Run a freelist with confidence. We show you how and exactly where it stops.
One good, plain question opens up how a group really thinks about a problem. Here is what a freelist is good at, five phases to run one yourself with tools at each step and the limits worth knowing before you start.
What it is good at
Where a freelist earns its place.
It is quick and it travels
You can run one in an afternoon, in person or in a browser. Anyone can add it to the work they already do.
It stays in people's own words
A freelist asks one open question. People answer in their own words rather than picking from a list someone else wrote. So what comes back is how they actually think about the problem, not how anyone guessed they might.
It surfaces how people respond
The quiet fixes people create may not reach anyone who could spread them. A freelist that asks how people respond to a problem can surface bright spots that may help many others.
It points somewhere to test
It ends with the smallest changes worth trying, so discovery turns into a next step, not one more finding.
FAQs
Questions people ask before they start.
How to run it
Five phases. Dip in and out at any point.
A check rides inside every phase, inspired by the work of We All Count. Done well, these checks make the research better.
- Ask a good question. A freelist starts with one open question that asks people to list everything that comes to mind, such as, “What makes it hard to keep SNAP and Medicaid?” The wording matters: it names one clear topic, in the words your participants use, without suggesting any answer.Check: Would the people answering frame the question this way? Test the wording with them before you field it.
- Gather the lists. Ask on paper, in your own sessions or inside surveys you already run. Our open source questions carry you: the freelist, then the top three, what they feel causes each, how they respond and what other ideas they have, starting with the no-cost ones. Prefer the interviews run for you? Our automated interviewer holds the conversations.Check: Mode, language and reading level decide who gets to answer at all. Private answering protects the people with the least power in the room.
- Bring the lists in and confirm the reading. The import step reads your lists and shows you exactly what it understood before anything runs.Check: Every automatic change is shown before anything runs, so you decide whose words survive.
- Review and decide. The Review Desk puts every phrase in front of you. You can group, split and move. The tool finds the overlaps and the gaps within one group or across roles, like the teams running a service and the people using it. This is where a gap shows up. Two roles can name the same thing and mean something different by it. The Review Desk keeps both readings rather than merging them into one.Check: Findings stay within each role rather than averaged across roles, so a small group is never dismissed as noise.
- Leave with a plan. Your findings turn into a plan with milestones at 30, 60 and 90 days. A shareable slide deck like the one shown here is on its way for this free path. Our automated interviewer already produces one alongside the spreadsheet for the teams testing it with us.Check: Who benefits from what you found and do the people who shared it learn what came of it?
What to know before you start
Six limits and what we do about each one.
These come from the peer-reviewed literature, not from us. Almost all of them are manageable. Here is each one in plain words and what people.to.policy does about it.
A salience score is like a poll number
Ask twenty people what comes to mind about breakfast. Eggs scores high because many name it and name it early. Ask twenty different people and eggs moves, the way a poll result moves between samples. A poll never prints its number alone. It adds a margin, the few points the result could be off by. An interval is the same idea written as a range, from the lowest to the highest value that is still reasonable. That range shifts for reasons the research names: who was asked, how hard the interviewer probed, how long each list ran and how answers were grouped.
The consensus model takes no salience score. Salience belongs to the free-list step before it, where the method's authors measure it by how many people name an item and how early.
What we do: In the reports we write for clients, we show items in the order the numbers give, with people's own words. We are building Cultural Consensus Theory for All to print the scores with a range beside each one and the pages we relied on, so anyone can read them.
The science: Major-Smith and Purzycki (2026, page 2), Weller and colleagues (2018, pages 11 and 15), Weller (2014, page 351), Weller and Romney (1988, pages 11 and 15), Quinlan (2005, pages 226 and 231), Bernard, Wutich and Ryan (2016, page 411) and Keddem and colleagues (2021).
A freelist is a floor, not a census
People forget things they know and skip what feels too obvious to say. What nobody listed is not proof it does not matter.
What we do: we keep the rarely-said items visible instead of cutting them and treat a list as a start, not the whole map.
The science: Brewer (2002) and Meireles and colleagues (2021).
What comes to mind is not the same as what matters
Listing something first or often can mean it came to mind quickly, not that it is most important. The two usually line up, but not always.
What we do: we read order as a strong hint worth examining, not as a final ranking. We check it against what people say.
The science: Smith and Borgatti (1997) and Chaves and colleagues (2019).
Grouping answers is a judgment call
Deciding that two differently-worded answers mean the same thing changes the result. Lump too much or too little and the picture shifts.
What we do: we show our grouping, let you open any item to read the original words and let you re-group anything that does not fit.
The science: Keddem and colleagues (2021).
The question shapes the list
A broad prompt gives a scattered list that misses things. A narrow, well-aimed prompt gives a usable one.
What we do: we help you write one tight, plain prompt and we can chain a second list off the first when the topic is large.
The science: Quinlan (2005) and Ryan, Nolan and Yoder (2000).
How and where you ask matters
Lists spoken with other people present, reading level, language and even age all change what gets said.
What we do: we let people answer privately in their own words, whether typing or speaking. We never suggest an answer for them.
The science: Quinlan (2005) and Meireles and colleagues (2021).
Optional
Going deep with math and statistics.
When you want numbers to stand behind the words, more phases follow. A freelist on its own cannot feed the cultural consensus model. Weller gives the reason on page 351 of her 2007 guide: that model needs one answer per question, where a freelist gives many answers at once. A freelist still gets an analysis of its own, which she names on that same page. The deeper track turns those items into a short structured task, asks a new group to complete it and then checks the answers with Cultural Consensus Theory. Most projects do well with the freelist alone and nothing here pushes you deeper besides your own research needs.