Trueness.ai business plan: the calibration programme for billion-dollar consumer brands
Version 3, 25 October 2026, enterprise re-cut on the owner's instruction. Claude drafts, Codex reconciles the numbers; the financial case is canonical once rerun on the parameters in section 8. All figures USD. Prices, volumes and win rates are hypotheses until buyers have paid them. Wording rules from version 2 stand: no first, only, certified or guaranteed; Calibration Record, not score; blinded validation, external review of our own builds; inconclusive is a paid outcome.
1. What changed and why
Version 2 priced Trueness for a research manager with a $12,500 synthetic tool and a question about whether it was wrong. That buyer exists, but the decisions, the reputational stakes and the money sit elsewhere. The reference organisation, a consumer business with about $1bn of revenue, spends $30–40m a year with its integrated agency, $10–15m of it in fees, about $1.5m on one-off brand strategy before a campaign ships, $0.2–0.4m on each strategy workstream, $0.2–0.4m a year on a brand tracker and seven figures across fifteen research sources in six months. Its owner was trying to hold non-working agency spend under $15m while increasing research and validation. In that budget a $35k engagement is invisible and a $0.6–1.2m programme is one line among twenty. The benchmark detail is in enterprise spend benchmarks. Trueness is re-cut to be that line.
2. The proposition
What the client's team gets every day: a panel of people who behave like its audiences, convened in a minute on any piece of work, who react, disagree, give a verdict, argue with the verdict, and can be interviewed one at a time; a research bench that answers questions with citations; a one-page export for the room. The reference build showed a marketing team using exactly this across brand, content, email and channel within a fortnight, with no training and no plan for the uses it found. That usefulness is the product; see what it is like to use Trueness every week.
What Trueness adds that the reference build lacked: an answer to "how do you know".
For consumer businesses of $1bn and above, Trueness runs the calibration programme that makes synthetic audiences usable for real decisions. We build the brand's synthetic audiences or take the ones its agencies and vendors already use, measure every one of them against fresh, blinded human research each quarter, and keep a client-owned Calibration Record current for every audience, market and use. The brand, its agencies and its AI tools then work from the same audiences through the web, an API and an MCP server, with the record's scope of use attached to every answer.
The promise to the chief marketing or customer officer: you can test as much as you like on Tuesday, because you know from real people, every quarter, exactly how far the synthetic audience is from the ones you are paying to reach. The promise to the board and the regulator: here is the record.
3. Why this buyer, and what they already pay for
| What the $1bn brand buys today | Typical annual spend (observed, rounded) | Trueness relationship to it |
|---|---|---|
| Brand health tracker, four markets | $0.2–0.4m | The programme's quarterly human waves are tracker-scale fieldwork; part of the tracker budget moves to Trueness or the tracker is integrated as a holdout source |
| Segmentation, positioning, benefit ladder, territories (one-off) | $0.4–1.5m | Consumed as the grounding corpus; the programme keeps the segmentation alive as calibrated audiences instead of a deck |
| Strategy workstreams (CRM, creator, MMM, SEO) | $0.2–0.4m each | The programme is priced and procured like one of these |
| Ad hoc quant and qual studies, DIY testing platforms, user testing | $0.3–0.6m | Mostly replaced for early-stage decisions; retained and used as holdouts for consequential ones |
| Agency strategy and creative fees | $7–10m | Agencies test their work against the calibrated audiences before the client sees it; the programme is written into the agency's scope of work |
| AI delivery partner | $0.5–2m | The audiences and records are exposed through MCP to the brand's AI workflows |
The buyer is not a research manager. It is the executive who owns the $15m non-working line and is already trying to turn agency opinion into evidence. Secondary buyers are the holding-company agencies and consultancies that serve those executives and need an independent check they can put in front of the client, and the synthetic-audience vendors who want their product measured by a neutral party for a credential.
4. Offers
| Offer | What it is | Scope | Price (USD, hypothesis) | Anchor in the buyer's budget |
|---|---|---|---|---|
| Enterprise Audit | A one-time measurement of every synthetic audience the brand and its agencies use, vendor and in-house, against fresh blinded human holdouts; consolidated Calibration Records; a board-level read of what can and cannot be relied on | Priced per audience-market cell: ~240 completes plus interviews per cell, 4 stimuli families, 6–8 weeks, up to 18 cells | $15k per cell, $75k floor; a typical seven-cell audit $105k | One quant A&U study plus one qual round; one agency strategy workstream |
| Trueness Programme, Foundation | Annual: 3 calibrated audiences in 2 markets; quarterly human waves (~1,000 completes per market per wave plus 12 interviews per audience); Calibration Records kept current; web, API and MCP access for the brand and its agencies; 12 bounded studies a month; named research lead; quarterly executive read | 3 audiences × 2 markets | $600k a year | One and a half trackers; 4% of a $15m non-working line |
| Trueness Programme, expansion | Further audiences and markets added to a Foundation programme, each priced at its own fieldwork and research cost (a market is a cost stress, not a free extension); semi-annual full waves with monthly pulses where quarterly is unaffordable | Per audience, per market | Quoted; the model shows a six-audience, four-market programme needs $3.0–4.4m to reach 60% margin | A CRM strategy plus MMM plus tracker |
| Trueness Programme, Group | Multi-brand portfolio: shared and brand-specific audiences, regional markets, group-level governance of synthetic research across agencies | Portfolio | From $3m a year | A fraction of group agency fees |
| Agency and consultancy licence | A holding company, consultancy or research house runs Trueness audits and programmes for its clients under the Calibration Record standard, with Trueness as reviewer | Per client programme | Licence fee plus 20–30% revenue share | Holding companies already resell simulation |
| Vendor credential audit | A synthetic-audience vendor has one of its audiences measured by Trueness with the client's consent and the result published | Per audience | $50–100k | A marketing line for the vendor |
Not sold: single-audience $15k audits and $35k builds. Those scopes exist inside the programme as units of work; the $75k five-cell audit is the entry. Recalibration on a model change is inside the programme, not an upsell.
