# Round 4 Synthesis - Concept Reveal: Every Pitch Died, Every Persona Rebuilt It (9 interviews + 3 retirement checks, simulated)

*Study lead synthesis, 2026-08-07. Simulated personas: every number below is a
directional prior, not validation. Round 4 of max 5. Not converged.*

## Headline finding

The first concept contact of the study produced zero acceptances of any wedge
as pitched and five spontaneous redesigns, and the redesigns agree with each
other more than the pitches did. Every persona with real pain took the concept
apart and handed back a narrower, cheaper, structurally different product:
Dana rebuilt the inspector network into escrow-released pay-after with a
visible "what he caught" record; Tom stripped lead verification down to a
single silent question, has this print been shopped; Kelly split the
visibility product away from commission pricing entirely because "eight
percent of nothing is nothing"; Priya reframed the verification network as a
registry-checked desk pass that tells Deepak where to point the car; Jenny
moved the monitoring flag from during the delay to before the queue bump and
added an invisibility bar no vendor may be able to prove; Derek converted a
monitoring subscription into one deposition-grade report at month 11; Aisha
converted a standalone product into a Beacon Ridge branded module. Meanwhile
the round's predicted failure landed exactly on schedule: Dana wired $5,500
to Grupo Herrera unverified because both available Monterrey inspectors
demanded 50% upfront, the single cleanest proof in four rounds that the
binding constraint is trust structure, not price, not evidence format, not
discovery. The common thread across all seven redesigns: trust never
transfers through a platform credential. It transfers through structure
(pay-after, escrow, shared stake, an incumbent's letterhead, provable
invisibility), and any product that asks the buyer to take the leap of faith
first is re-running Canton, Dongguan, and the two inspectors Dana refused.

## Scorecard

| # | Persona | Wedge revealed | Fit | Pain | New $ this round | Verdict on the concept as pitched |
|---|---|---|---|---|---|---|
| 01 | Dana, DTC hardware sourcing lead | W1 network | strong | 4/5 | $5,500 wired unverified July 8 (outcome pending late Aug) | Rejected as pitched; rebuilt around escrow + catch record |
| 02 | Marcus, OEM procurement | W1 network | medium | 3/5 | ~$1,100 Dave floor trip + $900 plater report; >$6,500 YTD | Describable but untested against the Dave gate; shared stake is his only precedent |
| 03 | Rita, DTC control (retired) | none | weak | 1/5 | none | Control closed clean; declined the reveal herself |
| 04 | Tom, 63, CNC shop owner | W1-flavored lead check | medium | 4/5 | none new; $650 Vandermeer anchor | Rejected "real and solvent"; redesigned to shopped-print check only |
| 05 | Kelly, 34, fab shop successor | W5 quote visibility | strong | 4/5 | Rio Grande $28k still unresolved | Wants it; commission pricing structurally cannot pay for it |
| 06 | Hank, foundry sales (retired) | none | no | 3/5 | $10,830 cumulative since April | Denise heard the real number and said "not now"; no wedge at any price |
| 07 | Priya, OEM sourcing engineer | W1 desk pass | medium | 4/5 | $1,250 second Deepak trip (realized) | Closest hit of the round; capped at desk-pass-only, registry-verified, criticality-tiered |
| 08 | Sal, rep control (retired) | none | no | 2/5 | none (no interview; retirement note) | Control carried forward unchanged |
| 09 | Jenny, trading co owner | W4-adjacent monitoring | medium | 4/5 | none new; $80 goodwill discount near-miss | Right gap, wrong timing window; invisibility bar may be unclearable |
| 10 | Derek, search fund CEO | W4 post-close monitoring | medium | 3/5 | Higgins closed, $46,800 transaction cost; $150k holdback live | Subscription dead; redesigned to one $2-3k month-11 report |
| 11 | Aisha, PE roll-up associate | W4 embedded module | medium | 3/5 | $200 own-pocket people-search; Vantage $2.1M/yr tie unverified | Would champion it, but only under an incumbent's name |
| 12 | Greg, EMS control (retired) | W1 (pressure test) | no | 1/5 | none | No cracked under direct concept pressure; QE liability cannot be outsourced |

Roster shape check: 2 strong, 6 medium, 1 weak, 3 no. Four retirements
executed as designed (Rita, Sal, Greg, Hank), and the reveal did not rescue
a single one of them. The study is still not skewing positive: the reveal
round subtracted product shapes rather than adding enthusiasm.

