Workforce & caregiving research advisory

Thousands of survey responses, themed, scrubbed, and analyzed, in a single project

Launch by Lunch built ELIXR a research project in Claude that turns raw open-ended survey data into themed, privacy-safe, analysis-ready findings, with a human checkpoint at every step. Work that used to be slow and manual now runs end to end, and it is reusable for every new client.

2,300+open-ended responses coded

0.3%needed a human privacy check

3 wks →
6 hrsmanual coding time, before & after

Industry: Workforce & caregiving research
Team: 2 co-founders + small research & ops team
Engagement: Research project build
Timeline: May → June 2026

The challenge

Incredible research insights, done manually, that couldn't scale

ELIXR helps organizations understand the caregivers in their workforce, work that hinges on reading thousands of open-ended survey responses and turning them into honest, usable insight. Doing that well is painstaking: every response has to be read and themed consistently, scrubbed of anything that could identify a person before it ever reaches a client, then woven together with the quantitative data.

It was repetitive, time-consuming, and high-stakes, one missed identifying detail in a client-facing file is a real privacy risk. It also hit a ceiling. Growing the client base or the volume of responses meant adding people and time, not just pressing go. ELIXR needed a way to do this faster and more safely, without giving up the human judgment the work depends on.

What we built

Three skills that move raw data to client-ready, with checkpoints

A reusable research project in Claude that's easy to replicate client over client.

Themer: consistent coding at scale

Reads every open-ended response and codes it against ELIXR's own theme taxonomy, the same lens applied to thousands of open-ended responses in a cycle, with a human checkpoint to confirm the themes before anything moves forward. Responses submitted in Spanish are translated inline into English and coded with the same lens, no separate workflow.

PII Scrubber: safe by default

Automatically finds and removes identifying details to produce a clean, client-facing copy. It over-redacts on purpose and logs every change, so only the rare genuinely ambiguous case lands in a short human review queue.

Analysis: qual + quant, together

Merges the coded themes with the survey crosstabs and the team's intake knowledge into staged findings, with built-in guardrails (small-group suppression, significance gates) so the deliverable is never built on partial or unsafe data.

Reusable for every new client

The skills resolve each client's data by a simple naming convention, so a new study is a fresh workspace and a couple of swaps, not a rebuild. The team owns and maintains it going forward.

The results

Faster, safer, and built to scale across clients

24themes applied consistently to every response

287identifying details auto-redacted

2,900survey responses analyzed end to end

Built to scale across clients: a new study is a setup, not a from-scratch build, so the project replicates client over client.
Gets smarter every cycle: as ELIXR grows its theme taxonomy and learns across client projects, the system compounds and improves over time.
Privacy protection runs by default: identifying details are removed automatically, with every change logged and a tiny review queue (0.3%) for edge cases.
Consistent, defensible coding: every response is themed against the same taxonomy, no drift between cycles or reviewers.
Accurate and consistent in two languages: all 2,300+ responses, including those submitted in Spanish (translated inline into English), are coded with the same lens.
Human judgment kept where it matters: checkpoints at every stage keep the team in control of what ships.

Have a repeatable, high-stakes workflow?

If your team runs the same careful, manual process for every client or cycle, we can build something that does the heavy lifting and keeps your judgment in the loop.

Book a call with Launch by Lunch