Household — AI-Powered Personal Finance for Dutch & European Households

Every app can tell you spent €847 at Albert Heijn last month. Household goes further: it knows whether those AH runs were necessities (bread, milk, vegetables — feeding your family) or luxuries (pre-cut mango, imported cheese, that fancy wine you grabbed on impulse). It also knows whether those groceries were personal (your household) or business (you run a catering side-hustle and bought supplies).

This isn't a gimmick. It means you can answer real-life questions like: How much did we really need to spend on food versus how much we splurged? or What portion of my car expenses are tax-deductible business kilometers? No other personal finance app can do this — it's a capability you normally only find in enterprise accounting software costing hundreds per month.

Household classifies every transaction across three independent (orthogonal) dimensions simultaneously. They never conflict because they answer different questions:

expense_type

WHAT was purchased. Spans Income, Fixed Expenses (rent/mortgage, insurance, utilities), Household Expenses (groceries, cleaning supplies), Occasional Expenses (dining out, entertainment, travel), and Transfers (internal account movements). 20+ subcategories provide granularity comparable to a Chart of Accounts (COA) in double-entry bookkeeping — but fully automated.

essentiality

WHY it matters financially. Every transaction is bucketed into Need, Want, or Savings. This is the dimension that powers Household's emergency-buffer calculation (only Needs count toward your minimum reserve) and the subscription watchdog (most subscriptions are Wants, so they're flagged when your discretionary spending ratio climbs).

classification

WHO it belongs to. Personal transactions stay private; Business transactions feed the ZZP Tax Agent for quarterly Btw-aangifte. Household-scope transactions are visible to family members; Personal-scope transactions are cryptographically isolated at the database row level.

NIBUD (Nationaal Instituut voor Budgetvoorlichting) is the Dutch government-backed institute that publishes reference budgets — how much a typical household should spend on housing, food, transport, clothing, and other categories based on your income and family size. These are the same numbers Dutch municipalities, mortgage advisors, and debt counselors use.

Household is the only personal finance app with NIBUD benchmarks built in. You fill in a 30-second profile (single or couple? kids? renter or homeowner? annual income?), and Household instantly shows you how your real spending compares across 9 NIBUD categories. If you're over-benchmark on housing, you see it. If you're under-benchmark on groceries (good job!), you see that too.

Even more valuable: the Toeslagen radar checks whether your household profile makes you eligible for any of 4 Dutch government subsidies — Kinderopvangtoeslag (childcare), Kindgebonden Budget (child benefit supplement), Huurtoeslag (rent subsidy), and Zorgtoeslag (healthcare allowance). These are benefits many Dutch families qualify for but never claim because they don't know they're eligible. Household tells you.

NIBUD publishes reference budgets annually based on CBS (Centraal Bureau voor de Statistiek) consumption data. Household maps your actual categorized spending onto NIBUD's 9-category framework using the official NIBUD household-type matrix: Single, Single Parent, Couple, Couple with Kids — cross-referenced against housing status (owner/renter) and income bracket.

Toeslagen eligibility is calculated against current Belastingdienst thresholds for the calendar year. The radar checks income caps, asset limits (vermogenstoets), child ages, and rent ceilings (huurtoeslag maximale huur). Results are advisory, not a guarantee — always verify with the Belastingdienst before claiming.

This is the part that makes Household unlike anything else. Most finance apps are passive — they show you what already happened. Household is active — it has 10 autonomous agents running in the background, each watching for something specific:

• Subscription Watchdog: Notices when Netflix, Spotify, or your insurance raises prices. Tracks contract renewal dates so you never auto-renew at a higher rate. Even generates a Dutch Wet van Dam cancellation letter if you want out. Average user savings: €180/year.
• NIBUD & Toeslagen Radar: Continuously benchmarks your spending against official norms and re-checks subsidy eligibility when your income or family situation changes.
• Emergency Buffer Agent: Calculates how many months you could survive on savings if income stopped tomorrow, based on your actual essential spending. NIBUD recommends 3-6 months — Household shows exactly where you stand and how much more you need.
• Cash-Drag Detector: Finds money sitting idle in zero-interest checking accounts and calculates what you'd earn by moving it to a savings vault. Most users discover €2,000-5,000 of idle cash within their first month.
• ZZP Tax & VAT Agent: For freelancers: estimates quarterly Btw-aangifte at 21%, 9%, and 0% VAT tiers. Identifies deductible expenses (kilometervergoeding at €0.23/km, home office costs, professional subscriptions) and summarizes them for your boekhouder.
• Scenario Simulator: What-if projections over 1-36 months. Income drops by 30%? Child #3 arrives? Six-month sabbatical? Buying a house? See how each scenario impacts your net wealth trajectory before you commit.
• Tikkie Settlement Agent: Tracks shared expenses with your partner or housemates, calculates 50/50 splits, and generates payment links for unsettled balances. No more I'll-get-the-next-one ambiguity.
• Receipt Matcher: When you scan a receipt, this agent cross-references every line item against your bank transactions to verify every charge. Flags discrepancies so you catch billing errors (they're more common than you think).
• SEPA Parser: Native CAMT.053 XML and SWIFT MT940 parsing — the same formats banks use internally. Handles Dutch (ING, Rabobank, ABN AMRO, SNS, ASN, Triodos, Bunq) and international banks (Revolut, N26).
• Export Agent: Generates mortgage-ready Financieel Overzicht PDFs (income statements, asset summaries, categorized annual spending) formatted for Dutch hypotheekaanvraag submissions. Also handles GDPR data export in CSV and JSON.

