UX Researcher · B2C + B2B · 7 years
I'm the voice of the customer in the room where the product is created.
I turn messy qualitative indicators into evidence clear enough to change a roadmap and confident enough to change a leader's mind.
- Currently
- Senior Design Researcher, Klaviyo
- Focus
- Customer Needs · Skill Building · Research AI Automation
- Methods
- Interviews · Usability Testing · Surveys · Synthesis
- Superpower
- Stakeholder management under pressure
About me
I've spent 7+ years in UX research for B2C and B2B products. I always say the same thing about the role: I am the speaker for our customers. When a team is deciding what to build, I stand up for the user to make sure the needs are represented.
In practice that means thoroughly planning and running the research from begining to end; synthesizing the data and managing the stakeholders expectations so the findings are easily accepted even when the data contradicts the stakeholders' assumptions.
I am a father of 2 kiddos and currently live in Trumbull CT. In my free time I enjoy gaming (tabletop and video), and exploring nature wtih my family and freinds. I have an encyclopedic knowledge of movies, music, and history which make a great teammate for trivia nights!
How I work
- 01Mixed methods, aligned to the question. Interviews and usability testing for the “why,” surveys to quantify and validate it.
- 02Evidence over opinion. I let customer stories, clips, and verbatims do the persuading — especially when the data contradicts an assumption.
- 03Shareouts tailored to the audience. Execs receive the human story; product gets the actionable next steps.
- 04Speed when it counts. I've delivered decision-ready findings in an 8-day sprint without cutting the rigor that matters.
Selected case studies
Five research projects across Autodesk and Klaviyo — a fast AI-agent adoption sprint, a deep new-user activation study, a forward-looking “big bets” discovery program, a foundational construction-software study, and a 3-week RITE sprint to validate a new product. Tap any card to read the full story.
Case Study 01 · Klaviyo · Q3 2025
Essential Flows:
the AI agent people didn't know was a builder.
An AI agent analyzes a user's account and pre-builds the automations it judges essential — turning days of setup into seconds. Three weeks into early access, almost no one was adopting it. I had 8 days to find out why.
- Role
- Lead Researcher — plan, run, synthesize, share
- Timeline
- 8 days, end to end
- Methods
- 15 participants · Interviews + usability
- Impact
- +13% adoption of AI Agent
Adoption stalled below 1%
The agent was built to take our users' most complex, time-consuming jobs — building automations — and collapse it to seconds. It was aimed at new users but scaled up to more advanced recommendations for power users.
After three weeks in early access, adoption sat at 1%, well below leadership expectation. Nobody knew why. The leading theories: poor discoverability, users not understanding what the tool was, and users disliking the flows it produced.
Leadership was heavily invested and held strong assumptions about how the tool should look and be adopted. The evidence needed to be undeniable.
Two lenses in one hour
15 participants across 60-minute sessions, split between users inside the early-access program and users who'd never seen it — allowing me to separate “new-tool confusion” from “this-specific-tool confusion.”
Recruited by behaviour, not just status
Early-access users were segmented by exactly how far they'd gotten with the recommendations — so each drop-off point had a voice in the room.
- G1Saw a recommendation card but never clicked
- G2Clicked and previewed, but didn't publish
- G3Turned a flow on without editing
- G4Turned a flow on after editing
- G5Turned one on, then set to draft / deleted
Interview + usability, back to back
Interviews explored how people already use AI for flow creation, and — for early-access users — what actually happened when they met the tool.
Usability testing put each person through two prototype versions, randomized per participant, thinking aloud, with checkpoint questions at each key moment of the flow.
The plan included a large early-access survey. With only 8 days, I removed this survey and put full focus into getting the qualitative story right.
The experience under test
Participants were shown prototype versions alongside the live recommendations inside their own accounts. Only one of six participants could even recall interacting with the AI Agent before. This was the first signal that the problem started long before the automation preview was ever shown.
It's just two lines and it looks like a picture of one of my standard flows.
