Trueness: the plan

The whole plan, distilled to what an investor needs. 25 October 2026. Everything below is a hypothesis until a client has paid. The working papers behind each line are listed at the end.

1. One line

Trueness makes synthetic audiences usable for real decisions at billion-dollar consumer brands: a panel the team uses every day, calibrated every quarter against fresh real people, with a client-owned record of how far it is from the truth.

2. The problem

A $1bn consumer brand spends $10–15m a year in agency fees, about $1.5m on brand strategy before a campaign ships, and runs fifteen research sources in six months. It still cannot test a homepage image on a Tuesday.
Synthetic audiences fix the speed. Nobody knows how wrong they are. Vendors report their own accuracy (86–96% on incomparable metrics); independent studies find individual-level correlation of 0.2, collapsed variance and failures on subgroups and novelty.
The industry's own bodies now say: validate against humans and disclose. ESOMAR guidance is due January 2027. No neutral party sells the validation.

3. The product

Every day: drop any ad, email, page or deck; twelve people who behave like your audience react in their own voices with a score and a reason; a chair gives a verdict, a second reader challenges it; interview any one of them; ask the research behind them and get a cited answer; export one page for the room.
Every quarter: fresh, sealed human research in every market measures every audience; the engine recalibrates.
Always: a Calibration Record per audience, market and use, versioned, signed by an external reviewer, owned by the client, with scope of use on every answer.
Everywhere: web workspace, API and an MCP server, so the brand, its agencies and its AI tools work from the same calibrated audiences.
Proof of the pattern: a reference build of exactly this daily experience was adopted by a brand's marketing team across brand, content, email and channel within a fortnight, with no training. It lacked only the record.

4. Why now

Over $1.5bn of venture money has gone into simulators (Simile $100m Series A, Aaru at a $1bn headline valuation, Electric Twin $14m) that each validate themselves.
Incumbents (Kantar, Ipsos, Toluna, Qualtrics) calibrate inside their own panels; the record is not portable.
Standards are arriving (ESOMAR 2027, AAPOR 2026, ARF benchmark end-2026) that will ask buyers to show validation.
Brands are moving money from agency production to evidence. The owner of a $15m non-working line is the buyer.

5. The market

Serviceable: consumer businesses of $1bn and above in high-consideration categories (travel, financial services, automotive, healthcare, education, premium durables), roughly 300–500 globally, at $0.6–1.2m a year each: $300–600m.
Plus the channel: holding-company agencies and consultancies licensed to run the standard for their clients.
Plus vendor credential audits.

6. Business model

Offer	Price	Anchor
Enterprise Audit	$15k per audience-market cell, $75k floor; a typical seven-cell audit is $105k against about $94k of direct cost	One strategy workstream
Trueness Programme	$600k a year for three audiences in two markets; further audiences and markets priced at their fieldwork cost; multi-brand Group programmes from $3m	One and a half brand trackers; 4% of the non-working line
Agency licence	Fee plus 20–30% revenue share	Agencies resell simulation today

Gross margin is an output of the model, not a promise: 41% at the mid price, 61% at $1.0m, and 60% needs about $1.1m per three-audience programme. The largest cost is fieldwork, which is the point: the client is paying for real people. Delivery is a research lead per two programmes, an external reviewer, and agents running the back office.

7. Competition

Pure-play syntheticIncumbent with own human panelReal respondents, AI-assistedTrueness, proposed position (not yet built) Marker size, indicative annual spend:Under $75k$75k to $250k$250k+ or undisclosed

Everyone on the map validates its own product or calibrates inside its own panel. The top-right, neutral and record-first, is empty. Every feature is replicable; being the neutral party first, with the body of records, is the position.

8. The plan

Months 0–2: one anchor design partner signed at $0.5m+; research lead and reviewer contracted; legal clean-room review done.
Months 1–6: first two quarterly waves; the anchor's agency on the MCP; two Enterprise Audits sold in adjacent categories.
Month 6 gate: audits convert to programmes; agency channel opens; second research lead.
Platform in four phases from a thin workspace to portfolio tenancy over 14 months; SOC 2 Type I in year one.

9. Financials (USD; Codex's enterprise model, final, eight test families passing, 25 October 2026)

	Low	Mid	High
Programmes signed, years 1 / 2 / 3	3 / 6 / 10	5 / 14 / 28	8 / 24 / 50
Average contract value	$500k	$750k	$1.0m
Year-1 revenue		$1.3m
Year-3 revenue	$6.1m	$24.1m	$62.9m
Year-3 EBITDA	-$2.4m	$6.7m	$33.8m
Mature programme gross margin	-7%	41%	61%
Funding need, incl. 20% contingency	$8.6m (stop case)	$2.9m	$2.2m

What the model says plainly: year one is slow because enterprise sales take six months and activation two, and no contract value is recognised on signature. The low case does not work at $500k and is a stop signal. The 60% margin the plan targets needs about $1.1m per Foundation programme at the modelled costs, so the mid case is a 41% margin business unless price or fieldwork cost moves. The six-audience, four-market programme is not viable at $1.2–1.5m: first-year direct margin is 21% on semi-annual waves and negative on quarterly, and 60% would need roughly $3.0–4.4m. The product is therefore the three-audience, two-market programme, with audiences and markets added at their real fieldwork cost. The six-month pilot (one $600k anchor plus two seven-cell audits at $105k) contracts $810k, recognises $410k and needs about $1.24m of funding including platform build and buffer, with the first wave consuming about 41% of first-year fieldwork.

10. The ask, and what could go wrong

Ask: $2.9m seed; introductions to three chief marketing or customer officers at $1bn+ consumer brands; one named independent methodologist as reviewer.
Team: a founder who built the reference panel inside a $1bn brand programme (anonymous by design; a named commercial lead fronts the venture), two AI agents running research drafting, statistics and operations, a contracted research bench.
What could go wrong: procurement has no line for a new category (we price it as one it has); fieldwork cost before cash (milestone billing, panel terms); an incumbent ships a portable record first (publish first, invite them to license); conflict of interest on our own builds (external review, separate custody); the method's provenance (clean-room rule, legal review before the first contract).

Working papers

The documents behind the plan. The full index with notes is at /plan/papers.

brand

Trueness brand platform/plan/brand/trueness-brand-platform

Trueness: the plan

1. One line

2. The problem

3. The product

4. Why now

5. The market

6. Business model

7. Competition

8. The plan

9. Financials (USD; Codex's enterprise model, final, eight test families passing, 25 October 2026)

10. The ask, and what could go wrong

Working papers

brand

finance

operations

product

research

review

strategy