PFML 101 · Kevin Middleton

The setup

A state paid-leave system, one team, a fixed runway.

The exercise itself is Nava's, so it is not reproduced here. What matters for everything that follows is the shape of it: a state agency wants people paid faster, wants a claim to cost less to process, and wants to hold the line on operating cost. Those three pull against each other.

So the work is not picking a feature. It is working out which problem is load-bearing, and what has to be true for the other two to follow.

What's actually true

It is not an eligibility system. It is a document-collection system.

Every instinct says a benefits program decides who qualifies. The data says otherwise. In every state that publishes its denial reasons, most denials are paperwork and timing failures, not people who were ineligible.

78%

of Washington's denials are hours-documentation, missing event documentation, timing errors, or ID problems. Only 2% are "non-qualifying event."

71%

of Connecticut's denials were simply "did not file requested documentation."

34%

of Massachusetts' 35,964 denials were documentation problems — either nothing submitted, or documents that didn't comply.

The reframe that should drive your MVP. These are not people the program decided to turn away. They are people the program failed to collect a form from. That is a product problem, and it is the one a single scrum team can actually move in six months.

Look back at the brief with this in hand. Two of the five bullets they gave you are this exact problem, stated in the agency's own language: applicants "have no way to correct information which can lead to processing delays," and case workers "must reconcile conflicting information from applicants and employers." Those are the same defect seen from two sides of the desk. The agency has already told you where the bleeding is — they just didn't rank it.

Sources: WA ESD FY25 report · MA DFML FY25 · CT FY25 coverage

The journey

Five gates. Any one of them ends your claim.

Nearly identical across states. Notice how many depend on someone who is not the applicant.

Notify your employer — before you can even apply

Massachusetts requires 30 days notice for a planned event. Miss it and leave "might be delayed or denied." 335 MA denials in FY25 were for this alone.

File inside a narrow window

MA: no earlier than 60 days before, no later than 90 days after. CA: no earlier than day one, no later than day 41. In Washington, applying too early is its own denial category — 11% of all denials.

Prove identity

MA requires an unexpired color copy, front and back. Under 4.5MB. And .heic — the iPhone default — is rejected. Color faxes aren't accepted either.

Get a doctor to complete a certification

The hardest gate. MA explicitly rejects MyChart screenshots, discharge papers, prescriptions, test results, and doctor's notes. And in California, "physicians cannot be required to certify a claim." The state has no leverage over the one person who controls the gate.

Wait for your employer to respond

MA gives them 10 business days. Minnesota blocks all processing for up to 7 days. Oregon budgets 10 of its 29 average days to employer review.

Do the arithmetic. Massachusetts' own published happy path is 5 business days + 10 business days + 14 calendar days + a 7-day unpaid waiting period. That is five to six weeks before the first dollar — if nothing goes wrong, and on top of however long the doctor took.

Sources: MA approval timeline · MA required documents · CA EDD claim process · CA EDD physicians FAQ · WA ESD FY25 · MN after you apply · KATU on Oregon's 29 days · 458 CMR 2.08

The pattern

The harder the paperwork, the higher the denial rate. Every time.

Two natural experiments prove the failure is documentary, not substantive.

Washington, by leave type, FY25

Leave type	Approved	What you must prove
Family Bonding	95%	A birth certificate
Medical Postnatal	93%	Recent, documented event
Medical (own condition)	85%	Clinical judgment
Family Care	85%	Someone else's condition
Medical Prenatal	72%	Ongoing clinical judgment
Family Military	61%	Third-party military docs

Rhode Island runs the cleanest experiment. Same agency, same month, two programs. Disability claims: 16% denied. Caregiver claims: 41.6% denied. The only difference is that caregiving requires a certification about another person's health — a document you have less power to obtain.

And the one that should make you angry. 3,578 Massachusetts applicants were denied because the state could not find their employer in its own tax filing data. That is a state data-integration failure, charged to the resident.

Sources: WA ESD FY25 report, Figs. 7–8 · RI DLT monthly update · MA DFML FY25 annual report
Note: the RI figures are from a 2016 monthly update, the most recent RI split found. Treat as historical. Cross-state denial rates are directional — states count "filed," "approved," and "denied" differently.

What nobody measures

People who never finish don't show up in your metrics.

5,597

Massachusetts residents who started an application on the website and never submitted it in FY25. They are counted as denials, under "did not complete."

Massachusetts is the only state that publishes this. Everywhere else, the person who opens the form, hits the document-upload wall, and closes the tab is simply invisible.

