Part 6 of 6 in The Good Side of 311

Three of Philadelphia’s five big-volume 311 agencies solve their tickets at a pace residents actually appreciate — the Community Life Improvement Program, Parks & Recreation, and the Streets Department. Two of them don’t: License & Inspections, and the Police Department. Five named agencies in total, with enough closed tickets and enough equity-gap data to actually score on every feature.

I’d have guessed the agencies that outperform just have an easy job — or just handle the most tickets. The big agencies have figured out how to run 311 at scale, the small ones haven’t. That is the conventional wisdom, and it has it the other way around. Volume doesn’t separate the agencies that solve tickets at a good pace from the other two. Four specific features do — and they are not the ones I would have predicted.

The five agencies

Philadelphia’s 311 system routes work to more agencies than five. The cohort in this article is the five that have enough closed-ticket volume and enough zip-code resolution to compute an income-equity gap on every one of them — CLIP at 40,043 closed tickets, Parks & Recreation at 13,065, the Streets Department at 293,532, License & Inspections at 78,101, and the Police Department at 46,435. The Fire Department and the Water Department do not have enough zip-level volume on their 311 data to compute an income-equity gap, so I can’t include them apples-to-apples. Five agencies, then.

Of those five, three cluster at the top of every distinguishing feature and two cluster at the bottom. CLIP, Parks & Recreation, and the Streets Department are the three agencies solving tickets at a good pace. (Earlier articles in this series looked at CLIP, Parks, and Streets individually. This article is the one that puts them on the same scale.)

What the three agencies share

I measured five things for each agency and averaged within each group. On four of those five things, the three agencies that work well clearly come out ahead. On the fifth — volume — neither group has a clear edge.

• Days taken vs days promised — the three agencies deliver about three-quarters of their promised time on average; the other two take almost twice as long as they promise. For every kind of work the city publishes a promised close time — say 30 days. The three agencies that work well actually close those tickets in about three-quarters of that time: a 30-day promise that closes in 22 days. The other two agencies take almost twice as long as they promise: a 30-day promise that closes in 58 days. The other two are overpromising and underdelivering. The working agencies are honest about their capacity, or slightly conservative.
• Share of tickets closed inside their promised time — the three agencies close 75.0% on time, the other two close 41.5%. CLIP meets its own promises nine times out of ten; Parks meets its promises about three-quarters of the time. L&I meets its promises four times out of ten. The gap between the two groups is wide.
• Number of specialized workflows they run — the three agencies run 7.3 such workflows on average; the other two run 1.0. You might assume the agencies that do well handle fewer types of work — switching between many different things usually slows you down. But that’s not what this number measures. It counts how many kinds of work each agency has carved out as its own dedicated operation, each with its own crew and promised time. Working agencies run more of them. Streets runs 19: Street Lights, Illegal Dumping, Potholes, Snow Shoveling, and more. Parks runs 2: Street Trees separately from Parks Safety. CLIP runs 1: Graffiti Removal as a dedicated operation. The other two average 1.0 — and the worst case is L&I, with zero. The reason isn’t that L&I’s work is split across many small pieces; it’s that almost every one of L&I’s 78,101 closed tickets isn’t filed under any formal work category. The city hasn’t built per-type promised times or specialized crews for L&I’s work at all. Police has 2 (Abandoned Vehicle, Police Complaint). Specialization is what lets an agency meet its promises — but only when the city has carved out the categories to specialize around.
• Same speed across rich and poor neighborhoods — the three agencies run a 4.2-day gap between rich and poor zips; the other two run a 34.7-day gap. This is the typical wait-time difference between a lower-income zip code and a higher-income zip code on the same kind of ticket. Parks’s gap is essentially zero (slightly favoring lower-income zips); Streets is 0.8 days; CLIP is 11.85 days. The other two agencies run 28- to 41-day gaps — meaning a lower-income zip waits weeks longer than a higher-income zip on the same ticket. Working agencies treat every neighborhood the same way.

