Dialogue Pulse · Episode 3 · Two waves, 347 interviews
We asked 191 strangers one open question about LeBron James and offered them no options to choose from. They described six different men. Four weeks later, a week after he signed with Philadelphia, we asked 156 different strangers the same question and the same six men came back. The largest of them held inside a point, and the rule that sorts people between them reproduced within four points on every cell. What did move, by nine points, was how they felt.
Every particle in the field is one interview, 347 in all, coloured by the frame that participant reached for. The field settles into wave one's shape, dissolves, and re-forms as wave two. The two waves are independent samples, so no particle represents a person changing their mind; what repeats is the shape of the distribution.
Start here: pick a name. Each of the six is a pattern people arrived at on their own, unprompted, illustrated below in a participant's own words.
The high-level takeaways
Evidence for each is in the sections that follow. These are the mechanisms underneath the numbers, not a summary of them.
Three words carry most of the weight below, so quickly. Sentiment is how people feel about you. A frame is the story they use to explain you. Positioning is which of those stories your market is running, and in what proportion. The next section takes all four apart, including the one people forget.
A note on terms
These words get used interchangeably and they should not be. In plain terms: sentiment is how people feel about you, a frame is the story they use to explain you, positioning is which of those stories your market is running and in what proportion, and salience is whether you are in the conversation at all. The distinction is the whole reason this study produced a result, because three of the four held still while one moved nine points. A tracker that measures only the moving one would have reported a transformation that did not happen.
One clarification, because it carries every positioning number on this page. We did not score positioning on rating scales or attribute grids. We measured it as the mix of frames across the population. So when we quote a number like 0.8 points, that is one frame's share of the population changing, not a rating on a scale. The mix of stories barely changed at the top. The feeling inside those stories changed a great deal.
Moved 9.4 points across our two waves. Wistfulness roughly halved, personal grievance roughly doubled. Sentiment is real, but it is weather, not climate: it responds to events within days and decays within weeks.
The oldest and cheapest measure in the business: valence, positive to negative, sometimes with an intensity dial. Every social listening tool ships it by default because it is the easiest thing to automate.
Runs across all three below, on its own clock.
We operationalised positioning as the distribution of frames across the population, rather than as ratings on axes we chose in advance. The advantage is that the axes are the respondents' own. The cost is that it takes conversation rather than a grid, which is precisely the tradeoff this instrument exists to change.
Ries and Trout's formulation in Positioning (1981) still holds: positioning is not what you do to a product, it is what you do to the mind of the prospect. Conventionally it is measured with attribute batteries and perceptual maps, which impose the researcher's dimensions on the respondent.
The Tenant and The Businessman are not opinions about LeBron. They are competing interpretive templates applied to an identical set of facts. Two people can hold the same sentiment and be running completely different frames, which is why sentiment alone tells you almost nothing about what someone will do next.
Goffman called these schemata of interpretation in Frame Analysis (1974): the mental templates that let a person locate, perceive, identify and label what is happening. Entman (1993) sharpened it for communications: framing is selection plus salience, promoting one problem definition, one causal story, one moral evaluation over the others available.
Positioning is the aggregate of frames, not a separate thing measured on separate axes.
The precondition beneath all three
51 of 156 respondents said they had encountered nothing about a story with unlimited free coverage, and reading all 51 transcripts confirmed 46 of them. For those people there was no stored frame to retrieve, so they built one on the spot from whatever arrived with the news. Salience is upstream of everything else on this page.
The prior condition beneath both frame and positioning. In the awareness literature it is the difference between recognition, prompted recall, and unaided recall, and the three are routinely conflated in reporting.
The practical consequence. The four run on different clocks, so no one number can stand in for the others. Sentiment responds to an event within days and decays within weeks. A frame is a template a person keeps for years. Positioning is the population-level mix of those templates, which makes it the slowest thing here to move and the most expensive to change. Salience sits upstream of all three, because someone carrying no stored frame will build one out of whatever arrives with the news.
Most brand trackers measure sentiment continuously, positioning once a year on attribute batteries the researcher wrote, and salience almost never. That combination is close to the worst available. It samples the fastest-decaying signal most often, measures the durable one on axes the respondent did not choose, and ignores the precondition entirely.
Here is the test, using our own two numbers. Sentiment moved 9.4 points. The largest frame moved 0.8. Read the first alone and you would report that the market turned on him. Read the second alone and you would report that nothing happened. Both are accurate, and they point to opposite decisions: feeling that moves and reverts is a communications problem, a mix of stories that will not move is a product and distribution problem. If your tracker moved this quarter, find out which of the four moved before anyone rewrites a strategy.
Act one · What the signing moved
Four weeks, a surprise destination, wall-to-wall coverage. The largest frame moved less than one point.
