PensionBrug

RSM Fintech Business Models & Applications, 2026 Amalia Nimeskern & Camila Munoz

What is PensionBrug?

PensionBrug is a pension aggregation MVP that brings together all three pillars of the Dutch pension system into a single personalised dashboard. It detects retirement income gaps, runs a Monte Carlo simulation of the Pillar 2 pot under the WTP contribution system, and recommends Pillar 3 products to close any gap.

The product follows the same strategic logic as Tink before PSD2: building the aggregation layer and the user trust ahead of the regulatory mandate (FiDA) that will eventually require pension funds to open their data.

Features

• Pillar 1, AOW: computed from user inputs (age arrived in the Netherlands, years spent abroad), based on the official 2% per insured year formula.
• Pillar 2, Employer Pension: UPO PDF upload with automatic parsing.
• Gap Detection: compares the total projected pension against your chosen retirement income target.
• Monte Carlo Simulation: 10,000 simulations of the Pillar 2 payout, reflecting the investment uncertainty introduced by the WTP shift from defined benefit to defined contribution.
• Risk Classification: On Track, At Risk, or Critical Gap, based on coverage percentage and probability of meeting target.
• Pillar 3 Recommendations: estimates the monthly contribution needed to close the gap and shows matching providers.
• Pension Chatbot: rule based question and answer assistant covering AOW, WTP, Pillar 2 and 3, gaps, and expat specific topics. No external AI API is used.

Architecture

pensionbrug/
  app.py                Streamlit UI, main application and styling
  pillar1.py            AOW calculation logic
  pillar2.py            UPO PDF parsing (pdfplumber and regex)
  gap.py                Gap detection and Monte Carlo simulation (numpy)
  providers.py          Pillar 3 recommendation logic
  chatbot.py            Rule based pension question and answer assistant
  create_sample_upo.py  Generates a sample UPO PDF for testing
  assets/logo.png        PensionBrug logo, shown in the header and sidebar
  requirements.txt      Python dependencies
  CLAUDE.md              AI agent instructions

Architecture

• Each pillar has its own module (pillar1.py, pillar2.py, and the Pillar 3 logic in providers.py) that normalises pension data from a different source into a common annual euro figure.
• gap.py is the intelligence layer. It takes the three normalised figures and the user's target, and produces a gap, a risk classification, and a Monte Carlo range.
• app.py is the user facing layer, structured as a six step guided flow (About You, Employer Pension, Retirement Goal, Your Pension Picture, Close the Gap, Ask PensionBrug), so the UX mirrors the conceptual journey from raw pension data to a personalised plan.

---

How to Run Locally

Prerequisites: Python 3.10 or later, pip

git clone https://github.com/cmunoz-hello/pensionbrug.git
cd pensionbrug
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
streamlit run app.py

The app opens at http://localhost:8501

---

How to Generate a Sample UPO

python3 create_sample_upo.py

This creates sample_upo.pdf, a realistic Dutch pension statement in the standardised UPO format, which you can upload in Step 2 to test the parsing.

---

Tech Stack

Component	Technology
UI	Streamlit
Styling	Custom CSS, Google Fonts (Poppins)
PDF parsing	pdfplumber and regex
Gap analysis and Monte Carlo	numpy
Charts	Plotly
Sample data generation	reportlab
AI coding agent	Claude (Anthropic)

---