A multi-agent run to position signalcollective.com. It started as a Signal Analyst product site — then two founder corrections moved the target. BigQuery ground truth showed the real buyer is a marketing team at a $5M–$100M+ DTC brand, and the site was re-scoped to the company site for the services firm — with Signal Analyst as the entry product and public demo, not the hero. Three company homepages came out of it — every market claim traced to a live source, every confidential number kept off the page.
01 · Two founder corrections
The run's first pass was rigorous and it ran clean — but it optimized for the wrong target. Two founder corrections on 2026-07-09 moved both the buyer and the whole purpose of the site. This recap keeps the first pass on the record (below) and ships against the corrected brief.
The pain hunt, the ICP tournament, and the positioning tournament all pointed at the sub-$5M owner-operator (Segment 1) and made real-time CAC the hero of a product site for the Aug-1 self-serve MVP. Internally consistent, well-evidenced — and, as it turned out, aimed at the person who complains loudest in public rather than the one who signs contracts.
Verified directly against BigQuery on the actual paid client roster: median ≈ $10M TTM, spanning $3M to $100M+, zero sub-$5M solo founders. The buyer is a small team — owner/CEO + in-house marketer, usually also working with an agency. "Five people, five different answers" is the pain precisely because multiple people look at the same numbers.
Open-forum evidence over-samples who complains in public and under-samples who signs contracts. The sub-$5M owner-operator survives — as the self-serve entry rung, not the center of the site's voice.
It leads with and sells strategic consulting, media buying, data analytics, custom data warehousing, and a growing AI advisory practice. Signal Analyst is the entry-level product and the public tech demo — the lowest rung of the engagement ladder ("start at $30/mo") and live proof the data layer is real — not the hero.
Primary CTA = the "Growth Signal Check" 30-min call. Secondary = start with Signal Analyst. The three tournament wedges survive one level down, as the Signal Analyst section's proof moments.
02 · The original brief (first pass)
The first-pass brief: find who Signal Analyst's customers actually are (real people, complaining about real pain, on the open internet), pick the winning segment and positioning through structured competition, then build the site around that verified buyer. The evidence work below is all still load-bearing — Correction 1 changed who it centers on, not whether the pain is real.
The entry product being sold
Signal Analyst — Claude MCP + Slack surfaces only (no new web chat); Shopify / Meta / Google connectors (Klaviyo coming soon); token plans $30 / $75 / $100 / $200 per month; ships Aug 1 2026. Now positioned as the entry rung and public demo of the firm's data layer — not the whole site.
The three hard rules that governed every word
Publish nothing (all output local). Invent nothing (every market claim needs a source URL). Confidentiality — client names, financials, team comp, the internal margin, and exact enterprise/retainer pricing never appear in public copy.
03 · The pain hunt
Parallel researchers fanned out across Shopify Community, G2/Capterra reviews of the incumbents (Triple Whale, Northbeam, Polar, Lifetimely), agency and fractional-CFO communities, and MCP/Claude power-user threads. Every load-bearing quote was independently re-fetched and confirmed real before it counted. The distrust these quotes capture is exactly what the services firm gets paid to fix — the evidence outlives the ICP correction.
"When I have 100 sessions and 5 sales. I should have a 5% conversion rate. Shopify however… tells me it's 0.7%."
"Some sales are simply not recorded in Shopify at all."
"Most sellers do not really trust the performance numbers from Meta and Google."
"Net profit per order, including product costs (COGS), shipping costs, transaction fees, taxes, and advertising expenses — how can I accurately calculate this?"
"Paid plan is extortionate and free plan is pathetic." Lifetimely
"[requires] a sales call [before you can even try it]." Polar
04 · ICP tournament overridden by Correction 1
A judge panel scored all eight segments and ratified Segment 1 (sub-$5M solo owner-operator) as the beachhead. Correction 1 overrides that ratification: the real paid roster has no sub-$5M solo founders. The lesson is baked into the table below — the segment the tournament ranked #2 (a $5–10M owner-operator team) was directionally right about who pays; the founder correction widened it to marketing teams at $5M–$100M+ brands.