The finding
Wave 1 established that audience segmentation here runs on information state rather than fandom. Respondents who already knew he was leaving converge on The Tenant. Respondents who learned it during the interview converge on The Famous Name.
Wave 2 re-ran the same instrument on 156 different respondents four weeks later, one week after the signing. The split reproduced within four points on every cell.
The split
Information state, not how much someone cares, decides which story you get. People who already knew he was leaving land on The Tenant; people who found out from us land on The Famous Name.
The Tenant moves one point among the informed and one point among the uninformed. The Famous Name moves four and three. A different set of people, after the biggest news of his career, reproduced the same split.
Wave 1 also ruled out the obvious objection, that this is just fandom: inside engaged fans alone, those who learned the news in the interview chose The Famous Name over The Tenant by 12 to 4, while engaged fans who already knew chose The Tenant by 30 to 2. What sorts people is not how much they care. It is whether they were carrying the story.
Wave 1 argued the uninformed half was the part of a market still open to being written. It has narrowed and not closed: 41% of wave 1 with clear awareness met the story in the interview, against 35% in wave 2. Directionally correct, not resolved.
Why the frame is the commercial variable here, and not the sentiment. A frame behaves like a decision rule. It governs which arguments get heard and which bounce off, because new information tends to get fitted into the story a person is already running rather than replacing it. That turns the table above into a budget question rather than a curiosity.
For the informed half, the story is already filed. Spend aimed at that half moves how they feel about a verdict they have already reached, which is precisely the pattern in our two numbers: feeling moved nine points while the largest story held inside a point. For the uninformed half there is no stored story to overwrite, so the first frame to arrive alongside the news becomes the frame, and arriving first is cheap.
The uncomfortable version for a brand: the part of your market that knows you best is the part your messaging can least reposition, and the part that is cheapest to write on is the part your targeting is least likely to reach, because optimisation pushes spend toward people who already engage.
One limit, stated plainly. This measures the story people tell, not what they bought. It tells you what will be heard, not what it earns.
What held
Wave 1 found that people keep two ledgers. Greatness travels with the person; belonging stays with the jersey, and belonging is barred by a rule participants state unprompted, in near-identical terms: legends don’t leave their team. If that is the real structure, then no destination could have moved it, because the verdict was never about where he went next.
Wave 2 is the test. Same instrument, sections byte-identical, matched fielding window, a fresh sample of different people. Below, the hollow ring is where each frame stood before he signed and the solid dot is where it stands now. The gap between them is the entire effect of the biggest news in his career.
What held
The signing barely touched the mix of stories. The largest frame moved +0.8 points, and four of the six named frames moved by under 1.3. Two moved more: The Famous Name, which tracks interview length rather than a real migration, and The Father, on a base of five respondents.
One thing did move, and we had predicted it in print. Wave 1 named Philadelphia as one of two ring-chase destinations and argued the destination would decide which exit script got activated, naming the mercenary reading as one of three candidates. The frame shares above barely moved, but a frame's share is not the only thing that can change: the register inside a frame can harden. So we went back and coded every interview in both waves, under one written definition, for the mercenary register specifically: money or ring-collection as the operative explanation, delivered with an edge. Not the no-fault version, which is The Businessman working as designed, but the sneer.
The vocabulary sharpened too: wave 1 mostly hedged (“probably a money thing”); wave 2 says serial ring chaser, rental player, buys championships, waste of money.
The surprise is where it lives. Not one of the 18 is coded The Businessman; 15 are Tenants. The Businessman is defined by the absence of anger, so anyone running money logic with contempt reads, correctly, as denying that he ever belonged, and lands in The Tenant. The mercenary register is the Businessman’s ledger delivered as the Tenant’s verdict — a bridge between the two frames, which is exactly why no frame count ever surfaced it.
It is not a seventh frame, yet. We are tracking it as a named register inside the six. The wave-3 test is written down now, before the season answers it: if Philadelphia starts well and the explicit rate keeps climbing, the register earns its own frame, and because it lives inside The Tenant, promoting it would re-carve this page’s largest frame, not its money frame. A ring would supply the material outright.
Method: every interview in both waves (347 people), coded from full interview records against a definition written before the sweep. Every explicit case was verified against the audited verbatim bank or the source transcript, and anything that appeared only as agreement with a statement our moderator read aloud was excluded. The count is conservative: three apparent cases were removed when the source transcript did not support the summary that claimed them.
Every one of these respondents is plotted individually, with their verified words, in the map below.
Setting the stage
On 30 June 2026 LeBron James's agent announced he would not re-sign with the Lakers, ending eight seasons and one title in Los Angeles. No destination was known. On 24 July he signed with the Philadelphia 76ers, two years at roughly $8M after earning about $53M, and called it his last decision.
