66%
Coordination present
170/256
This showcase is private. Enter the password to continue.
“I liked the last one the best because it gives you the community aspect of being able to engage with your friends, but it doesn’t, um, risk your winnings.” P006 · 25-34 woman Denver · [12:34]
Kalshi asked one question. The team building it asked five more. We put the concept in front of 256 bettors and let them answer all six, then hand back the questions nobody thought to write down.
Five steps, streamlined. 256 AI-moderated interviews, sized and batched, coded one by one, normalized into one verified source of truth, pressure-tested by a council, then rendered here. Every number and quote traces back through this chain.
One script. Twenty-nine questions. Twelve sections. The same for all 256, all three cards shown in the same fixed order. The primary measure was the choice at question 19: of those three, which one, and why. Here is what came back, in the order it was asked.
Both, and they travel together. Coordination surfaced for 170 of 256. Communion for 154 of 256. A hundred people showed both at once. The either/or the study was built on is not how people hold it. The real fault line runs somewhere else: autonomy and control against shared fate.
Just Share It, the solo build, took the plurality at 110 of 256 (43%). Locked In was a strong second at 89 (35%). Pass the Slip trailed at 40 (16%). Seventeen rejected all three or picked none. The plan hoped for a solo landslide. It got a plurality and a loud refusing tail: 71 of 256 (28%) hold a hard never-mix-money-and-friends boundary.
Choice at Q19, within-subject, n=256.
On the exit binary the population divides almost evenly: locked until settlement 123 of 256 (48%), sell anytime 109 (43%). But cut it by experience and it tilts. Combo-fluent participants, the people who have actually bought a Kalshi combo, prefer sell-anytime 24 of 46 (52%) over locked 16 of 46 (35%). Locking is a retention risk for the users you already have.
Share of each group choosing sell-anytime vs locked-until-settlement.
The refusers reason arithmetically. Each of us can place it separately and each of us makes the full payout, so why pay extra to bundle it. That position is real, and it is a minority. The numbers correct the loudest voices in the transcript.
The card someone points at is not the thing they want. The same button pulls two opposite people. Here is the surprise, the proof you can click, and the edges the plan never budgeted for.
Locked In pulls a togetherness person and a control person toward the same button. One wants everyone in it together. The other wants a wall so nobody can touch their pick. They arrive at the same card from opposite motives. And some solo picks are not enthusiasm at all: they are confusion, or a forced least-bad choice. Among the 89 who picked Locked In, 28 still want a sell-anytime exit. Among the 110 who picked solo, 48 want a locked exit. The mechanic and the meaning cross-cut.
“either the second or third one because I have the choice and, um, w-well, nobody can take it from underneath... Well, the one where nobody can take it from underneath me.”
P233 · 45-54 woman Grandview · [13:12]
“No, this I would actually be okay with. Um, all four of us are putting our money in... and you can’t cash it out unless all four of us agree. So it would kind of be a consensus thing. That I could work with.”
P239 · 35-44 woman Oberlin · [13:45]
“The first one where I would buy and purchase on my own.”
P097 · 45-54 woman Baltimore · [12:42]
Each dot is one interview. Columns are the exit the person wanted. Rows are whether communion surfaced. Color is the card they chose. Click any dot for the exact quote.
The dots are packed inside each cell for density; position within a cell carries no meaning. Filter to Locked In and toggle the split: of the 89 who picked it, communion surfaced for 65 and stayed absent for 24. Same button, two different reasons.
Click a dot to read that person’s exact words.
The plan feared shared fate would produce a villain, not a bond. It mostly did not. Friendships ended in 2% of cases, 5 of 256. Blame was absorbed in 159 of 256 (62%) and joked off in 144 (56%). Others’ picks costing you is recalled more vividly (96) than your own pick costing someone else (43), but own-blame is far from empty: 130 had an answer.
Neither of these was on the script. Both came up on their own, unprompted, often enough to note. We are not sizing them; we heard them.
The same person is warm about sharing a few dollars and private about the big play. Group fun money is social. The real book stays solo. We heard this as a say-do line drawn inside a single participant, not across the sample.
Every sample has voices at the far ends. These are exact.
“I don’t think I would ever do anything like this with friends. I, I know not to mess with friends and money and stuff like that. That just complicates friendships. I’m never gonna mix money and friends.”
P002 · 25-34 man Bernalillo · [11:08]
“I would use the first one because I think all three are mechanically the same, but the first one is the clearest about what’s going on, which, uh, reduces the chances for, uh, conflict and people getting upset.”
P077 · 35-44 man Omaha · [11:44]
“Equality is everything to me. I want my friends to feel like they can trust me. I want everyone to feel like they can trust me.”
P029 · 18-24 other Monroe
“’Cause then you can rub it into your friends.”
P213 · 55-64 man Little Egg Harbor Township · [16:39]
One linear script, twelve sections, twenty-nine questions, identical for everyone, no branching. All three concept cards shown in the same fixed escalating order. Tap any section to open it.
256 participants coded, August 2026. Panel-recruited, AI-moderated, structured items inside the transcript. Planned N was 300; we fielded 256, roughly 15% short. The PRIVATE cell undershot badly: 34 against a floor of 70, so anything cut on privacy is underpowered.
One cell per interview, colored by quality band, sorted by IQS from best to worst.
IQS mean 3.83/5, with 76% strong or above (195/256). Two interviews dead, two authenticity-capped. Verbatims verified exact against source. Anonymized to P## plus age, gender, and city.