— Participant, on the entry bannerThree signals, two that mattered most
The entry point read as a reminder, not a builder
Users thought the banner was a nudge to go build a flow themselves — that clicking would dump them into the standard flow builder. They didn't realize a purpose-built agent was offering to do the work for them. Discoverability was the wall; most never got past it.
People expected to see the flow's logic first and distrusted it when they didn't
To these users the conditions and timing are the automation; the email content is secondary. Not seeing a tree/logic view early created confusion and outright distrust — one participant said they would exit the experiance immediately. The old design didn't reveal the automation flow until the very end, so people discontinued use before they understood what the tool had presented.
I would immediately hit exit… I want to know what the flow looks like before.
— Participant · 5 of 6 expected a tree view firstMaybe it can be a preview of the flow… in terms of the logic.
— Participant“Why should I care?” needs more than a dollar figure
A projected revenue number wasn't enough on its own. Users wanted the reasoning — why this flow matters to their business and how it grows their brand. A secondary finding, but a clear factor for making the recommendations feel worth acting on.
The evidence changed minds — and the metric
Because this went to very different audiences, I built two shareouts: a findings deck, and a 2-minute AI-voiced podcast for leaders who wanted quick consise findings on their time.
For execs, I leaned hard on customer stories, clips, and verbatims. These mattered most where the findings contradicted the CEO's assumptions. Letting users speak for themselves reframed the conversation and helped leadership see the value of researching before shipping.
Both key recommendations were built and tested in production against the original experience.
Production A/B · adoption
My key takeaways
This study taught me I could resepctfully push back on leadership to do right by the customer and that the win lived in the communication as much as the findings. Tight, ongoing context-sharing with the VP of Product kept the project aligned all the way to a shareout that landed despite strong prior beliefs.
Case Study 02 · Klaviyo · Q2 2025
Activation Runway:
why new users stall before takeoff.
More than half of new users weren't completeing onboarding in their first two weeks. I co-led a 3-month, mixed-method study to find the barriers and give six teams a shared, evidence-based picture to build against.
- Role
- Lead Researcher working alongside a mid-level researcher
- Timeline
- 3 months
- Methods
- 12 interviews + 281-user survey
- Impact
- 12-month roadmaps + pricing overhaul
Adoption varied wildly — and no one knew why
Klaviyo gives new users plenty of tools to get set up, but who actually used these tools varied enormously. The goal: understand what holds people back in their first 7–14 days, and turn that into the 2025 activation OKRs.
We defined activation concretely — sending a campaign, setting a flow live, or adding a form to a live site — so “success” meant the same thing to every team reading the findings.
A two-researcher, months-long study is rare at Klaviyo. We had to push to get it approved, then conduct it as background work — the reason it stretched across a quarter.
Depth first, then breadth
We ran the study in two phases so the numbers had stories behind them:
- P112 interviews with users who had been on the platform under 3 months — exploring the onboarding experience and what shaped whether they activated in 14 days.
- P2281-user mixed-method survey of users who had been on the platform under 6 months — to validate and size the themes the interviews surfaced.
Every group spanned a range of GMV and company size, and split cleanly into early activators (within 14 days) and late activators — the comparison that carried the whole analysis.
The barriers were mostly psycological, not usability related
People weren't blocked by usability. They were blocked by readiness and confidence — feeling they lacked the content, strategy, or subscribers to be “ready,” and fearing they'd do it wrong.
Flows are the draw — and the fear
Automated flows were the top reason people signed up and the biggest long-term value they named. They were also the most intimidating feature — even early activators weren't sure they had set them up correctly.
Barriers are emotional before they're operational
Late activators are stuck on readiness and nerves; early activators moved beyond that and were stuck on execution — content, assets, time. Onboarding has to support the emotional journey, not just the technical one.
The barrier progression
You get huge self-doubt because it's sold to you like it's so easy.
— P6, late activatorI am in great need of mentorship. Someone to tell me “you're doing it right.”
