or more people are processed in the same hour, the system sometimes creates overlapping records that inflate population counts by two or three. The fix is to run a de-duplication query on the booking table before generating the monthly report. The query takes about three minutes and removes the duplicates without affecting any other metrics. Nobody on my team knew about this except for one veteran analyst who mentioned it casually during a training session. It wasn't in any of the manuals. If you need the raw data for analysis, the facility management portal allows exports in CSV format. The export includes all fields but requires you to select a date range. Maximum range is ninety days. There's no bulk export option for historical data older than three months unless you submit a formal request to the IT department, which can take two to three weeks. Plan ahead if you're doing longitudinal research.

The Score isn't a perfect reflection of facility operations. It never will be. It's a tool for resource allocation and accountability, not a diagnostic instrument. Understanding what it captures and what it misses is the difference between using it effectively and being used by it.