Methodology

What ChatGPT doesn't know,
AI Market Twin does

Ask a generic AI chatbot "Will my premium skincare brand sell in Vietnam?" and you get a generic guess. AI Market Twin runs simulations on a persona pool that's actually grounded in real government and public statistics — Vietnam's Bộ Y Tế food registration process, Indonesia's BPS household income distribution, Japan's 厚生労働省 cosmetics regulations, and 21 more.

24-Country Reference Database

Each country's personas come from
that country's official statistics

Income distribution, profession-level wages, household composition, consumption patterns, and core regulations — measured data pulled from each country's government and public statistics offices is the ground truth for persona generation.

🇰🇷
Korea
KOSIS · Korea Customs Service
🇺🇸
United States
U.S. Census Bureau · BLS · Nielsen
🇯🇵
Japan
e-Stat · 総務省 · 厚生労働省
🇬🇧
United Kingdom
ONS · Family Resources Survey
🇩🇪
Germany
Destatis (Statistisches Bundesamt)
🇫🇷
France
INSEE
🇮🇹
Italy
ISTAT
🇪🇸
Spain
INE Spain
🇳🇱
Netherlands
CBS Netherlands
🇨🇦
Canada
Statistics Canada
🇲🇽
Mexico
INEGI
🇧🇷
Brazil
IBGE
🇦🇺
Australia
ABS (Australian Bureau of Statistics)
🇨🇳
China
NBS · 国家统计局
🇹🇼
Taiwan
DGBAS · 行政院主計總處
🇸🇬
Singapore
SingStat · HSA
🇲🇾
Malaysia
DOSM (Department of Statistics)
🇹🇭
Thailand
NSO Thailand · Thai FDA
🇻🇳
Vietnam
GSO Vietnam · Bộ Y Tế
🇮🇩
Indonesia
BPS Statistics Indonesia
🇵🇭
Philippines
PSA Philippines
🇮🇳
India
MoSPI · NSSO
🇦🇪
UAE
Federal Competitiveness Authority
🇸🇦
Saudi Arabia
GASTAT (General Authority for Statistics)

Refreshed annually via an automated GitHub Actions pipeline · Additional countries can be onboarded in 4–6 weeks on request

Why Not Just Ask ChatGPT?

What sets us apart from a generic AI chatbot

You'll get an answer either way. The question is whether the answer is something you can actually take into a boardroom.

Generic AI Chatbot (ChatGPT/Claude/Gemini)

"Will our product sell well in Vietnam?"
  • ×
    General knowledge from training cutoff. No specifics on Vietnam GSO 2024 household income distribution or Bộ Y Tế food registration procedures.
  • ×
    Reasoning isn't traceable — can't answer "why did you conclude that?"
  • ×
    One persona. No intent distribution, segment diversity, or minority-opinion visibility.
  • ×
    No quantitative outputs — no price curve, CAC estimate, or market prioritisation.
  • ×
    No executive PDF or charts — you have to assemble the deck yourself every time.

AI Market Twin

200 personas grounded in government statistics · traceable sourcing
  • Live integration of 24-country official statistics + category-specific regulations (Bộ Y Tế · 厚生労働省 · Thai FDA · UK CPSR, etc.)
  • Every persona statement is traceable to its source data cell — sources listed in every PDF.
  • 200-persona intent histograms, segment-level rejection factors, minority-champion visibility.
  • Price-vs-conversion curves, country-level CAC estimates, automatic HIGH/MEDIUM/LOW risk classification.
  • Executive PDF generated in one click (Korean and English).
Verifiable Accuracy

Measurable accuracy,
open scoring rubric

The most reliable way to reduce LLM hallucination is external grounding plus self-evaluation. AI Market Twin runs six public-data anchors and a 5-metric self-scoring pipeline.

Six external grounding anchors

Hofstede 6-dimension cultural indices
28 countries × Power Distance · Individualism · Uncertainty Avoidance, etc. — calibrates persona decision priors
World Bank macro indicators
GDP per capita PPP · population · household consumption — the base for price-sensitivity and market-size estimation
UN Comtrade trade flows
Korea → partner export flows by HSCode — quantifies existing market interest per category
Korea Customs export records
data.go.kr 관세청 OpenAPI · monthly granularity / 10-digit HSCode precision (complements Comtrade)
DART corporate disclosures
Consolidated financials + region-segment revenue for Korean listed companies — brand-level overseas-revenue anchor
KOTRA registered Korean entities
86-country KOTRA registry of Korean companies abroad — category-matched presence signal that calibrates brand recognition

5-metric self-scoring pipeline

30%
top3Hit
Fraction of simulated top-3 markets that match the actual top-3 ground truth
25%
rankCorrelation
Spearman correlation between simulated market ranking and measured revenue ranking
20%
rejectRecall
Whether markets the brand actively avoided are also rejected by the sim (false-positive avoidance)
15%
confidenceCalibration
Whether STRONG / MODERATE / WEAK labels are calibrated against actual accuracy
10%
trendMatch
Whether predicted market-trend direction agrees with measured data
Honest measurement principles

The 5-metric scoring runs as open, auditable logic. Every code change is re-measured on the same product set (multiple K-product fixtures) automatically, and results are verified with paired t-tests for statistical significance. The measurement-improvement cycle runs weekly.

Every sim result lists the grounding anchors used and a per-metric score breakdown directly in the PDF. Accuracy improvements roll out to existing customers automatically — no separate upgrade cost per release.

Pipeline

The 6 stages every simulation runs through

From wizard input to PDF output in 5–7 minutes on average. Each stage uses a different LLM model tuned to its role.

01 — VALIDATING

Input validation + slot planning

Validates product info and pre-allocates 200 persona slots from category-specific profession pools.

02 — REGULATORY

Up-front regulatory check

Inspects sales bans and labelling rules per country and category. Markets where launch is structurally impossible are auto-excluded.

03 — PERSONAS

Pool sampling + voice generation

Reuses matching personas from the workspace pool; only new slots get fresh generation. Each persona's first-person voice quote is generated alongside.

04 — SCORING

Country-level prioritisation

Aggregates intent distribution, rejection factors, and trust signals across 200 personas to score each market on demand, CAC, and competitive intensity.

05 — PRICING

Price curve (3-sample median)

Three parallel pricing simulations with median selection eliminate single-call variance. Per-market price sensitivity broken out separately.

06 — RECOMMEND

Synthesis + self-critique

Vision model analyses any uploaded creative. Self-critique pass automatically verifies macro consistency (best-country alignment, etc.) before finalising the result.

Try it

See an actual result — sample report

An anonymised K-food 5-market validation simulation. PDF, no signup.

Download sample report (PDF) Start free