## Cross-cutting patterns

**1. The pitch dies, the persona's redesign survives. Seven for seven.**
Nobody accepted a concept as described, and nobody rejected the underlying
job. Every active persona rebuilt the product mid-interview into something
smaller, with a sharper trigger, priced against a specific dollar figure they
already know. The R5 products are the personas' versions, not ours.
> "Don't sell me 'are they real.' Sell me 'has this print been shopped.'
> That's the only sentence in your whole pitch that would've saved me six
> hundred fifty bucks." - Tom

**2. Trust does not bootstrap from a credential; it transfers only through
structure or stake.** Dana demands escrow and a visible catch record ("a
caught, not a testimonial"). Marcus's single precedent for accepting a
stranger's floor eyes was an auditor whose employer had money in the same
outcome. Aisha will only put a finding in front of IC if a deposable firm
signed it. The platform badge is worth zero; the payment structure and the
name on the hook are the product.
> "The one time anybody's eyes but Dave's ever cleared a supplier for us, it
> was a customer's own auditor who happened to also buy from that supplier.
> He wasn't a stranger with a badge, he had money in the same outcome we
> did." - Marcus
> "If some company I've never heard of offered me the identical report for
> less money, I wouldn't put it in front of IC, because the day it's wrong,
> I need a name that can sit in a deposition chair, not a PDF." - Aisha

**3. The predicted failure happened on schedule, and it prices the trust gap
at $5,500.** R3 flagged that Dana would either book a Monterrey inspector or
wire blind. She wired blind, eyes open, told her ops director out loud as she
did it. Not because verification was unavailable or unaffordable ($600-750),
but because the available verifiers' trust structure (50% upfront from a
stranger) was worse than the factory's. This is the study's cleanest
behavioral proof that pay-after is not a preference, it is the gate.
> "I wired fifty-five hundred dollars to a factory nobody has ever laid eyes
> on for me. I knew exactly what I was doing when I did it, that's the part
> that bugs me." - Dana

**4. Timing beats intensity: a flag inside the risk window is worth a
fraction of a flag before it.** Jenny reconstructed the Ningbo timeline
herself: a flag during the eleven-day delay saves maybe $300 of freight
timing; a flag when the bigger order landed, weeks earlier, is worth the
whole $4,050. Derek generalized it: a monthly memo nobody owns reading trains
its reader to stop opening it; only event-triggered signals get acted on.
Kelly added the third leg: a flag without odds attached "is closer to a
feeling than a decision."
> "If your flag only fires during my eleven days, you've built a
> slightly-faster way to find out I'm already screwed." - Jenny
> "A monthly memo with nothing in it eight months running trains me to stop
> opening it. That's what happened to my own Google News digest before
> close, and I built that myself." - Derek

**5. The shopped-print check crystallized from both sides of the table, and
it now has a named mechanism problem.** Tom made it the entire product and
attached it to a per-referral trigger and his $650 anchor. Kelly wants it on
every quote but immediately asked where the data comes from: no shop
volunteers its own pipeline, so the signal either needs reciprocal
contribution (which she might do, "but that's a real ask, not a free lunch")
or it does not exist. Hank's exit interview added the failure case: Deb now
asks the shopped-print question on every call and a prospect simply lied,
costing $1,900 anyway. Self-reported answers do not close this gap; only
cross-shop data does, and cross-shop data has a cold-start wall.
> "I don't know how you'd know a print's been shopped unless the other shops
> are telling you, and why would a shop in Laredo tell some outside service
> what's crossing their desk?" - Kelly

**6. Avoidance value cannot be commission-priced, which splits W5 into two
products with two bills.** Kelly, the only persona who ever validated
commission pricing, ruled it out herself for the visibility product: the
value lands on quotes she walks away from, where there is no revenue to take
a cut of. Information products anchor to her Bluebeam precedent ($240/yr,
does one job well), roughly 20x below her standing commission structure.
> "If the flag says pass, what am I paying you eight percent of? There's no
> job to take a cut of. That math doesn't work for this part." - Kelly

**7. Onboarding friction is a hard numeric gate, now quantified.** Marcus set
it at one phone call plus one email, about twenty minutes, before he defaults
back to the backlogged incumbent, exactly how Dayton died in ninety seconds.
Any intake questionnaire, portal account, MSA, or NDA review is a kill. This
compounds pattern 2: the product must be trusted like an incumbent AND
frictionless like a phone call, on first contact, simultaneously.
> "You're not competing against nothing, you're competing against a phone
> call to a guy who already picks up." - Marcus