What the four features mean in plain English

The four features are not a list. They fit together as a chain. An agency that takes roughly as long as it promises then hits those promises, because it has carved out the kinds of work it does — each with its own crew and promised time — and it treats every neighborhood the same way. Each step makes the next one more achievable. An agency that runs everything as a catch-all (or hasn’t built per-type crews and promised times at all), with a published close time it never hits, and that closes tickets faster in richer zip codes than in poorer ones — that agency is below the bar on every feature at once, because the features are not independent. They are the same problem seen from four angles.

The reverse is also true. The features the three agencies share are concrete enough to act on. They are choices an agency head can make: rewrite the published close times to match the operation, carve out each kind of work the agency does (with its own crew and promised time), staff the operation to meet the published close time, and run the same crew on the same promised time in every zip code. None of those is a mystery. The agencies that do well made those choices. The ones that don’t make them are the ones below the bar.

What this is and is not

The feature matrix is descriptive, not causal. I have shown that the three agencies share four structural features; I have not shown that those features cause the upper-group outcome, and I have not ruled out confounders (citywide management turnover, budget cycles, agency-head tenure). The other two are not necessarily badly managed; the data does not measure management. The five-agency window is also small — including Fire and Water once their equity gaps can be measured would tighten the picture. The headline finding — four features cleanly separate the three from the other two, and volume is not one of them — is robust to those caveats. The specific gap on each feature is not.

Where this leaves the series

This is the sixth and last article in a series I started after the one I wrote about the parts of 311 that aren’t working. Across the five articles before this one, the data kept surfacing the same shape: there are parts of 311 where the city genuinely delivers. Four of the thirty service codes close twice as fast as the city promises — Street Lights in 7.9 days instead of 90, Hydrants in 3.1 days instead of 30, Street Trees in 1.8 days instead of 15, Complaints in 3.9 days instead of 30. CLIP meets its own promises 91.3% of the time on 40,043 tickets, the highest of any big agency in the city. Parks runs two operations that have almost nothing in common — Park Safety, gradual and reliable, and Street Trees, fast on six in ten of them and brutal on the rest. Streets handles 293,532 tickets, and 76.6% of them meet what a normal Philadelphian would consider acceptable. And this article names the four structural features that distinguish the agencies that do them well from the ones that don’t — tight promises, met-rates, specialized workflows, equitable wait times.

Over the course of my six months of research into the Philly311 system — the first series that highlighted inequities and failures, and this one, which highlights the strengths of some agencies and service types — I’ve tried to explore what is empirically happening, and where possible, why it’s happening. I want to pause for a second at the capstone of "The Good Side of 311" to note an observation: this article named an observable set of habits that the high-performing agencies share, and that can likely be transplanted to other agencies to improve performance.

In fact, I would encourage any agency leader, commissioner, or elected leader to observe these habits and adopt them in the other agencies. I believe they will improve the quality of service for residents and, equally importantly, improve the satisfaction of city employees for having done the work well.

Notes, Sources, and Methodology

All data is the OpenDataPhilly 311 Service Requests table public_cases_fc. The cohort is the closed-with-SLA universe filed 2024-01-01 through mid-2026. Agency-level equity-gap data is sourced from the 311-agency-gap table maintained for the income-gap article in this series. Five agencies qualify (CLIP, Parks and Recreation, Streets Department, Police Department, License and Inspections); two are excluded because they do not report enough zip-level volume to compute an equity gap (Fire Department, Water Department).

Methodology. For each of the five qualifying agencies I compute five features on the closed-with-SLA cohort: SLA tightness (ratio of published sla_days to median actual close, per code, aggregated as the volume-weighted mean), mean met-rate (volume-weighted), number of service codes with at least 100 closed tickets, volume share of the closed-with-SLA universe, and equity gap (median close-time difference between lower-income and higher-income zip codes, in calendar days). Each feature is converted to a percentile rank across the five agencies, with lower-is-better features inverted so higher percentile = better outcome. Agencies are grouped into the upper half (above median on most distinguishing features) and the lower half (below median). The falsification check confirms that four features — SLA tightness, mean met-rate, number of service codes, and equity gap — cleanly separate the two groups, and that volume share does not.

Limitations. The feature matrix is descriptive, not causal; confounders (citywide management turnover, budget cycles, agency-head tenure) are not measured. The five-agency window is small; including Fire and Water once their equity gaps can be measured would tighten the picture.