Beyond Affordability

All corpus · 52 sessions, 963 minutes · regenerated 2026-08-28 against the fixed values classifier and the fixed report prompt (no more starter lines for non-persuadable domains) · every number below pulled live from BigQuery, never generated

The question this run was asked: beyond affordability as the default message, what else should a candidate focus on? Each row below is one topic ("domain") scored on its own, not blended into one corpus-wide average — a candidate can be locked out on one issue and wide open on another, and a single number would hide that. The narrative (the read on each domain, the recommended frames) is an AI read constrained to only what's actually in the transcripts — nothing invented.

How to use this report

Status / Readiness answers whether to engage a domain at all. OFF-LIMITS isn't a smaller opportunity than CONDITIONAL — it's a different finding: the room is locked, not conflicted, and no message moves it today.

Starter line is a real, usable opening — except on an OFF-LIMITS domain matched to Core MAGA, where you won't find one. That's deliberate, not a gap: there's nothing to open with on a cluster explicitly classified as off the persuasion axis, so this report doesn't invent one.

Do / Don't stays populated even for OFF-LIMITS domains — it's harm-reduction guidance (how not to make it worse if the topic comes up), not a persuasion tactic.

Values are grounded differently from villains — real participant speech matched against a moral-value lexicon, with a session count and a verbatim quote, not an entity's leftover sentiment score.

Who's actually in the room

Every domain below maps to one of four real, poll-tested audience profiles. Defined once here so the jargon doesn't repeat eight times below.

Core MAGA

Locked in, identity-committed, distrustful of the other side. Not conflicted — just certain. Don't try to move them.

Diverse Disaffected

Mad at everyone — both parties, low trust across the board. Not loyal to a side, which makes them reachable if you name what's actually broken.

Tuned Out Persuadable

Not ideological — just stressed and checked out. Reachable with a concrete, near-term fix. No lecture, no labels.

Anti-Woke Traditionalist

Pro-work, allergic to moral lectures and activist language. Reachable through fairness and respect for work, not identity framing.

Domain snapshot

010254570→100

OFF-LIMITSLOW RETURNCONDITIONALHIGH VALUEPRIMARY TARGET

This is the field, 0–100 — the five bands above (OFF-LIMITS through PRIMARY TARGET) are the lines marked on it. Each domain below is plotted twice against that same field: Readiness (how open it is right now) and Sentiment (which way the room is leaning, -0.6 to +0.6).

How to use this table

Every domain in this corpus lands in the bottom third of the field — nothing clears CONDITIONAL by much. Worth knowing before reading bar-by-bar: there's no PRIMARY TARGET or HIGH VALUE domain to find here yet.

Read the two bars together, not separately. A short Readiness bar paired with a long, red Sentiment bar (Economy, Law/free speech) means loud and locked — the angriest domain in the room is not automatically the best opening; here it's the opposite.

A longer, green Readiness bar paired with a short, near-center Sentiment bar (Elections/process) means calm and open. That combination — not raw volume — is what's actually worth chasing.

Coverage matters as much as the score: a single-session domain (Health) can point the same direction as a solid one, but it's a hint, not a finding, until more sessions land.

Domain	Readiness	Sentiment	Speaking to
Economy / cost of living11 sessions · solid coverage	8/100 · OFF-LIMITS	-0.416	Core MAGA
Law / free speech2 sessions · thin coverage ⚠	8/100 · OFF-LIMITS	-0.511	Core MAGA
Partisan identity10 sessions · solid coverage	19/100 · LOW RETURN	-0.217	Diverse Disaffected
Immigration4 sessions · solid coverage	21/100 · LOW RETURN	-0.164	Diverse Disaffected
Foreign policy / war16 sessions · solid coverage	24/100 · LOW RETURN	-0.253	Diverse Disaffected
Elections / process5 sessions · solid coverage	27/100 · CONDITIONAL	+0.08	Tuned Out Persuadable
Health1 session · very thin ⚠⚠	33/100 · CONDITIONAL	-0.352	Anti-Woke Traditionalist
Trump / administration3 sessions · solid coverage	34/100 · CONDITIONAL	-0.119	Tuned Out Persuadable

⚠ thin data, ⚠⚠ very thin — hold that row's read loosely. Sentiment bars: red = negative, green = positive, centered on zero.

Domain by domain

Economy / cost of livingOFF-LIMITS · 8/100

11 sessions · solid coverage · Core MAGA

Loudest, angriest domain in the whole corpus — and locked. This isn't a conflicted room waiting to be reached; it's a certain one.

