Risk Analysis # What if the risk pipeline could see the territory? Geospatial data as complementary variables in credit, insurance, onboarding, compliance, and KYC pipelines — income, border, territorial context, and proximity to sensitive areas. No CPF required, via API. 588 municipalities in Brazil's border strip 11,5 mi fraud attempts in Brazil in 2024 R$ 3,76 bi seized in smuggling and contraband in 2024 The problem ## Risk pipelines are blind to territorial context Credit, insurance, and compliance models rely on registration and behavioral variables — but ignore **where** the customer is, what their surroundings reveal, and in what territorial context the operation takes place. A credit score treats a person in rural Minas Gerais the same as one in a twin city on the Triple Border. 0 territorial variables in most traditional risk models Credit scores don't distinguish border regions, prisons, or urban areas 16.886 km of land border with virtually no integrated digital control Source: SEDEP / 150 km strip with 11 million inhabitants +24% more financial intelligence reports (FIRs) issued by COAF in 2023 — suspicious money laundering operation reports Source: COAF / Metrópoles, 2024 — Triple Border is a risk area recognized by FATF The scenario ## Two identical profiles. Opposite territorial contexts. Without geospatial data, your risk pipeline cannot distinguish a customer in a stable residential area from one in a border zone with a money laundering history — both arrive with the same score. Profile A ### Savassi neighborhood — Belo Horizonte/MG **Declared income:** R$ 10.000 **Regional median income:** R$ 9.152 **Z-score:** +0.51 (compatible) **Border strip:** NO **Nearest prison:** 8,4 km **Nearest community:** 1.2 km (outside) → Compatible income. Stable territorial context. Proceed. Profile B ### Downtown — Ponta Porã/MS (dry border) **Declared income:** R$ 10.000 **Regional median income:** R$ 1.890 **Z-score:** +4.28 (anomaly) ⚠ **Border strip:** YES — Paraguay (1.1 km) ⚠ **Nearest prison:** 2,3 km **Twin city:** YES — Pedro Juan Caballero ⚠ → Highly incompatible income. Dry border. Twin city. Escalate for review. Without PrismaData, both have the same score. With it, Profile B is immediately flagged — income incompatible with the surroundings, border zone, and twin city with a money laundering history. The solution ## Multiple layers of context. One coordinate. Send latitude and longitude. Receive back estimated income with z-score, border proximity, twin city indicator, community context, and prison proximity. Each variable is a feature ready for the risk pipeline. GET /v1/location/personal_income_pdf GET /v1/location/border GET /v1/location/slum GET /v1/location/prison Combined response — Ponta Porã/MS (lat=-22.5302, lng=-55.7162) { // — Income (personal_income_pdf) — "renda_media_regiao": 1890.45, "renda_mediana_regiao": 1420.30, "renda_comparada_z_score": 4.28, "renda_comparada_cdf": 0.9991, "qtde_pessoas": 312, // — Border (border) — "ponto_contido_na_faixa_de_fronteira": true, "pais_mais_proximo": "PARAGUAI", "distancia_pais_mais_proximo_km": 1.112, "municipio_possui_cidade_gemea": true, "nome_municipio_de_fronteira": "Ponta Porã", // — Community (slum) — "favela_distancia_m": 890.2, "favela_dompp_total": 342, // — Prison (prison) — "ponto_dentro_presidio": false, "distancia_borda_m": 2340.5, "nome_estabelecimento": "Unidade Penal Ricardo Brandão" } Income z_score + cdf Is the declared income compatible with the surroundings? How atypical is it? Border strip + twin_city Is it in the border strip? Is it a twin city? Which country is closest? Community distance_m + context Continuous variable — not a label, it's distance. Context, not judgment. Prison inside + distance_m Operation originated from inside a prison facility? Which prison? Applications ## Which pipelines to apply it to? Every operation that captures an address, ZIP code, or GPS coordinate is an opportunity to add layers of territorial context to the risk decision. #### Credit Income z-score as a consistency variable. Is the declared income compatible with the region? If the z-score is +4, the income is well above the local standard — a red flag. #### Insurance Community context as a pricing variable — which can even **reduce** the premium. Border and prison data as variables for claims and fraud. #### Onboarding / KYC Registration from a twin city in the border strip? With incompatible income? Strong signal for manual review or additional verification layer. #### AML/CFT and Compliance Continuous monitoring of customers in risk zones recognized by FATF — Triple Border, twin cities, smuggling routes. Auditable variables for COAF reporting. #### Transaction fraud prevention Bank transaction from a favela in Rio — from a person who never frequents the area? Strong signal of express kidnapping. Community resident making the same