— P8, late activatorThree more patterns shaped the recommendations:
Users lean on external AI like ChatGPT over Klaviyo's own AI because “it already knows my business”;
Roughly 40% seek help outside Klaviyo, and mentorship — real or perceived — measurably accelerates activation;
Upgrading feels like friction, not progress, with profile-based pricing the single biggest source of confusion.
Nine POVs the org could build against
The study resolved into a shared set of positions — each one a direction, not just an observation.
Serve beginners and explorers both
Offer multiple paths — step-by-step for the overwhelmed, open exploration for the confident.
Early value = quick wins, not full setup
Guide to 1–2 high-impact actions and reinforce with visible results fast.
Support the emotional journey
Reassure, normalize “good enough,” and make it safe to preview and test.
Make upgrading feel like success
Frame hitting limits as growth; clear the path for high-intent users.
Mirror the consumer journey, not the product tree
Frame features around end-customer touchpoints, especially for SMS.
Meet users where they are on analytics
Early on, simple reassurance beats complex dashboards.
Let users dictate their own setup
Allow skipping irrelevant steps; keep surfacing helpful next actions.
Human mentorship accelerates onboarding
Make webinars, community, and guidance far more visible and accessible.
Reposition Klaviyo AI as a strategy partner
Offer next-step prompts and setup confirmation to win back from external AI.
What the research moved
12 months of roadmap
The onboarding and activation teams took the low-hanging fruit for quick wins and folded the bigger recommendations into their yearly roadmap and staffing.
A pricing overhaul
Once the pricing team saw how the upgrade path was suppressing conversion, they ran multiple sprints and reworked how users upgrade in-platform.
My key takeaways
A study this big is only as useful as its shareout. Feeding findings to six teams over months taught me to reshape the same research for each audience — the spark that pushed me to start building tailored artifacts (and to use AI to help produce them). Next time I would push to make a project this valuable a primary focus, not background work, so momentum and attention never fade.
Case Study 03 · Klaviyo · Q4 2025 → Q2 2026
PLG & Big Bets:
finding the bets before the roadmap is written.
A generative discovery program I proposed and led. The brief wasn't to evaluate a feature — it was to find the opportunities that would make existing customers expand with Klaviyo and stay. The hard part was turning open-ended exploration into bets leaders could actually act on.
- Role
- Lead Researcher — proposed, scoped & drove the program
- Shape
- Multi-quarter, 3-phase — Round 1 delivered
- Methods
- Workshop · 8 tool-exploration interviews · 2×2 prioritization · survey
- Outcome
- Set direction for a phased PLG roadmap
Research before the roadmap exists
Most of my work answers a question about something already being built. This was the opposite: a foundational, future-facing study to uncover the “big bets” — where Klaviyo could expand its footprint inside a customer's business and become the tool they reach for instead of the five others they stitch together.
That framing changes the job. There's no prototype to test and no single metric to move. Success is giving product and design a defensible point of view on what to explore next — and doing it early enough to matter.
Aligning stakeholders on where they'll be in 6–8 months, not where they are today. I de-risked that with incremental updates that showed the path the findings were tracing — so direction never arrived as a surprise.
A future-facing program, built in phases
Rather than one big study, I structured the work as a phased program — each phase narrowing from broad exploration toward testable solutions, so the organization could commit incrementally instead of betting everything up front.
Phase 1 · Q4
Get customers moving faster
Explore the whole stack
A team workshop to surface existing assumptions, then interviews + a survey mapping every tool customers use inside and outside Klaviyo, and why.
Phase 2 · Q4 → Q1
Customer exploration & deep dive
Pressure-test the bets
Concept interviews on the highest-potential ideas — persona tooling, analytics — then quant validation with a representative pool to size friction and demand.
Phase 3 · Q1 → Q2
Experimentation & evaluation
Build & test tactically
Design iterates on prototypes and flows; we test at the tactical level against the earlier findings — a rapid loop to prove what's worth staffing.