⛔ Do not engage. Non-persuadable.

No starter line, no frame — there's nothing to open with on a cluster this locked. What follows is harm-reduction guidance, not a pitch.

Stay concrete and respectful
Talk corruption + costs
Ask questions, don't preach
Prioritize harm-reduction asks

Don't

Argue facts as a first move
Trigger cultural identity conflict
Moralize

Law / free speechOFF-LIMITS · 8/100

2 sessions · thin coverage ⚠ · Core MAGA

Same profile as Economy, on thin data — only 2 sessions. Directionally consistent, but hold it loosely.

⛔ Do not engage. Non-persuadable.

Same reasoning as Economy — a locked cluster, not a conflicted one. No starter line for the same reason.

Focus on concrete issues, harm reduction

Don't

Argue facts
Trigger cultural identity conflict

Partisan identityLOW RETURN · 19/100

10 sessions · solid coverage · Diverse Disaffected

Mad at both parties, not attached to either. Not the same as Economy's certainty — this is an opening, if you name what's actually broken instead of picking a side.

“Both parties talk — but prices still climb. That's not an accident.”

Start with cost-of-living + fairness
Name corporate profiteering
Offer a specific, local win

Don't

Assume partisan identity
Lead with ideology

ImmigrationLOW RETURN · 21/100

4 sessions · solid coverage · Diverse Disaffected

Same profile as Partisan identity — frustrated with everyone, not partisan-loyal.

“Regular people need a team that actually fights back.”

Start with cost-of-living + fairness
Name corporate profiteering
Offer a specific, local win

Don't

Assume partisan identity
Lead with ideology

Foreign policy / warLOW RETURN · 24/100

16 sessions · solid coverage · Diverse Disaffected

The biggest domain by volume in the whole corpus (16 of 52 sessions) — and it's the same "mad at everyone" profile, not a fringe read.

“Both parties talk — but prices still climb. That's not an accident.”

Start with cost-of-living + fairness
Name corporate profiteering
Offer a specific, local win

Don't

Assume partisan identity
Lead with ideology

Elections / processCONDITIONAL · 27/100

5 sessions · solid coverage · Tuned Out Persuadable

The calmest domain in the corpus — near-neutral sentiment. Not because nobody cares; because this room is checked out, not entrenched. That's the best opening on this page.

“Feels like groceries and rent jump every month, right?”

Start with bills/rent
Short causal story
Name corporate price-gouging
Offer a near-term fix

Don't

Lead with labels
Shame voting history
Abstract system-talk
Assume policy fluency

HealthCONDITIONAL · 33/100

1 session · very thin coverage ⚠⚠ · Anti-Woke Traditionalist

One session. Treat this as a hint, not a finding — not enough data to act on alone.

“If you work hard, you should be able to afford a decent life.”

Lead with respect for work
Name greedy corporations + corrupt politicians
Talk fairness and rules

Don't

Lead with identity-first frames
Sound like scolding
Use activist jargon

Trump / administrationCONDITIONAL · 34/100

3 sessions · solid coverage · Tuned Out Persuadable

Same checked-out, cost-stressed profile as Elections — the highest-scoring domain in the corpus, for the same reason: low attention, not high certainty.

“It shouldn't take two jobs to stay afloat.”

Start with bills/rent
Short causal story
Name corporate price-gouging
Offer a near-term fix

Don't

Lead with labels
Abstract system-talk
Assume policy fluency

Recommended frames

How to use this table

Match a frame to the audience in its own row — a line that lands with Diverse Disaffected doesn't automatically transfer to Anti-Woke Traditionalist. These are matched pairs, not a menu to mix.

This is what to say once you know who you're talking to, not the order to say it in. For the sequence — when to open, when to escalate — that's Terrain, not this table.

Every frame below is for an engageable cluster only. Core MAGA (Economy, Law/free speech) is deliberately absent — see "Domain by domain" for why.

Frame	Who it's for	Starter line
Fairness / broken rules — corporations and politicians rigging the game	Diverse Disaffected, Anti-Woke Traditionalist	“Both parties talk — but prices still climb. That's not an accident.”
Concrete, near-term needs over abstract promises	Tuned Out Persuadable	“Feels like groceries and rent jump every month, right?”
Corruption and fairness, for audiences distrustful of institutions	Diverse Disaffected	“Too many politicians take corporate money and forget regular people.”
Respect for work / afford a decent life through hard work	Anti-Woke Traditionalist	“If you work hard, you should be able to afford a decent life.”