| # | Segment | Tournament verdict | Post-correction reading |
|---|---|---|---|
| 1 | Self-serve solo / small DTC owner-operator sub-$5M, ~$500–2K/mo ad spend | was "Primary ICP" | Best-evidenced in public forums — but public forums over-sample complainers. Demoted to the self-serve entry rung (the Signal Analyst $30 tier), not the site's center. |
| 2 | DTC owner-operator team, $5–10M+ now: teams at $5M–$100M+ | was "Secondary" | Corrected primary ICP Real paying precedent + tightest pain-to-service match. Correction 1 widened it to the full roster: owner + in-house marketer ± agency, sweet spot ~$8–40M. |
| 3 | In-house growth lead / media buyer | Hero user | Now a member of the buying committee — the daily user of Signal Analyst and of the retainer's reporting. Threaded into every company page's team voice. |
| 6 | Agency / growth-consultancy owner | Untested channel | Reframed: the agency is now part of the buyer's committee (the reporting that gets reconciled), not a resale channel. |
| 8 | Fractional CFO / finance lead | Referral bet | Still a plausible referral partner; unchanged. |
| 4 | VC-backed Series A–C founder | Persona mismatch | Verified founder voice is bootstrapped DTC, not VC. Unchanged. |
| 7 | Enterprise "AI-ready data" buyer | Unproven demand | The AI-advisory practice serves the upper end of the roster — a growing engagement, honestly hedged, not a self-serve claim. |
| 5 | Strategic CMO | Weakest | Unchanged; supported only by abstract survey stats. |
05 · Positioning tournament re-scoped by Correction 2
A skeptic pass found four of the five concepts shared one spine — "your three numbers don't agree." Three finalists advanced as full product pages: real-time CAC, analyst-in-Slack, and anti-disagreement. Correction 2 re-scoped them: the three wedges survive as Signal Analyst section framings and proof moments on the company site — not as page heroes. The company site's hero is the firm.
| Rank | Concept | First-pass status | Where it lives now |
|---|---|---|---|
| 1 | real-time CAC | was Hero | Signal Analyst proof moment — "your CAC the same day." Powers the operator-console demo on company page 2. |
| 2 | analyst-in-Slack | was Supporting | The surface-honesty framing for the Analyst section (Claude MCP + Slack, no web chat). |
| 3 | anti-disagreement | was Supporting | Became the editorial spine of company page 3 — "three answers → one reconciled number." |
| — | ai-ready | Cut | Its enterprise-scale anxiety now maps to the honestly-hedged AI-advisory service, not a $30 buyer. |
| — | decision-smarter | Re-elevated | "Make every decision smarter / embedded, not outsourced" was too big for a product wedge — but it is exactly right for the company hero. It became company page 1. |
06 · The three company pages
Each is a full, self-contained company site: five named services, the engagement ladder with Signal Analyst as rung 1, a dedicated Analyst "entry product / public demo" section, the "Growth Signal Check" primary CTA, and "Start with Signal Analyst" secondary. The red team certified all three ship-ready — zero unresolved CRITICAL/MAJOR — and genuinely distinct on every visual axis.
It executes the brand identity (deep green + lime on a warm field) so it doubles as the firm's first public statement, and it operationalizes the exact direction Correction 2 re-elevated ("Make every decision smarter / embedded, not outsourced"). Its partnership-warmth register reads to both the owner who signs and the marketer who uses it daily, and it is the only page whose Signal Analyst secondary CTA already routes to a real signup surface. Keep Company 2's operator console and Company 3's reconciliation apparatus as supporting proof sections beneath it.
Brand identity executed, services-led, warmest to the full buying committee, Analyst signup already wired. Ship as the homepage.
Sharpest operator credibility. Best kept as the "we run it" proof band + the same-day-CAC Signal Analyst demo.
Most visceral, most on-brand for luxury DTC. Best as the reconciliation section — keep the third-party quotes attributed.
Solo-operator product hero. Now demoted to a Signal Analyst proof moment (company page 2's console).
concept-1 →Surface-honesty product hero. Now the Analyst section's surface-explainer framing.
concept-2 →Reconciliation product hero. Re-aimed and re-scaled into company page 3's editorial spine.
concept-3 →06b · Claude Design refinements
After the company-site pass, the "One Number" positioning was iterated in Claude Design (claude.ai/design) into three visual treatments. Each was imported via the Claude Design MCP and converted to a standalone page. Placeholder image slots await real photography.
Raw .dc.html sources in site/design-src/ · Claude Design project retains the editable originals.
07 · Signal Analyst — the entry product
On the company site, Signal Analyst is not the hero — it is where a brand starts ("start at $30/mo") and live proof that the firm's data layer is real ("the same technology behind our client work — try it"). Plans are named by shop stage, priced for the whole team (not per seat), and honest about a token-metered balance.
Launch-week play: the "Growth Signal Check" 30-min review is the top of the funnel; founder-disclosed thread-answering in the exact Shopify Community threads where owners ask for this tool feeds the Analyst entry rung (zero paid acquisition), with the already-approved App Store listing as the always-on companion asset. The self-serve product is the trojan horse; the services ladder is the business.
08 · Red-team honesty summary (company-site pass)
The final critic re-attacked the three re-scoped company pages against the company-site claims map and both founder corrections. All three per-page passes closed every CRITICAL/MAJOR they raised; the critic re-verified the fixes and checked the four things this pass owns: distinct aesthetics, services-led architecture, claims compliance, and team voice. Zero unresolved CRITICAL or MAJOR on any page.
09 · What to do next