## Vetoes and dealbreakers (added or hardened this round)

- Upfront payment from a stranger verifier is now behaviorally proven fatal,
  not just stated: it cost the study's strongest-fit persona $5,500 of
  unverified exposure. Escrow or pay-after is table stakes (Dana).
- One wrong answer on the exact question the product was bought for ends the
  relationship permanently: a cleared prospect who ghosts (Tom), a false
  "no outside processing" (Priya), a cleared factory with a Dongguan defect
  rate (Dana), two noisy false alarms (Jenny). The tolerance is one, and for
  Priya's specific miss it is zero.
- Provable invisibility to the watched party is a new structural bar:
  Jenny's factory owner must never learn the watch list exists, and she
  doubts any vendor can prove that; Derek would not even run his month-11
  check until Ray's transition ends. Surveillance products carry a
  relationship-detonation risk the buyer prices higher than the loss.
- No-name evidence in legal contexts is worse than nothing: it "looks like
  diligence happened when it actually didn't hold up" (Aisha). Deposition
  survivability requires a named, credentialed, insured signer.
- Generic instruments are auto-rejected: not Dave's checklist verbatim means
  Dave already said no (Marcus); one-size-fits-all criticality means the
  builder "doesn't understand how a quality org signs off" (Priya).
- Scheduled-cadence deliverables die unread; only event-triggered flags with
  the dollar consequence attached get acted on (Derek).
- Third-party evidence cannot discharge personal liability: Greg's QE signs
  the AS9100 finding herself or the customer's supplier-quality team rejects
  the package. Process-gated orgs stay structurally closed (control, final).
- Standing vetoes hold: vendor self-reported anything, agents speaking in
  the user's name, commission to the Xometry-burned, subscriptions to the
  ThomasNet-burned.

## WTP read (with required skepticism)

Simulated WTP is a directional prior at best, and this round's numbers are
the most contaminated of the study: all were volunteered minutes after a
concept description, which is exactly when social desirability and
anchoring-on-the-pitch run hottest. Discount them harder than the R2 revealed
prices, not less. What was volunteered, verbatim, unsolicited:

- Marcus: $1,000-1,200 per verification, pegged to what a Dave floor trip
  already costs, explicitly no premium for speed. This is a substitution
  price, not a value price, and it assumes the Dave gate opens, which has
  never been tested.
- Priya: $500-600 for the outside-process desk pass if registry-verified,
  $300 if it is "a nicer directory listing." The only number in the study
  attached to a named, just-realized loss ($1,250 and three weeks).
- Derek: $2,000-3,000, one time, month-11 holdback-resolution report,
  Higgins operating budget. He priced it as insurance on $150k and can
  defend the math "in one sentence."
- Aisha: $3,000-5,000 flat, sub-threshold deal-cost line, unchanged in two
  years, and never once approved. Two years of a stable unapproved number is
  evidence about the buy path, not about the price.
- Jenny: "maybe a few hundred dollars a year on top of VeriFact," double
  conditional (pre-window flag, provable invisibility), explicitly not a
  promise.
- Tom: $150-250 flat per shopped-print check, "every time Gary hands me a
  name." At his referral volume this is a low-hundreds-per-year account.
- Kelly: no number, but a structure: flat annual or per-check, Bluebeam
  shaped ($240/yr), "closer to what Chris costs me a month than what I'd pay
  a broker."
- Dana: no new number; her R1 $300-500 stands but the round reframed the ask
  entirely as structure (escrow, pay-after, catch record), not price.

Four rounds, zero purchases, zero commitments. The revealed-spend floor from
R2-R3 ($150-900 per event to trusted incumbents) remains the honest anchor,
and every volunteered number above sits at or below the incumbent-equivalent
cost, which means the roster is consistently pricing this as substitution,
never as premium. Assume real-world numbers come in under these.

