Washington State's Department of Labor & Industries wanted an honest read on how customers experienced their offices. I designed and led a 347-person intercept survey administered in both English and Spanish, ran statistical analysis across six research questions, and delivered five high-confidence findings with a prioritized improvement roadmap that tied every recommendation to a measurable KPI.
Washington State's Department of Labor and Industries serves hundreds of thousands of residents every year across workers' compensation, workplace safety, and licensing programs. Leadership wanted to know how those customers experienced their offices: not anecdotally, but with enough rigor to make decisions and track improvement over time. I designed the research program, built the survey instrument, oversaw data collection, ran the statistical analysis, and presented the findings to the executive team.
The study collected 355 surveys between May 15 and June 2, 2026, with 347 usable responses after applying the consent filter. The sample included English and Spanish speakers across phone and in-person contact channels, with a QR self-serve option at select office locations.
I led every phase of this engagement: research design, instrument development, bilingual administration oversight, statistical analysis, equity review, and executive presentation. The survey instrument was built with six structured research questions that tied each data point to an actionable decision. Analysis was run in Python using pandas and scipy, with Spearman rank-order correlations for satisfaction drivers and chi-square and Mann-Whitney U tests for equity subgroup comparisons. The final deliverable was a 67-slide executive report with findings organized by confidence level: high-confidence findings as immediate action items, moderate-confidence as near-term priorities, and directional findings as monitored gaps.
Rather than running a generic satisfaction survey, I structured the study around six specific research questions that L&I leadership needed answered before they could make resourcing and process decisions. Each question had a defined analytical approach tied to it before a single survey was collected.
I organized findings by confidence level so the executive team could immediately distinguish between what required action this week versus what needed monitoring over the next quarter.
94.8% of customers rated their experience 4 or 5 out of 5. Mean satisfaction is 4.75 out of 5. For a public agency managing workers' compensation claims and licensing, this is a genuine strength and a defensible baseline for legislative and budget conversations.
Rep Helpfulness showed the strongest correlation with overall satisfaction (Spearman rho = 0.75), followed closely by Getting to the Right Person (rho = 0.70). 89% of customers gave staff the top rating. The strategic implication: L&I's people are the asset. The gaps are in the systems around them.
Customers without clear next steps scored 3.86 out of 5, compared to 4.87 for fully resolved customers. That 1.01-point gap is the largest single gap in the entire dataset. Notably, customers whose issues were unresolved but who had clear next steps scored 4.66, nearly as high as fully resolved customers. The intervention is not resolution. It is communication.
"There was no additional information given on next steps. The L&I staff did not know what next steps were."
Phone is where 40% of customers start, but it is the lowest-performing channel across every metric. The channel gap on resolution (66.1% in-person vs. 53.3% phone) is statistically significant. Phone customers also tend to have more complex, claims-related issues, which partially explains the gap but does not eliminate it.
Spanish-speaking respondents (n=81, surveyed in Spanish) scored higher on overall satisfaction (4.85 vs. 4.70) but resolved at a lower rate (73.7% vs. 81.0%). The satisfaction score alone does not reflect service parity. Qualitative data documented customers waiting 30 minutes on Spanish-language lines before giving up and calling the English line to get faster service.
"In Spanish he waited 30 minutes then hung up. Called the English line: 2 to 3 minutes to get a rep."
The equity analysis tested outcomes across disability status, age, language, race/ethnicity, and gender. One gap was statistically confirmed and required escalation beyond monitoring.
Customers with disabilities required four or more contacts to resolve their issue at 22%, compared to 7% for non-disabled customers. That is a 3x gap with a p-value of 0.002, making it the only disparity in the study with enough statistical power to treat as a confirmed finding rather than a signal. The recommendation was immediate operational review of how the office handles complex cases involving customers with disabilities, not additional study.
Age and language subgroups surfaced directional gaps: seniors showed the lowest first-contact resolution rates and the lowest wait time ratings, and Spanish-speaking customers faced documented access barriers on dedicated Spanish language lines. These were flagged as monitor-and-investigate gaps pending Wave 2 sample sizes large enough for confirmatory analysis.
One of the most important methodological decisions in this study was building three distinct resolution metrics rather than a single number. Each answers a different question, and conflating them leads to bad operational decisions.
The gap between the 63% single-contact rate and the 39% true FCR represents customers who contacted L&I only once but left without resolution. These are silent non-resolvers who never called back. They do not show up as repeat contacts, but they are not satisfied customers either.
I structured the recommendations to give the executive team decisions they could make immediately alongside a phased plan for systemic improvement. The framing was deliberate: all four quick wins required no new technology and could be initiated in the same week the report was presented.
The longer roadmap targeted: Top 2 Box satisfaction at 95.5% or above within 12 months, repeat contact rate reduced from 34% to 28% or below, and a continuous listening infrastructure so CX becomes an ongoing capability rather than a one-time measurement exercise.
The study delivered a 67-slide executive report with full methodology documentation, replication instructions, and a Wave 2 instrument redesign guide. Every metric in the report is reproducible from the raw data file using explicitly documented field definitions and filter logic. The research gave L&I leadership their first statistically grounded baseline for customer experience: a defensible starting point for budget conversations, program investments, and continuous improvement tracking across future survey waves.
The equity finding on disability contact frequency was the highest-urgency output. It identified a confirmed, statistically significant gap that was invisible to the organization before this study and prompted an immediate operational review of how complex cases are managed.
The most important design decision in this study was building three distinct resolution metrics rather than one. Organizations almost always default to a single satisfaction number. That number is easy to report and easy to game. Separating first-contact resolution from eventual resolution from single-contact rate gave the team a much more honest picture of where the system was actually failing customers versus where it was succeeding.
The equity findings reinforced something I think about on every research project: aggregate scores can hide the people who most need attention. A 94.8% top-box score looks excellent until you break it out by disability status and discover that a specific group is coming back three times as often. Research that doesn't look for those gaps isn't really measuring service quality. It's measuring the average experience of people who didn't fall through the cracks.
Whether you're rethinking a flagship product, scaling a research practice, or trying to make sense of complex user behavior, I'd love to talk.