Miguel Ángel Pérez-Holdsworth

• On the team modernizing Colorado's 40-year-old state payroll system: the federal Quality Assurance Surveillance Plan adapted as the quality standard, an automated test suite at unit, integration and end-to-end levels, and biweekly usability testing with state employees. The State published the approach as a case study in October 2024.
• Helped facilitate the assumption mapping that put correctness, not speed, first; the State's case study credits the team with the facilitation.
• Second workstream: identity and sign-in for the State's resident identity platform, past two million registered users by June 2026.
• Technology procurement consultation: advised agencies and vendors on solicitation and contract language, applying evaluation design to procurement. Built a design-practice survey from two of Jeanne Liedtka's instruments; started a twice-monthly Lean UX book club whose takeaways reshaped the weekly design sync.

A staffed research and design consultancy rather than a solo practice: subcontractors under signed agreements since 2015, client MSAs and SOWs, conversion optimization and experimentation as the core business.

• Arrow Electronics, via ClientSolv · 2019 – 2020. A 90-day discovery engagement before an investment in a self-service portal for partners: 39 recorded interviews across stakeholders, employees, resellers and suppliers; four ideation workshops across Sales, IT, Finance, Operations, Marketing and Engineering; an atomic research repository of 1,600+ observations and 50+ co-created insights; eight ways a portal could drive profitability, three on revenue and five on cost. The process and repository stayed with the team.
• Whistle Labs · 2020 – 2022. Re-hired by the Love With Food founder four years on. Research through redesign: seven recorded interviews into personas and test scripts, a product-market-fit survey and competitive analysis, A/B infrastructure and Heap analytics, and training the web and mobile teams to run research. Owners of flight-risk animals emerged as the loyalty segment; the homepage and campaign were rebuilt around it.
• Turo, via Huge · 2021 – 2022. A monadic multivariate test exposing 400 participants to four TV commercials and out-of-home banners with segment-based sampling, ten listening sessions, and a repeatable campaign-evaluation method.
• Autodesk, via Tomorrow Partners · 2021. Coded interview recordings, storyboarded the current state, and co-created future-state scenarios and stage diagrams for a ten-stage generative-design workflow.
• Tenasol · 2023 – 2024. The full design surface for an AI text-analytics startup ahead of its raise, with two subcontractors: brand identity, product UI, investor decks and website.
• Smaller engagements · 2011 – 2021. Usability testing, analytics audits, customer surveys and SUPR-Q for an e-commerce retailer; CRO and UX audits and roadmaps under MSA and SOW; learning-platform UI; design systems and voice-and-tone documentation.

• Ran community-centered engagements and action research for the STARs-as-customers team (workers Skilled Through Alternative Routes), funded by the Gates Foundation; conducted the intake interviews for the community fellowship's design-a-thon.

• Built the business case that replaced the design vendor's $687k-a-year contract with four in-house designers: total design spend fell 26% while delivery capacity doubled (modeled). Design became an official corporate goal in 2019 with approved headcount.
• Ran a design-maturity survey (100+ responses across four business units), identified a founding cohort of 20 practitioners, and designed the certification program: 20 per cohort, quarterly, toward 100 actively leading research. Owned hiring for the new team: job descriptions and the compensation analysis with HR; 200+ qualified applicants answered the claim that designers would not want to work there.
• Two years of discovery research and strategic design avoided roughly $2.4M in code rewrites (own estimate). One instance: a prototype of a medical natural-language-processing product, built for about a tenth of the production budget, shipped for the annual client event in three months; contextual inquiry found clinicians did not fully trust automated tools, so the product was repositioned as an enabler and its interface simplified on that finding.
• Built the design-ops infrastructure for the in-house transition: a documented, version-controlled design system, front-end governance, onboarding and compliance processes, and an atomic research repository; research ran on protected health information under HIPAA with de-identification built in.
• For a market-analysis product built on external research alone, captured the assumptions, evaluated them with clients, and convinced the sales team to ask research questions on calls; contract sizes grew and a new department formed to find repeatable services.

• First UX hire, with no direct access to users. Ran the company's first rigorous usability study, of the intranet, with five employees on ten tasks; the readout taught HR leadership what user research is, and that alliance became the business case for the Center of Excellence. Introduced human-centered design company-wide through lunch-and-learns tested with colleagues first.

• 19+ in-depth interviews with community leaders, journalists, analysts and developers, a usability study of the open-data portal, and an intercept survey on it; synthesized three systemic challenges, proposed interface concepts and a service blueprint, and presented the findings at a City press conference.

• A 90-day market-entry strategy for the festival-fan market on lean and design principles; roughly 60 customer interviews across the team, the first wave run personally with the scripts, guides and recruiting funnel the team then used. Shipped Curadora, an MVP that converted prospects into paying customers, with a behavioral-tracking and innovation-accounting plan.

• Led research, design and front-end for a responsive redesign that lifted sign-up conversion 60%, measured by A/B test; the full method is published in UX Collective (2019).
• A 118-variable member questionnaire with correlation and factor analysis derived two personas; nine interviews per segment; six documented homepage generations from wireframe to live across five breakpoints.
• Built the front end: a Handlebars template system, a responsive build pipeline, versioned assets on S3, and a design system engineers could extend.
• Redesigned account management on a support-cost signal and tested a high-fidelity prototype unmoderated: top-task completion rose from below 30% to eight of eight participants.

• Shipped front-end for Free People, BHLDN, Terrain, One Stop Plus and others on an enterprise e-commerce platform; pioneered a design-system and style-guide program, borrowed from game tilesets, that became the company standard.
• Led the Free People China internationalization: tokenized strings, layouts rebuilt for variable phrase lengths and both text directions, then rolled out to further countries, with the internal i18n guide. Built the new-hire onboarding program and mentored the developers routed through the team.

• Built CMS-powered sites in HTML, CSS, JavaScript and PHP from agency comps, two or three at a time; trusted with client conversations and technical estimation at 18.

• Co-created Efencea, a PvP tower-defense game, at 16: first playable in eight days, rebalanced on player feedback, still live.

• Benchmarked against expert human coding over 13 fidelity cycles: agreement rose from 48% to 72%, persona coverage from 50% to 92% (V-measure 0.61 to 0.86). Ran a blind human-versus-machine coding study with pseudonymized speakers and a pre-registered rule locking the human read first.
• The harness twice overruled the result its author wanted: a NO-GO on a statistically significant win, and a circular self-metric caught (98.7% claimed against 55% measured). A source-comparison study measured 34% speaker-attribution bleed on found data against 0% on interviews; a second data class ran at 0.00% contamination against a 5% bar, for $25.44.

• 48 pre-registered experiments, 34 falsifications; tournament evaluation with Wilson-scored tiers, entropy, Nash support and Hodge decomposition; a CI-enforced citation graph on the design record. A $101 paid fake-door test of 15 concepts (9,205 impressions, 9.4% click-through) that also caught its own blind spots.

• 4,300+ GitHub stars, 2,600+ live sites a decade on; a CSS-Tricks guest post and a Hacker News launch.

• Extended atomic UX research with the FAIR data principles; launched on Instagram with video lessons, templates and check-for-understanding segments, teaching a practice already run in production at Inovalon and Arrow.