5. How Trueness is distinct, and what it is not
The same facts as version 2, applied at this scale:
- Electric Twin, Simile and Kantar already validate against humans on their own products. Trueness measures any audience, including theirs, with a fresh prospective holdout and gives the client the record. At programme scale this becomes governance: one record standard across every audience the brand uses, whoever built it.
- Qualtrics, Toluna and Ipsos calibrate inside their own panel contracts. The programme is portable across vendors and survives an agency or vendor change.
- Trackers measure the brand. The programme measures the instruments the brand uses to think about its customers. They complement each other and can share fieldwork.
- Agencies sell judgement. The programme gives the client the evidence to test the judgement, and gives the agency a safer way to be right.
Every element is replicable. The durable assets are the body of records, the client relationships at executive level, the reviewer bench and the standard itself. Trueness is not a simulation vendor (it is model-agnostic and will run the client's chosen simulator), not a panel (it buys fieldwork), and not a consultancy (it does not recommend the campaign; it tells you how far the rehearsal is from the audience).
6. Go-to-market
- One anchor design partner at $0.5m or more, signed before anything else is built. A $1bn consumer brand mid-transformation with a live synthetic panel and a research lead who already wants to calibrate it. The programme is scoped jointly, the first quarterly wave doubles as the launch case, and the partner's agencies are enabled in the first ninety days. The partner's identity is never used in marketing without written consent; the record is.
- Two Enterprise Audits in adjacent categories (financial services, automotive, healthcare, higher education, premium durables) within the first six months, each sold at executive level through warm introduction, each a candidate to convert to a programme.
- Agency channel from month six. One holding-company network and one consultancy licensed to run the standard for their clients; Trueness reviews and signs. The channel is where volume comes from in years two and three.
- Vendor credential audits as content: invite the simulators to be measured; publish what happens, including inconclusive results, with consent.
- Standards work: map the Calibration Record to ICC/ESOMAR 2025 and the forthcoming ESOMAR synthetic-data guidance; offer the template to working groups; aim to be the named example when guidance is issued.
- No self-serve, no freemium, no paid media in year one. Executive relationships, published records and the channel do the work.
Sales cycle assumption: six to nine months from first conversation to signed programme at enterprise procurement; three months for an Audit. Win rate assumption: one in four qualified executive conversations becomes an Audit; one in two Audits converts to a programme within twelve months.
7. Operating model: a delivery organisation, run by agents, led by people
The pilot of version 2 assumed one human and two agents. A $1.2m programme assumes a named research lead the client can call, an external reviewer who signs, and a delivery bench. AI does the preparation, drafting, simulation, statistics, record generation, monitoring and administration; humans design, interpret, sign and sell.
| Role | Per programme | Notes |
|---|---|---|
| Research lead (human) | 0.5 FTE | Owns design, interpretation, client readouts; can carry two programmes |
| External methodological reviewer (contracted) | ~10 days a year | Signs every record for audiences Trueness built; separate holdout custody |
| Fieldwork and recruitment (bought) | $150–300k a year | Panels and recruiters; incentives; moderation |
| Engineering and operations (shared) | 0.2 FTE | Platform, MCP, tenancy, security; grows with client count |
| AI usage | $10–20k a year | Metered; model and version recorded on every record |
| Client and channel management (human) | 0.2 FTE | Executive relationship, QBRs, agency enablement |
| Venture lead | Across all | Contracts, pricing, reviewer bench, standards |
Target gross margin 55–65% at Foundation and Enterprise; Group programmes and the agency licence carry higher margin because fieldwork scales sub-linearly across brands and the channel supplies the delivery labour. The back office (proposals, contracting, panel orders, fieldwork QA, invoicing, bookkeeping, renewals) stays agent-run with human approval gates where money moves or a record is signed, as specified in operations.