Nobody publishes a correction rate either. No state reports what percentage of applications need a fix-and-resubmit round. Minnesota gives you 14 days to answer a Request for Information and says your application "cannot move forward" until then. Massachusetts gives you 90 days. But neither reports how often it happens, or how many people give up mid-loop.

If you propose instrumenting one thing on day one, make it this. You cannot manage a funnel whose largest leak is unmeasured.

Sources: MA DFML FY25 · MN RFI process · NJ DOL 2024 report — NJ is statutorily required to report insufficient-information claims and states the data "were not available for calendar year 2024."

The doom loop

Status opacity is a demand generator for your most expensive channel.

Washington states the causal chain plainly in its own legislative report:

"When customers receive claim decisions faster, they are less likely to call to check on their claim, thereby reducing call volume." Washington ESD, FY2025 Annual PFML Report

Then watch what happens when a state doesn't break the loop.

52 min

Oregon's average call wait in late 2023 — up from the 7 seconds officials touted at launch.

50%

of calls to Washington never entered a queue at all in FY25. They got a "high call volume" message and were disconnected. Median hold for those who got through: over 30 minutes.

Washington admits the trade explicitly: "Paid Leave reprioritized staff who typically answer phones and return calls to focus on processing claim applications." They robbed the phones to pay the queue. Both got worse.

Oregon's is worse than a trade — it's a design defect. The state mailed paper identity-verification letters, then denied people for not responding, while messages in its own online portal went unassigned. One couple's son was born prematurely and airlifted; they came home to find verification letters and denials for not answering mail they weren't there to receive. 1,500 people were stuck in ID verification after officials had said fewer than 100 were affected.

Minnesota's counter-pattern is one sentence long, and it's the best product decision in this entire brief: "The information in your account is the most up-to-date and is the same information our Contact Center can see if you call." That sentence answers "should I call?" before the thought forms.

Sources: WA ESD FY25 · WA JLARC oversight report · Governing / Oregonian on Oregon · KGW on Oregon benefits · MN Paid Leave · WA status states

The tell

When half of appeals get overturned, the appeal is your intake process.

51%

of decided New York PFL arbitrations went for the employee in Q1 2025 (18 employee / 17 carrier).

38%

of those disputes were categorized "PFL claim incomplete" — the single largest reason.

For comparison, the national unemployment-insurance benchmark is a 28.7% reversal rate at the lower appeal authority. New York PFL is running nearly double that.

What that actually means. In a well-calibrated system, appeals mostly affirm. When half of contested denials flip, and the top dispute category is "incomplete," the appeal is not correcting judgment errors. It is performing the document collection that intake failed to perform — months later, with an arbitrator in the room.

And note the window: Massachusetts gives you 10 calendar days from receipt of a mailed notice to appeal. For someone medically incapacitated or caring for a newborn.

Sources: NY WCB arbitration report, Q1 2025 · NY WCB Q2 2023 · US DOL ETA appeal reversal rates · MA appeals process · WA appeals · Alight on CO FAMLI changes
Note: New York is the only state found that publishes routine appeal-outcome data, because its arbitration is contracted out and reported quarterly. Do not assume NY's overturn rate generalizes.

The physics

Three inputs the state does not control.

1 · The healthcare provider certification

Top denial reason in New Jersey TDI (26.2%). Second in Washington (24%). The cause of Rhode Island's caregiver gap. The state has no authority over the person who must sign.

Best answer found: Colorado's provider portal. Providers register and verify their license, get notified in-app when a patient applies, complete the form in-portal, and can delegate to staff. Colorado's own reasoning: "reduce the risk of human error and reduce the amount of patient paperwork during stressful life events." Note it is a portal, not an EHR integration — nobody has cracked true EHR/FHIR yet.

2 · The employer response

Massachusetts made one decision that matters more than any other: the 10-day window lapses rather than blocks. Adjudication proceeds either way. Their actual average employer response is 2.69 days, and their median determination is 9 days. Minnesota, by contrast, blocks all processing for up to 7 days regardless.

3 · Wage data

Massachusetts reuses the Department of Revenue contribution feed rather than asking anyone to re-report. Nava validated that integration in pilot one, before anything else — 100% of records delivered with zero errors. Minnesota did the same thing with its unemployment-insurance wage data and the UI statutory definition of "wages," specifically to avoid new employer data collection.