transaction? Normal context. #### Context-based pricing Adjust rates, limits, and conditions based on the real socioeconomic profile of the surroundings — not just the ZIP code. Census tract granularity, not city-level. Nuance ## Community data is not a label. It's context. Favela proximity is not synonymous with risk. It's a context variable that can **increase or decrease** risk perception depending on the scenario. Two real examples show why: Less risk ### Auto insurance — vehicle parked overnight in the community When an insured vehicle is parked on the street overnight, the insurer's risk perception increases and the premium goes up. However, if the vehicle is parked in a community, the premium tends to increase **less** — because there is a strong code of respect among residents that protects property within the community. → Knowing that the customer lives in a community can **reduce** the insurance premium. More risk ### Bank fraud — express kidnapping A tourist from BH, visiting Rio, was the victim of an express kidnapping inside a favela. Forced to hand over passwords, he suffered R$ 16,000 in losses across 4 institutions. If the banks had cross-referenced the transaction GPS with community data and movement history, the transaction would have been blocked — **he never frequented that location**. → Transaction from a favela by a non-resident = strong signal of coercion. The data doesn't discriminate. It contextualizes. PrismaData's community variable is a **continuous distance in meters** , not a binary "favela / non-favela" label. Combined with the customer's movement profile, it becomes a powerful and fair risk feature — protecting both the resident and the institution. Integration ## How it works Simple integration via REST API. Send the customer's address, ZIP code, or coordinate. Receive complete territorial context to feed your risk pipeline. Operation / Registration Address, ZIP code, or device GPS PrismaData APIs /income + /border /slum + /prison lat + lng Territorial context Income, border, community, prison Risk pipeline Composite score, policy rule, model feature or AML/CFT flag REST Standard API < 200ms Typical latency Lote Batch endpoint available OpenAPI Spec + Sandbox SDK Python available AI-ready Claude Code, Cursor, Copilot PrismaData's OpenAPI documentation is compatible with coding agents like **Claude Code** , **Cursor** , and **GitHub Copilot** — which drastically reduces integration friction. Your engineering team can connect the API to the risk pipeline in minutes, not weeks. Differentiator ## Combinable layers. Updated data. Exclusive. No credit bureau or data provider offers, on a single platform, estimated income with PDF + border data + community context + prison geointelligence. At PrismaData, everything is combinable by coordinate. Monthly Income model updates via PNAD Contínua. Not Census 2022. 4 layers Income + border + community + prison. Combinable by coordinate. Continuous Distance in meters, not binary label. Z-score, not generic range. Ideal for ML models. Reference data: PNAD Contínua (IBGE), 2022 Demographic Census, SENAPPEN/SISDEPEN, official cartographic data. Proprietary models with ML inference. Prison and border geofences manually created by the PrismaData team. Governance ## Territorial data, not personal. LGPD-compliant. Auditable. #### No personal data The query key is the geographic coordinate — not CPF, name, or any personal identifier. The data returned describes the profile of a **region** , not an individual. #### LGPD-compliant There is no personal data processing in the enrichment layer. The input is a coordinate; the output is territorial context. This simplifies the compliance discussion. #### Auditable and explainable Every variable is traceable — income comes from the PDF model, border from official cartographic data, prison from proprietary geofences. Ideal for supporting human decisions and meeting regulatory requirements. Results ## More variables in the pipeline. Fewer blind spots in risk. 4 independent geospatial layers added to the pipeline 588 border municipalities now automatically identifiable 0 CPFs required — 100% enrichment by coordinate, no personal data **References:** ¹ Serasa Experian, _Fraud Attempts Indicator_ , 2024. 11.5 million attempts recorded in Brazil. ² Federal Revenue Service, _Seizure Report_ , Feb. 2025. R$ 3.76 billion in smuggling and contraband in 2024. ³ FATF, _Mutual Evaluation of Brazil_ , Dec. 2023. Triple Border recognized as an AML/CFT risk zone. ⁴ COAF / Metrópoles, Sep. 2024. 24% increase in financial intelligence reports in 2023. ⁵ Agência Brasil, Sep. 2024. 588 municipalities in the border strip, 11 million inhabitants. ## Ready to add territorial context to your risk pipeline? Talk to PrismaData. We can run a pilot with your own coordinates in minutes. [ Talk to PrismaData ]() [ Explore the API → ]()