Two waves of AI-moderated voice interviews on the same frozen instrument. Wave 1: 191 interviews fielded 3 July, entirely inside the pre-destination window. Wave 2: 156 interviews fielded 31 July and 2 August, after the signing, on a fresh matched sample with no re-contact. Three geographic rings: LA metro, California outside LA, and the rest of the United States.
Four questions: (1) does a destination change what a market has concluded about a person; (2) does the information-state split found in wave 1 replicate on a new sample; (3) which of wave 1's published forecasts survive contact with the real event; (4) what a market that missed the news entirely does when it meets it cold.
What moved
Same instrument, same questions, coded instead for the register people used at the news moment. Here there is real movement, and it runs opposite to Act One.
Wave 1 was the limbo week and its colour was elegiac, all "end of an era." Once he actually landed somewhere, that did not survive. It resolved into either transactional acceptance or personal grievance. The middle emptied and both ends filled.
What moved
Here the movement is real. Wistfulness roughly halved and the share reporting first-person hurt roughly doubled across the same four weeks, on the same frozen instrument.
Implication: positioning and sentiment are separate assets on separate clocks. An event can move how a market feels without moving what it has concluded. Measuring only sentiment would have reported a large change where none occurred.
The same people, two ways
Every dot is one interview. Switch between the two constructs and the map redraws the same 136 people: first by the frame they reached for, then by the register they used at the news moment. The point is that these are different maps of identical respondents. Choose any dot for that person's verified words.
Two lenses
136 of the 156 wave-two interviews, coloured two ways. Switch the toggle and the pattern changes without a single dot moving, because the story someone reaches for and the feeling they attach to it are two different measurements of one person.
The California view groups respondents by the county each one selected in the screener, which is the residence question that assigned their cohort in the first place, and it is self-reported rather than inferred from panel geolocation. Each cluster sits at its county's real position in the state; a dot's spot inside a cluster carries no meaning, because county is the finest resolution the question captured. Three respondents chose “another county in California” and are shown offshore, since we know they are Californian but not where. Panel geolocation disagreed with the screener for roughly a third of the LA metro cohort, which is the reason this view trusts the answer rather than the location.
Every quote on this map was re-selected on 4 August after an audit of all 122 of them found that the displayed line often did not show the code beside it. The quotes had been chosen for being vivid rather than for displaying the frame. Where nothing in a participant’s audited verbatims displays their coded frame without stretching, the dot now shows no quote: 56 of 136 are blank for that reason or because the participant is quote-restricted. We also removed any line that named the moderator, opened by agreeing with him, or repeated one of the four statements he read aloud, since agreement with our own prompt is not evidence. Nothing here changes any number on this page.
A blank dot is not a dead end. Click one and it shows how that interview was coded instead: the frame, the feeling, the coder’s confidence, the grounds for the judgment, and which of the four reasons the quote is withheld. That text is ours, not theirs. The coders’ own notes quote the transcripts, but only 27 of the 103 fragments in them could be matched word for word against the audited bank, so every fragment was rewritten in our words — a dot exists in this state precisely because we have nothing from that person we can put in quotation marks. For the quote-restricted participants we characterise only their read on LeBron, never the personal disclosure that restricted them. Writing these up caught two dots that were blank by oversight rather than by rule: one now shows the quote it always should have, and the study-wide restricted count above was ten and is twelve.
Dots are placed within the respondent's state, not at their exact address, so the map shows distribution rather than precise location. 19 respondents whose panel location contradicted their own screener answer are excluded here, and one in Hawaii sits off the frame; all 20 remain in every other number on this page.
Act two · The rule underneath
Underneath all six frames sits one rule people apply without being asked: belonging is earned by staying.
Mechanism
Act one rests on a claim: that people apply a rule about belonging which LeBron structurally cannot satisfy. We never asked anyone about that rule. It arrives on its own, in almost identical language, from people with nothing else in common.
Belonging is earned by staying. Not by winning, not by giving a city eight years, not by being the best player ever to wear the jersey. By staying.
Six respondents in wave 2 volunteered some form of this standard in their own words. Two arrived before any comparison was offered; the others came while discussing Kobe, and the moderator supplied the word loyalty in one of those exchanges. It is the machinery under every other number on this page: a market running a tenure rule cannot be moved by performance, and a man who has played for four franchises can never satisfy it no matter where he signs next.
Cleveland, Miami, Cleveland, Los Angeles, and now Philadelphia, against twenty seasons in one uniform. Career tenure from the public record, not from this study; the final segment is unplayed. The rule participants state is about staying, and this is what staying looks like next to not staying.
“Kobe spent his whole career with one franchise. LeBron didn't.”
“I don't really consider LeBron a Laker. He jumps from team to team.”
“Kobe's more of sort of the fans' ideal version of a player, someone who picks one program and sticks with it for his entire career.”
“I like it when a player stays with one team his entire career. I admire that kind of loyalty because sometimes switching teams can bring them more money.”