## Wedge status after round 4

- **W1 Verification network: ALIVE but split in two, and the lead half
  changed.** The round separated two products that were traveling together.
  (a) The brokered inspector network survives only as a trust-structure
  product: escrow-released pay-after, visible catch records, zero intake
  friction. Its central assumption, that trust transfers through a
  platform's vouching, remains unproven, and the round's hardest evidence
  (Dana wiring $5,500 rather than paying a stranger 50% upfront) shows the
  market's current answer is no. Marcus adds a second wall: his only
  precedent ran on shared stake, and the Dave gate has never been tested
  against a disinterested stranger's report. (b) The sub-tier desk pass
  (registry-verified outside-process mapping before anyone travels) emerged
  as the sharpest artifact in the study: Priya named the exact step it
  saves, the exact dollar figure ($1,250 plus three weeks), volunteered a
  price premium over her R1 anchor for it, and Dana independently called
  the subcontractor map "the part I care about most and didn't even know to
  ask for." It is desk-only, criticality-tiered, and its kill condition is
  a single false "no outside processing." The desk pass is now the lead
  product of the study; the inspector network is behind it, gated on trust
  mechanics nobody has demonstrated.
- **W2 Agent-amplified rep desk: DEAD.** Retired R3, no interview run,
  retirement note on file. Closed.
- **W3 Castings/forgings desk: DEAD.** Hank's exit closed the last thread:
  he finally said the number to Denise and lost to the melt floor in one
  sentence. Even the free fix only half-works (the index card catches
  honest mismatches, not lies). Ownerless pain is now confirmed against
  both a zero-cost fix and a direct budget-holder ask. The residue, the
  shopped-print signal, lives on under W5.
- **W4 Succession demand book: DYING, two narrow residues, neither is a
  company.** The continuous post-close monitoring product died in the hands
  of its best-case buyer: Derek, with exposure, budget authority, and a
  proven inability to run the routine himself, still said nobody at NewCo
  would read a monthly memo, and he would rather lean on a 14-year
  controller's invoice-cadence instincts. What survives: (1) a one-time
  month-11 deposition-grade holdback-resolution report, $2-3k, which is a
  productized service with a 12-month sales cycle per customer and no
  recurrence until the next deal; (2) an incumbent-embedded
  relationship-durability module, which Aisha confirms gets funded same-day
  when Beacon Ridge asks and never when anyone else does. Both residues
  require a credibility asset (deposable name, incumbent brand) that a new
  vendor does not have and cannot quickly manufacture. W4 as a standalone
  wedge is one round from dead; what remains is a partnership thesis.
- **W5 Commission rep in software: WOUNDED, formally split.** Kelly's
  standing commission structure (0% upfront, 8-10% capped on closed work)
  survives untouched but was not re-tested, and nothing currently ships
  against it. The visibility product she actually wants cannot be
  commission-priced by her own math and lands at Bluebeam-scale flat fees,
  which is a $200-500/yr/shop product, real but tiny per account. The
  shopped-print check is the strongest single feature: wanted on both sides
  (Tom per-referral, Kelly per-quote, Hank's dead-RFQ flavor), priced by Tom
  at $150-250 a check, and blocked by one honest question nobody answered:
  where does the data come from before the network exists. W5's fate now
  rides on that mechanism question, not on demand.

## What we still do not know

1. Whether escrow/pay-after actually unblocks a booking. Dana specified the
   mechanism but never got to use one. Grupo Herrera's first article lands
   late August; her next deposit-gated event is the live test, and R5
   should put a firm structured offer against it.
2. Whether the Dave gate opens at all. Marcus can describe the acceptable
   report in detail (his checklist verbatim, named photos, callable
   inspector), but no disinterested stranger's floor report has ever crossed
   Dave's desk. Until Marcus agrees to put one in front of Dave, the
   inspector network's second-best account is hypothetical.
3. Where shopped-print data comes from. Kelly named the cold-start problem
   and a possible reciprocity model in the same breath. Whether shops will
   contribute their own quote flow to get visibility is a testable ask, and
   R5 should make it explicitly.
4. Whether the desk pass is buildable to the bar that was set: registry
   verified, sub-tier complete, criticality tiered, near-zero false
   negatives on "no outside processing." This is a feasibility unknown, not
   a demand unknown; demand for it is the best-evidenced in the study.
5. Whether any volunteered number survives a firm, dated, yes-or-no ask
   with the condition attached. Four rounds of priors; R5's labeled asks
   are the only remaining instrument before human interviews.
6. Frequency times price per segment. The card-movers (Dana, Priya, Jenny)
   run 2-6 billable events a year at $300-600; the shops price in the low
   hundreds per year; the acquirer products are one-shot per deal. Whether
   any segment supports a business at simulated volumes is an arithmetic
   question R5 must close.
7. Whether Jenny's invisibility bar is clearable even in principle. If
   catching a queue bump early requires anyone local asking around, the
   product detonates the relationship it protects. This may be unanswerable
   in simulation and should be flagged as a top question for real humans.

## Convergence call

NOT converged. The wedge board moved materially again (the desk pass
displaced the inspector network as W1's lead product, W4's subscription form
died and left two residues, W5 formally split into two pricing motions, W3
closed), every surviving product now has a persona-authored spec that has
never been priced firmly, and the study has collected zero commitment-grade
WTP evidence across four rounds by design. Round 5, the final round,
proceeds: segmentation and firm WTP, putting one dated, structured,
yes-or-no offer in front of each of the eight active personas, built
verbatim from their own round-4 redesigns.