The Phase-1 survey was built to reduce bias: consistent 5-point scales for comparability, “select up to 3” on goal questions to force prioritization, and neutral phrasing that never implied a tool should be easy or valuable.
Map everything, then force a priority
I recruited 8 Klaviyo users across a spread of maturity — from entrepreneurs through upper-SMB, plus one agency. Each session started with their goals, then walked through every tool in their stack: key features, likes, dislikes.
The anchor was a 2×2 activity: participants placed each tool by how hard it is to use against how important it is to their goals. It turns a sprawling, subjective conversation into one shared, comparable picture — and makes the priority argument for me.
Reading the eight boards together, a pattern held: the capabilities customers cared about most kept landing in the hard-to-use half — 8 of 10 analytics tools sat squarely in the important-and-difficult quadrant. That became the spine of the readout: four areas worth betting on, ranked by the evidence rather than by opinion.
Four opportunities, sized and sharpened
Analytics — make users their own analytics master
8 of 10 analytics tools landed in the important-and-hard quadrant. Smaller teams don't have an insights function, and they don't want another dashboard — they want the “so what / what next.” Proactive recommendations, forecasts built on their own history, and live goal tracking.
We have a monthly meeting where they talk us through the numbers… which customer segments we should be focusing on.
— Participant, on the analytics tool they pay extra forSegmentation — reasoning, not just the mechanics
Every external segmentation tool sat in the hard-but-important zone. Users can build a segment; what they lack is the why — which audience to target, with which message, as part of which journey. The opening is persona-based recommendations that carry their own justification.
Sometimes I'm wondering what the segment should be for a certain follow-up email… it helps with that kind of thing.
— Participant, managing four different listsAI tooling — let small teams punch above their weight
AI was seen as the force multiplier for lean teams — saving time and money that goes straight back into marketing. The sharpest unmet need: turn a layout or a plain-language prompt into paste-ready HTML, plus specialist agents trained on the brand that grow with the business.
It takes our brand guidelines, I prompt it, it designs the email and gives me the HTML to paste into Klaviyo. It'd be awesome if Klaviyo had that.
— Participant, describing a tool they wish Klaviyo would buildAds — “a necessary evil” (and a retention risk)
All 8 participants were frustrated managing ads: constant platform changes, biased attribution, no way to compare campaigns side by side. Not a core bet — but a real source of churn pressure. I flagged it as a watch area rather than a lead investment, with a clear rationale for the lower priority.
How I made the next steps actionable
Generative research dies in the shareout if it lands as “interesting themes.” My goal was the opposite: every finding had to arrive as a decision someone could make on Monday. Three moves did the work.
Lead with the bet, backed by the matrix
The readout opened with a single headline — the high-impact, high-difficulty direction — before any detail. The 2×2 gave that claim defensible shape, so leaders debated the bet, not the method.
Show it, don't just say it
Each opportunity was paired with an AI-generated concept of what the solution could look like inside Klaviyo. Stakeholders reacted to something tangible instead of an abstract theme — which moved the conversation from “is this real?” to “how far do we take it?”
Split quick wins from the roadmap
Every bet broke into low-hanging fruit teams could ship now and larger plays to staff on the yearly roadmap — each with a clear next step, so the research handed off cleanly into delivery.
The “tailor the artifact to the audience” habit from my activation work showed up here too — concept visuals and a tight TLDR for leadership, the full evidence and tool-level detail for the product teams who'd build it.
Reflection & the road ahead
Round 1 set the direction; the program is built to keep earning it. Next is a larger survey to validate the four areas at scale, and running the same study with non-customers to see where the opportunities hold beyond our own base. Leading a generative effort end to end — proposing it, structuring the phases, and defending the bets — is the work I most want more of: shaping what gets built, not just refining it.
Case Study 04 · Autodesk · Construction Cloud
Handover:
where should a construction product live?