Sources: NJ DOL 2024 (medical evidence = top TDI denial reason) · Colorado FAMLI provider portal · MA DFML FY25 · Nava on the DOR integration pilot · Minnesota Reformer on UI wage reuse
Note: no verified example of true EHR/FHIR integration for PFML certification exists anywhere. Colorado's is a provider portal with license verification, not a clinical data feed.

The benchmark

Massachusetts, and what they actually did differently.

Nava calls it "a gold-standard PFML program." Here is what that means operationally, not rhetorically.

Employer response is time-boxed and lapses. Adjudication is never blocked on a third party.
Integration-first sequencing. The wage-data feed was validated in the first pilot, not in an integration-test phase near the end.
Pilots run in production with real users before public launch. The first pilot was deliberately scoped to have zero cross-vendor dependencies.
Self-service as a throughput strategy. Moving intermittent-leave reporting into the portal cut call volume 65% and freed 763 staff hours per month.
Teams organized around user groups — applicant, employer, staff, payments — each owning its own roadmap.
A spike team absorbs policy shocks so legislative change doesn't stall the roadmap. A statutory "top-off" change shipped in under four months.
They publish the numbers annually, denial reasons broken out. The transparency is itself part of being good.

9 days

Median initial determination, FY25 — down from 13. Median final determination fell from 18 days to 11, while volume rose 12.7%.

The counter-example. New Jersey has the oldest program in the country (1948) and now decides only 20% of disability claims within two weeks against a 75% goal. Their own report names the cause: the Disability Automated Benefits System, built in 1989, which "requires staff to complete additional manual processes." Also worth noting: 81.2% of their family-leave denials are categorized as "Other." The state cannot tell its own legislature why it denies most claims.

Sources: MA DFML FY25 · Nava: 65% call reduction, 763 staff hours/month · Nava: team topologies by user group · Nava: spike team, top-off in under 4 months · Nava: multilingual PFML · NJ DOL 2024 report (DABS, 1989)

Scoreable

The exercise asks you to demonstrate Nava's methods. Here they are, in their words.

"We are stubborn about pursuing long-term outcomes, but we are flexible and adaptable in how we get there... We tune our approach — we know when to be fast and iterate, and we know when to be methodical and get it right up front." Nava values — "Delivery over dogma"

"Our partners in government bring decades of knowledge, experience, and policy insight to the table... we invite feedback early and often, receive direct feedback as a precious gift." Nava values — "Build together"

"In a pilot, we incrementally develop pieces of the end-to-end experience and test them with real users before the service launches to the public. This helps us identify and mitigate risks early on." Nava case study — Using a pilot to minimize risk

Their planning guidance, verbatim

"Plan for cycles of discovery, testing and implementation — not just one planning phase followed by one implementation phase."
"Start with your agency and user needs, not functional requirements."
"Build modularly and incrementally." Modules: "as few as possible, but no more."
"Ask vendors to show, not tell."

And the USDS play they build against: "Ship a functioning minimum viable product that solves a core user need as soon as possible, no longer than three months from the beginning of the project." Your exercise gives you six months and one scrum team. Two three-month cycles, not one six-month plan.

Sources: Nava values · Pilot case study · Nava toolkit · USDS Digital Services Playbook

Bringing it back

What this means for your MVP.

The exercise hands you a trap and tells you it's a trap. Three goals: cut applicant time, cut staff processing time, cut operating cost. Then: "these goals can sometimes be in conflict." Most candidates will pick one and optimize it.

You already know how to refuse that tradeoff, because you wrote the argument for a completely different company two months ago:

"The right friction at the top filters customers who would have churned in the first 30 days anyway. Higher quality customers, not fewer customers. The answer isn't slower onboarding, it's right-sized onboarding." Kevin Middleton, Optimum Mobile take-home, May 2026

Same physics here. Validation and correction at intake look like they add applicant time. What they actually do is collapse the rework loop that is consuming staff hours, generating call volume, and producing a 51% appeal-overturn rate.

Three candidate MVPs, weighed

Option	Hits all 3 goals?	One team, 6 months?
Status visibility + in-place correction	Yes — fewer restarts, fewer calls, less rework	Yes
Provider certification portal	Big impact, but depends on providers adopting it	Risky
Dept. of Revenue wage integration	High value, invisible to applicants for months	Partner-dependent

The honest case for status + correction as MVP: it is the only option that moves all three stated metrics rather than trading them, a single team can ship it, it produces visible value for the client inside the first cycle, and it is the thing the real Minnesota program actually got wrong — where one applicant resubmitted eighteen times, each edit pushing her back in the queue.