“You had a chance to prove to LA that you are a Laker, and you took the opportunity to leave to Philadelphia... we don't want that. We want someone that's gonna stay loyal no matter what.”
“people were more attached to Kobe just because you were able to see him for his entire career.”
This is the machinery under every other number on this page: a folk rule about what earns a city's belonging, recited unprompted, that a man who has played for four franchises can never satisfy no matter where he signs next.
The awareness check
Every interview opened by asking what, if anything, people had heard lately about LeBron and the Lakers. A month after his agent announced he would not re-sign, and a week after he signed with Philadelphia, 51 of 156 said they were hearing it from us.
We did not trust that, so we read all 51 of those transcripts end to end rather than sampling them. Forty-six of the 51 held up: no prior exposure at all. Four had in fact encountered it, two of those knowing only that he had left and not where he went, and one case is unclear. Each of those four was then re-checked adversarially, because the first pass over-called exposure by counting unrelated LeBron news, golf-course comments and a Stephen A. Smith segment, as contact with this story.
The check
Free, saturating coverage still missed people. 51 said they had encountered nothing, and reading all 51 transcripts confirmed 46, while 63 already knew where he had signed.
The gap between what people said and what the transcripts showed is the finding. One had scrolled past it on Instagram. One volunteered that she had seen it but was "so focused on work" that it never landed. Asked directly, they reported no knowledge, and they were not lying: the item had reached their eyes without reaching them.
How it arrived
It arrived through people and feeds. Of 32 coded first-contact mentions, two were television, radio or a newspaper.
For those who did register it, how it arrived.
Friends, feeds, alerts, and a bet. Television, radio and newspapers barely register as the point of first contact for a story that ran on every sports network in the country.
Transfer
Two mechanisms here are not sport-specific and are carried by hard numbers in this study. Two others surfaced and are not shown: they were plausible readings we could not evidence separately, and a single signing event cannot carry them.
Act three · If you run research
Two findings here cost money in any category, and neither would have survived a closed-ended survey.
Method
Fifteen people in this study opened by telling us they had nothing to offer. They did not follow basketball, could not name an LA athlete, had no opinion worth having. Every one of them then produced some of the sharpest reasoning in their own transcript.
On the left is what a closed-ended instrument would have recorded for these four people. On the right is what they actually said, a few minutes later, in the same conversation.
What a screener misses
Fifteen people opened by telling us they had nothing to offer, then produced substantive reasoning anyway. These 4 are what a closed-ended instrument would have recorded instead.
These respondents are the ones a low-engagement screener removes and a rating scale flattens. Self-reported disengagement predicted nothing: eight others made the same disclaimer and had nothing to add. The signal is unrecoverable from a checkbox and recoverable in four minutes of conversation. If your screener drops non-category-users, quantify what that exclusion costs before defending it.
Scale
The scale
This is the scale line. Two waves and 111 hours of recorded talk, fielded in two afternoons, on one frozen instrument.
111 hours of primary qualitative data, fielded in two afternoons, at a fraction of the cost and cycle time of a conventional two-wave design. Coded, quoted, and verified against source.
The two-wave structure exists to make the first wave falsifiable. Wave 1 published a mechanism and a set of forecasts; wave 2 tested them against a real event on a fresh sample. The deliverable is not a finding. It is a finding with a check attached.
Two matched waves on one instrument. Sections 1–4 and 6 are frozen byte-identical across waves and were diffed before fielding; only the moderator's context line and one forward-looking section changed. Wave 1 n=191, wave 2 n=156. Voids removed; no quality filter applied to either wave, deliberately, so the comparison stays symmetrical. The two samples are different people by design, so nothing here describes anyone changing their mind.
A finding we cut. The raw national cut appeared to show a large migration, Tenant up about 17 points and Famous Name down about 15, a near mirror image. It is not real. Famous Name is defined by the absence of retrievable content, so it is mechanically sensitive to how much an interview extracts, and wave 2's interviews ran 36% longer. Holding interview depth constant, the migration collapses and then reverses. We found it, tested it, and removed it rather than publishing it.
Quotes. Every quote came from an audited bank: 437 candidates were checked line by line against source transcripts and 42 were rejected, including two that did not exist in the interview at all. Act four's quotes come from a later targeted re-mine of the signing sections, extracted under the same exactness rules. Nothing here is paraphrased, composited, or stitched across a moderator turn. Twelve participants are quote-restricted for volunteering sensitive personal information unrelated to the research question; they count toward every number and none of their words appear.
Limits. Qualitative and directional, not a poll. Los Angeles came in under our pre-registered floor in wave 2, and roughly a third of self-reported LA residents had panel locations elsewhere, so LA-specific cuts carry that caveat. Wave 1 sits about three days after the exit news and wave 2 about a week after the signing, so the comparison contains time-distance as well as the event.