From my time at Autodesk, on the Construction Cloud (ACC). Handover is the final stretch of a build — consolidating a project's data and transferring it from the construction team to the owner. The team wanted to build a solution for it; leadership needed to know whether it should stand alone or live inside Build. I led the research that answered it.
- Company
- Autodesk — Construction Cloud (ACC)
- Role
- Lead Researcher — embedded with a new, low-research-maturity team
- Methods
- 19 participants · Interviews · Journey mapping · Concept pitching
- Outcome
- Settled the build-vs-standalone decision & shaped the roadmap
A messy process, and a decision to make
Handover is one of the last steps of construction: the builder consolidates and structures everything about the project and hands it to the owner. Today it's done with spreadsheets, email, and whatever file store is handy — and it's painful enough that scale alone tells the story.
Autodesk's team wanted a dedicated Handover solution. The open question for leadership: should it be a standalone product, or a module inside Build (ACC's site-construction tool)? That's what the research had to settle.
One Handover saw us using 3 vans to deliver all of the documents!
— Participant, General ContractorA brand-new team that had never worked with a researcher and had low research maturity. Part of the job was delivering the study — and showing them what research could do.
Meeting a new team where they were
The kickoff made one thing clear: the team didn't know what they already knew. So before any sessions, I wrote a comprehensive guide to what handover actually is — which doubled as a map of the holes in our knowledge — and aligned product goals with research goals.
A 60-minute session, three parts
- 10mUser interview — break the ice and learn each participant's day to day.
- 40mJourney mapping — walk the entire handover process, end to end.
- 10mConcept pitching — react to solution-location workflows.
I ran a test session with an internal SME as “participant 0,” which showed the team what a session looks like and let me tune the wording to real-world jobsite lingo before going live.
Who I talked to
19 participants — 18 general contractors and one large multi-trade subcontractor; by role, 11 project managers, 7 project engineers, and 1 VDC manager. All were hands-on with handover as part of their normal work.
Recruiting ran three weeks through Research Ops, with a $50 incentive for a one-hour session. Four internal SMEs were met first to ground the plan in industry reality.
Mapping the entire handover process
The journey map was the centerpiece. Mapping all six stages together surfaced the pattern that decided the project: handover ideally starts at project kickoff, but realistically happens at the very end — often when the project is already 90% done — which is exactly how it gets forgotten.
Across the map, the work of handover overlaps heavily with the day-to-day of construction already happening in Build — the same people, the same documents, the same moment. Participants described themselves as the connective tissue of the project.
I'm the information brain of our project… and also kind of the paper-pusher.
— Project Engineer, General Contractor“It needs to be a module of Build”
The recommendation was clear and the evidence was lopsided: a Handover solution should live inside Build, not stand alone. It sits on the most overlapping workflows, saves time by avoiding tool-switching, needs less onboarding, and — critically — keeps handover visible so it isn't forgotten until the end.
I didn't bury the counter-case. A minority wanted it separate, for corporate alignment reasons and a “handover mindset” of keeping it distinct — a real concern I flagged alongside the recommendation.
I don't personally think there's anything that you could tell me that would make me like “Oh yeah, it's way better as a standalone.” You know, it needs to be a module of Build.
— Top 100 E&R Construction Company“So being able to reference that would be good, but… the main thing is keeping it separate, being able to see the progress.” — the standalone view, presented honestly next to the majority.
A decision made with confidence
On the decision
The findings let the board make an informed call on where the product should live and gave product the evidence to prioritize the roadmap. The journey map added a layer of confidence to that decision — and made plain how much users cared about getting this right.
On the understanding
The map reframed handover as more integral than the team had assumed, showed how easily it slips to the end, and laid out which features are needed when — the ideal versus the realistic. Its value reached well beyond this one project.
Feature prioritization via card sort and design validation, with a closed beta planned for six to eight months out, pending engineering availability.
What I'd carry forward
Leading a study for a team new to research taught me as much about building research maturity as about handover. The lessons I took: insist on low-fidelity concepts — detail pulled participants off the high-level questions I needed; back the story with export analytics next time; and update the extended engineering team throughout, with a mid-project retro rather than one at the end, so they can voice concerns while there's still time to act.