7a. The platform
The programme runs on a governed research platform with five parts: an audience engine that builds and runs synthetic audiences on any named model from the brand's own evidence; a calibration engine that recruits sealed human holdouts each quarter and measures every audience against them; a record registry holding the client-owned Calibration Record for every audience, market and use; three access surfaces (web workspace, API, MCP server) through which the brand, its agencies and its AI tools query the audiences with the record's scope of use on every answer; and an operations layer that runs the eighteen back-office procedures as agent workflows with human approval gates. Tenancy is brand plus agency sub-tenants with client-controlled sharing; enterprise security (SSO, isolation, audit, SOC 2 Type I in year one) is a procurement requirement, not a feature. The build runs in four phases from a thin workspace for the anchor's first wave (months 0–2) to channel and portfolio tenancy (months 8–14), at an assumed $0.6–0.9m of first-year engineering before the research bench. Full specification: the platform, with screens in the product specification and model routing in models and integrations.
8. The three cases (parameters for finance/ to rerun)
| Low | Mid | High | |
|---|---|---|---|
| Programmes signed, years 1 / 2 / 3 | 3 / 6 / 10 | 5 / 14 / 28 | 8 / 24 / 50 |
| Average programme ACV | $500k | $750k | $1.0m |
| Enterprise Audits, years 1 / 2 / 3 | 2 / 4 / 6 | 4 / 8 / 12 | 6 / 12 / 20, at $110k average |
| Agency licence revenue | none | 10% of programme revenue from year 2 | 25% from year 2 |
| Programme churn | 25% a year | 15% | 10% |
| Fieldwork per programme | $300k | $220k | $180k (volume terms) |
| Research lead cost | $180k fully loaded, 2 programmes each | same | same |
| Reviewer, engineering, AI, client management | as section 7 | ||
| Sales cycle | 9 months | 6 months | 6 months |
| Gross margin target | 50% | 60% | 65% |
| What breaks it | Procurement will not buy a new category at $500k without a tracker-scale track record; fieldwork cost overruns | Research-lead hiring lags signed programmes | Channel partners under-deliver; reviewer capacity |
Codex's final rerun (25 October 2026, eight test families passing): year-3 revenue and EBITDA of $6.1m / -$2.4m low, $24.1m / $6.7m mid, $62.9m / $33.8m high; year-1 mid revenue $1.3m because of a six-month sales cycle and two-month activation with no contract value recognised on signature; mature programme gross margin -7% / 41% / 61%; funding need including 20% contingency $8.6m low (a stop case), $2.9m mid, $2.2m high. The 60% margin target needs about $1.1m per three-audience programme. The six-audience, four-market Enterprise tier is not viable at $1.2–1.5m (first-year direct margin 21% on semi-annual waves, negative on quarterly; 60% would need $3.0–4.4m), so the offer table is corrected: the programme is three audiences in two markets at $600k, with audiences and markets added at their real fieldwork cost, and Group programmes from $3m. The Enterprise Audit is priced per audience-market cell at $15k with a $75k floor; a typical seven-cell audit is $105k against about $94k of direct cost. The six-month pilot (one anchor plus two seven-cell audits) 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. Wave design: 240 sealed plus 160 development per audience per market in wave one (about 1,200 completes per market for three audiences), 240 plus 80 thereafter. The canonical figures are in the financial case and the cadence comparison in programme-cadence-comparison.csv.
9. Pilot: one anchor, two audits, two quarters
| Period | Milestone |
|---|---|
| Months 0–2 | Anchor partner scoped and signed; research lead and reviewer contracted; legal review of IP and employment complete; venture identities and MSA ready |
| Months 1–3 | First human wave in two markets; audiences built or ingested; first Calibration Records; MCP exposed to the partner's agency |
| Months 3–6 | Second wave; drift read; executive QBR; two Enterprise Audits sold and delivered |
| Month 6 | Decision gate: convert audits to programmes; open the agency channel; hire the second research lead |
Success is not a statistical threshold. It is a renewed anchor, two audits delivered with honest records, and one signed channel conversation.
10. Risks (additions to version 2)
| Risk | Response |
|---|---|
| New category at enterprise price: procurement has no line for it | Price and scope it as a known line (tracker, strategy workstream); contract through the agency SOW where easier; anchor partner as reference |
| Fieldwork at tracker scale is a real cost before cash | Milestone billing; fieldwork quotes before signature; volume terms with two panels |
| Research-lead supply | Bench of contracted methodologists; reviewer network; AI absorbs the preparation so leads carry two programmes |
| Agency resistance (we are measuring their work) | Position as protection for the agency; enable them first; write the programme into their scope |
| Incumbents (Kantar, Ipsos, Qualtrics) launch a portable record | Be the standard first; publish; invite them in as licensees |
| Conflict of interest on our own builds | External review and separate holdout custody on every record |
| Owner anonymity versus executive selling | A named, non-founder commercial lead fronts the venture; the venture entity contracts |
| IP provenance of the method | Clean-room rule; legal review before the anchor contract |
Version 2's risk table (validity, inconclusive findings, supplier quality, rights, drift, cash) still applies and is kept in the private history.
11. Decisions for the owner
- Approve the enterprise re-cut and the three case definitions for Codex's rerun.
- Name the anchor design partner candidate and the route to its executive.
- Who fronts the venture commercially, given the anonymity constraint.
- Seed funding appetite at $2.9m (mid case), and whether to price programmes nearer $1.1m to reach the margin target.
- Whether the agency licence is offered in year one or held until the standard is proven.