Compiled 2026-07-29 from state agency reports, legislative testimony, auditor findings, and Nava's own published material. Every figure above is sourced. Where a number could not be verified, it was left out.

Citations

Every source, in one place.

No source, no claim. If a number isn't here, it isn't in the brief.

Nava PBC — methods and case studies

Nava values — "Be active stewards," "Build together," "Progress takes work," "Delivery over dogma"
Building technology to power your mission — HCD and agile definitions, procurement guidance, "as few modules as possible, but no more"
Using a pilot to minimize risk — MA, 13-month build, DOR wage integration validated in pilot one, 28 improvements found
Automating intermittent leave reporting — 65% call-volume reduction, 763 staff hours/month
Building the PFML API layer — integration-first sequencing, record 4,000-call day
Agile response to legislation — spike team, statutory top-off shipped in under 4 months
Ongoing user research — 20 interviews since Apr 2025, employer exemptions rebuild
Becoming a premier PFML vendor — teams by user group; 7 → 70+ people in MN in under 5 months
Multilingual PFML — 5.71% of MA applicants use a non-English language
Nava Labs: GenAI assistant for caseworkers · full report PDF
Nava wins Minnesota · Maryland FAMLI prime

Massachusetts

DFML FY2025 Annual Report — denial taxonomy, 5,597 abandonments, 9/11-day medians, $3.90B to date
Application approval timeline — the 5 + 10 + 14 + 7 arithmetic
Required documents — color ID, 4.5MB cap, .heic rejected, MyChart screenshots refused
Appealing a decision — 10-calendar-day window
458 CMR 2.08 — the statutory lifecycle

Washington

ESD 2025 Annual PFML Report — denial reasons (Fig. 8), approval by leave type (Fig. 7), call-queue data, the "faster decisions reduce call volume" quote
After you apply — the four published claim states
Appeals · Paid Leave Ombuds · Payments
JLARC oversight report — recommended the program adopt published performance measures
State Auditor findings on improper payments

Minnesota — the program behind this exercise

MN Paid Leave: after you apply — the "same information our Contact Center can see" line, 14-day RFI window
Minnesota Reformer: the rollout was a success — on time, $70M under budget, 340 stakeholder meetings, UI wage reuse
EP Local News — ~126K applications, ~75K approved, $598M paid, 40%+ rejection
KSTP: DEED on approval delays — certification mismatch named as most common issue
KSTP: the applicant who resubmitted 18 times
KTTC: first six months · MPR: launch week · Northern News Now
MN House: fraud-prevention hearing — three levels of identity proofing, MNIT liveness check

Oregon, New Jersey, New York, and others

Governing / Oregonian on Paid Leave Oregon — 52-minute waits, 1,500 stuck in ID verification, denials for unanswered mail
KATU: Oregon's 29-day average · KGW · OED dashboard
NJ DOL Annual FLI/TDI Report, CY2024 — 20% two-week determination rate, DABS built 1989, 81.2% "Other" denials
NY WCB arbitration report, Q1 2025 — 51% employee win rate, 38.1% "claim incomplete" · Q2 2023 for comparison
US DOL ETA: UI appeal reversal rates — the 28.7% national benchmark
Colorado FAMLI provider portal · CO regulatory changes
CA EDD claim process · CA EDD: physicians cannot be required to certify
RI DLT monthly update — the TDI vs TCI denial gap
Connecticut FY25 coverage — 71.33% of denials for unfiled documentation
Littler: Maryland delays again

Government technology practice

USDS Digital Services Playbook — Play 4 (MVP in no more than three months), Play 6 (a single empowered product owner), Play 2 (design online and offline together)
18F De-risking Government Technology Guide · State Software Budgeting Handbook (archived) — "Only 13% of large government software projects are successful"
National Partnership: state evidence for paid leave · A Better Balance on paperwork barriers

What could not be verified, and is therefore not claimed above: Colorado FAMLI denial rates or appeal outcomes · current (2025–26) Oregon processing statistics · California EDD denial-reason breakdowns · Rhode Island data after 2018 · appeal volumes for any state except New York · a formal correction or resubmission rate for any state · per-adjudicator caseload or straight-through-processing rates anywhere · any verified EHR/FHIR certification integration · a published Nava "client kickoff" playbook (the pilot case study is the closest proxy).

Independent concept work by Kevin Middleton. Not affiliated with, endorsed by, or commissioned by Nava PBC. All trademarks and product imagery belong to their respective owners.