City Hall Turns a Permit Backlog Over to Software, and the Council Wants the Log Files
An automated triage tool cut residential review times by a third. Nobody can yet say which applications it pushed to the back.
Port Meridian’s planning department has been sorting building permit applications with an automated triage tool since March, a fact that reached the city council on July 14 in a footnote to a quarterly performance report.
The council spent the following ten days finding out what the footnote meant.
The tool, licensed from Rampart Civic Systems for $214,000 a year, reads incoming residential and small-commercial applications and assigns each one a completeness score and a routing recommendation: express lane, standard queue, or flagged for a planner’s manual review. It does not approve or deny anything. What it does is decide the order in which human beings look at things, and in a department carrying a 2,900-application backlog, order is close to outcome.
By the department’s numbers, it works. Median review time for a residential addition fell from 41 days to 27. The express lane, which handles roof replacements, water heaters, and like-for-like window swaps, now closes 64 percent of its volume within four business days. Permit counter walk-in wait times are down by half.
Council member Ruth Adeyemi asked the question that produced the next three weeks of work: which applications got slower.
Nobody knew.
## No denominator
Planning director Hollis Ng told the council’s operations committee on July 20 that the department tracks median and mean turnaround by permit type, which is what the city’s performance dashboard requires, and does not currently retain the tool’s per-application scores after a permit closes.
“We kept the outcome and threw away the reasoning,” Ng said. “That is on me. It is a records decision I made without thinking of it as a policy decision.”
The distinction matters because the express lane’s gain has to come from somewhere. Applications the tool flags for manual review sit in a queue staffed by four senior planners. Committee staff pulled a hand sample of 300 flagged applications from April and May and found a median of 58 days — up from a pre-deployment baseline of roughly 44.
Flagging is not random. In the sample, applications submitted on paper at the counter were flagged at 3.4 times the rate of applications submitted through the online portal. Applications from addresses in the channel flats and the Vessey Road corridor — the two parts of the city with the oldest housing stock and the most nonconforming lots — were flagged at roughly twice the citywide rate.
Ng’s explanation is that this is largely legitimate. Old houses on odd lots genuinely raise more zoning questions. Paper applications genuinely arrive with more missing fields. “The tool is not wrong that these are harder,” she said. “The question the council is asking is whether harder should also mean slower, and that is not a question the vendor can answer for us.”
A flag is not a denial, but ninety extra days will read like one to a contractor with a crew on payroll.
## What the contract says, and does not
City attorney Bram Kessel’s July 21 memo found no procurement violation. The purchase fell $36,000 under the $250,000 threshold that triggers council approval, and the department routed it as an operational software renewal against an existing master agreement.
Kessel flagged three gaps. The contract contains no audit clause. It does not require Rampart to disclose the features driving a score. And its data provisions grant the vendor a license to use “de-identified application metadata” for product improvement, which Kessel wrote was probably harmless and definitely should have been negotiated.
Rampart’s director of public sector accounts, Sela Whitmore, said the company will provide feature weightings under a nondisclosure agreement and supports a written audit provision at renewal. Asked whether the model was trained on Port Meridian’s own permit history, she said it was fine-tuned on eleven jurisdictions including this one, and that historical approval timelines were among the training signals.
Adeyemi asked the obvious follow-up: if the model learned from a decade in which flats applications moved slowly, whether it had learned to keep them slow. Whitmore said the model does not predict approval and cannot encode that bias in the way the question implies. She also said she could not rule it out without the audit.
## The counter-argument, from the counter
The strongest defense of the tool at the July 20 hearing came from someone with no stake in the software.
Marisol Feen has run a six-person residential remodeling outfit on the ridge for nineteen years. She testified that she has spent most of her career on the losing end of the manual system and does not want it back.
“Before March, the order was whoever the planner liked, whoever called, whoever had a lawyer,” she said. “I have watched a permit sit ninety days and then move in an afternoon because somebody’s brother-in-law made a call. At least the machine is the same machine on Tuesday as it was on Monday. Audit it — please audit it. But do not tell me the thing you are replacing it with was fair, because I was standing at that counter.”
Adeyemi said afterward that Feen’s testimony changed the shape of her proposal. Her original draft would have suspended the tool pending an audit. The version she introduced instead does not.
## The ordinance
Adeyemi’s ordinance, co-sponsored by Glen Parry and introduced July 21, would require four things of any city system that scores, ranks, or routes a member of the public’s application:
Departments must file a one-page public registration before deployment, naming the vendor, the decision affected, and the human official accountable for outcomes. Per-application scores and the features behind them must be retained for four years and released on request to the applicant. Every such system must report quarterly on outcome disparities by submission method, council district, and permit type. And any contract must carry a right-to-audit clause survivable by an outside reviewer.
Parry, who has voted against most of Adeyemi’s recent proposals, signed on after the flagging numbers came out. “I do not care that it is software,” he said. “I care that we built a fast line and a slow line and did not write down who we put in each.”
The city budget office estimates compliance at $95,000 in year one, most of it storage and one analyst position. Mayor Teodora Sarn has not taken a position. The ordinance is set for first reading August 11.
Ng, for her part, is not fighting it. She has already ordered the scores retained going forward, effective July 15, and has told her staff that the sample of 300 will be a full audit by October whether the council passes anything or not. She was asked at the hearing what she would have done differently. “I would have asked what the slow lane looked like before I bragged about the fast one,” she said.
Reporting for this story was prepared for The Harbor Ledger’s politics desk. Tips:newsroom@theharborledger.com