Risks — what not to do, and where

How to use this table

Each row is tied to a specific domain and a specific named WFP failure mode — not a universal warning. "Culture-war baiting" is the trap in Partisan identity/Immigration specifically; it doesn't mean every domain shares that risk.

Match the row to the domain you're actually working before applying it — the "Any domain" row is the one real exception, worth keeping in mind regardless of which domain you're in.

Domain	Don't	Failure mode
Law / free speech	Engage directly on facts	Escalation first
Partisan identity, Immigration	Lead with ideological frames	Culture-war baiting
Elections / process	Abstract system-talk	Abstraction without agents
Health	Shame voting history, moral lectures	Shame-based scolding
Any domain	Assume partisan identity or policy fluency	Premature moral sorting / Policy dump

Named villains & values

How to use this section

Villains are named things people blame — Watson NLU entities, ranked by relevance and sentiment, not by how many sessions they recur in. Most chips below are real but thin (1 session) — check the note under each name before treating one as a corpus-wide theme, not just a strong single moment. Trump (13+ sessions) is the one real exception.

Values are a different, broader kind of signal — a fixed moral-value lexicon matched against real participant speech, with an actual session count and a verbatim quote behind every row. Don't read the two the same way: a villain chip can be one loud moment; a value row is measured across the corpus.

Villains

Trumpnegative, high relevance · sadness 0.20

Bidenstrong negative · sadness 0.27

MAGA Republicansnegative, high relevance · anger 0.51

Kimmelnegative, high relevance · sadness 0.29

Hakeem Jeffriesstrong negative · disgust 0.24

Raystrong negative · sadness 0.55

George Bushstrong negative · sadness 0.26

Pete Hegsethnegative, high relevance · disgust 0.16

UCMJnegative, high relevance · anger 0.42

"people"negative, high spread · sadness 0.28

WhatsAppnegative, high relevance · joy 0.70 (odd pairing — real classifier output, flagged as-is)

Democratsnegative, high spread · sadness 0.30

Values

Value	Valence	Real signal	What people are actually saying
Honesty	violated	22 sessions / 35 mentions	People describe others as lying and not being truthful, leading to distrust in information.
Security	contested	21 sessions / 42 mentions	Mostly Social Security's long-term solvency — the trust fund running out, need for new funding or tax increases.
Trust	contested	17 sessions / 51 mentions	Concern the Social Security trust fund can't pay benefits; also distrust of the administration on military/intelligence matters.
Accountability	contested	15 sessions / 41 mentions	Whether anyone will actually bear responsibility for problems like Social Security shortfalls.
Fairness	contested	13 sessions / 20 mentions	Whether the public suffered unfairly from political conflict; some tie this to immigration and the economy.
Community	contested	11 sessions / 20 mentions	Deferring to people who've served their communities; risk to military/intelligence professionals from the administration.
Corruption	violated	7 sessions / 8 mentions	Disappointment that corruption in government benefits certain people at real economic cost to everyone else.
Freedom	contested	5 sessions / 11 mentions	Wanting global AI regulation while still protecting freedom; general respect for liberty as an ideal.

"Contested" means real evidence pulled both ways in the sample — not a hedge, a measured finding. This replaces an earlier version of this section (2026-08-24) where every value traced to exactly one session, one already-listed villain — fixed 2026-08-28 (bq/classify.py's _value_evidence).

Data appendix

The exact source table behind every number above — All corpus, 52 sessions, 963 minutes.

Domain	Status	Readiness	Sentiment	Emotional openness	Confidence	Sessions	Coverage
Economy / cost of living	OFF-LIMITS	8/100	-0.416	0.312	medium	11	solid
Law / free speech	OFF-LIMITS	8/100	-0.511	0.222	high	2	thin ⚠
Partisan identity	LOW RETURN	19/100	-0.217	0.257	high	10	solid
Immigration	LOW RETURN	21/100	-0.164	0.244	medium	4	solid
Foreign policy / war	LOW RETURN	24/100	-0.253	0.270	high	16	solid
Elections / process	CONDITIONAL	27/100	0.080	0.248	high	5	solid
Health	CONDITIONAL	33/100	-0.352	0.345	low	1	very thin ⚠⚠
Trump / administration	CONDITIONAL	34/100	-0.119	0.295	medium	3	solid