Case Study 05 · BuildingConnected (Autodesk) · Construction
Building the right RFI tool in three weeks.
From my time at BuildingConnected, an Autodesk company. An RFI — Request for Information — is the one project channel that contractually requires an answer, and tracking them is a mess. Using the RITE method, our team went from assumptions to a validated product concept in three weeks: 15 hours of interviews, 50+ concepts, iterated weekly.
- Company
- BuildingConnected — an Autodesk company
- Role
- Researcher on a cross-functional RITE team
- Method
- RITE — Rapid Iterative Testing & Evaluation
- Scope
- 3-week sprint · 15 hrs interviews · 50+ concepts
The one answer you can't skip
On a construction project, RFIs are the only channel that contractually requires answers — yet the early signals of an issue are rarely shared with the people who could fix them. The result is predictable: lost emails, missed questions, and all parties hemorrhaging money when issues aren't caught in time.
There was no single tool to track, organize, and collaborate on RFIs across the project. And because accuracy protects against lost money, damaged relationships, and even litigation, everyone defaults to “cover yourself” mode. The research had three questions to answer:
- HMWFacilitate prioritization and add data to RFIs to flag and forecast project risk?
- HMWSet the tone from day one with a uniform, accessible log — from design coordination through facilities management?
- HMWEase the tension between protection and collaboration?
There are better ways of managing RFIs. I encounter a lot of pain trying to chase down the responsible parties in the allotted time.
— Project Manager, video interviewThey're very time consuming. Half the time you can answer the question quickly, but they have to track via a paper trail and this causes the slow down.
— Video interview participantThree weeks, the RITE way
The approach was RITE — Rapid Iterative Testing and Evaluation: test a concept, learn, change it, test again, in tight weekly loops. Across the sprint that meant 15 hours of user interviews and more than 50 different concepts, with a team pulled from across BuildingConnected.
Interviews
Hours of user interviews over three weeks — enough to start hearing the same things back.
Concepts
Different concepts floated, sketched, and put in front of users — most discarded fast.
Weekly loops
Each round sharpened the last, so by week three the signal was unmistakable.
Are we making the right thing?
Speed only works with a way to stay honest. Every concept was scored against the same lens — desirability, viability, and feasibility — so the team made evidence-based calls and stayed aligned instead of arguing taste.
Running the same criteria every round turned “which concept feels right” into a defensible comparison. Paired with what we heard in interviews, it kept the team confident we were validating real needs — not our own assumptions.
The tasks that earned their place
After three weeks of hearing the same needs validated, a core set of tasks was clearly worth building. The RFI track log became the one-stop shop — no more bouncing between Excel and Outlook.
- 01Set priority & assign owners — no more “do I need to do this?”
- 02Reminders on high-priority RFIs so nothing slips.
- 03Manage multiple RFIs in one tracked log with status at a glance.
- 04Attach documents, respond in-app, add observers, and keep a full history.
The ability to remind the design team is super nice… so we can keep an eye on the high-priority ones.
— Architect, video interviewWhat it does for customers
Save time by centralizing
Notify, answer, and respond from one place inside BuildingConnected — cutting the extra steps of checking Excel and Outlook.
Divide and conquer
Get work to the right people for quick, efficient responses instead of chasing down owners.
Cover yourself
Solid issue tracking is good risk management — flags and a clear log give users control of their RFIs.
If you solve it, sign me up!!!
— Video interview participantThe validated concept moved toward high-fidelity prototypes, ACS software integration, and engineering — with research continuing weekly rather than monthly.
How we grew at BuildingConnected
The lasting win wasn't just the RFI tool — it was proving a way of working. We tied what we built to actual user problems, concepted on user needs before settling on a product direction, and set the product on a “test and learn” foundation. Most of all, we streamlined research to run weekly — work that used to take a month.