Each agent is a stateless rule evaluator with configurable thresholds. The built-in agents use deterministic heuristics (regex-based merchant matching, numeric threshold comparisons, date-range arithmetic) as their primary path, falling back to LLM augmentation only for ambiguous cases like merchant name normalization or natural-language scenario interpretation. Custom agents created through the Agent Center accept plain-English rules that are compiled into structured monitoring queries.

Agents fire on a scheduled cycle (hourly for built-ins, daily for customs) and produce structured insight records with severity levels (info, warning, alert). Alert-level insights surface on the Dashboard and optionally via email. Every agent can be individually toggled on/off from the Agent Center — you control exactly what watches your finances.

When you upload a bank statement, here's the exact path it takes — and what doesn't happen:

1. Local redaction first. Before any text leaves your browser, Household strips out your name, IBANs, addresses, phone numbers, and email addresses. What gets sent to the server is an anonymized financial record — amounts, dates, merchant names — but no personally identifiable information.
2. Processing happens. The server runs the parsing engine (heuristic extraction + optional LLM fallback for complex layouts) on the redacted data. If an LLM is involved, the commercial API terms contractually prohibit the provider from using your data for model training.
3. Original values are restored locally. The redaction mapping is ephemeral — it exists only in your browser session and is never persisted to disk on any server.
4. Your data sits in PostgreSQL with Row-Level Security. Every row in the database is tagged with your household ID, and the database engine itself enforces isolation. Even if someone gained direct database access, they couldn't read across households.

No advertising networks. No data brokers. No analytics platforms. No social media pixels. No tracking. Zero.

• On-Device PII Redaction: Regex + NER-based pipeline runs in-browser before any network request. Strips IBANs (NLxx-format), BSNs, names from statement headers, addresses, phone numbers, and email addresses.
• Zero Model Training: Commercial developer API terms contractually prohibit training. The ephemeral redaction map is held in browser memory only.
• PostgreSQL Row-Level Security: Multi-tenant isolation at the database engine level. Each query is transparently filtered by household_id.
• GDPR Self-Service: One-click data export (Art. 20), one-click account deletion (Art. 17), per-agent consent toggles (Art. 7(3)). No email to support required.
• Authentication: JWT (HS256) via AWS Cognito, bcrypt password hashing (cost factor 12), optional email-based 2FA.
• Infrastructure: Docker containers on AWS Lambda, Alembic-migrated PostgreSQL, TLS 1.3 everywhere.

Instead of typing in every transaction by hand (nobody has time for that), you download your bank statement from your bank's website or app and drop it into Household. That's it. The platform reads the file — whether it's a PDF, CSV, Excel file, or a specialized banking format — and extracts every transaction automatically. Duplicates are detected and skipped so you never count the same expense twice.

Most uploads are parsed in under 30 seconds. For complex multi-page PDFs from banks with unusual formatting, the AI fallback takes a minute or two. You'll see a progress bar and can navigate away — processing continues in the background.

• Native formats: SEPA CAMT.053 XML (ISO 20022), SWIFT MT940 text
• Structured formats: CSV, TSV, XLSX, OFX/QFX, QIF
• Unstructured: PDF bank statements (multi-page, multi-column) via LLM-powered chunked parsing
• Deduplication: SHA-256 file hashing prevents re-import of identical files. Transaction-level deduplication by (merchant, date, amount) triplet skips overlapping entries across statement boundaries.
• Progressive parsing: Heuristic extraction first (regex + column mapping for structured formats, PDF text extraction + pattern matching). LLM fallback kicks in only when heuristics fail — minimizing API costs and latency.

Snap a photo of any receipt or invoice — grocery store, restaurant, hardware store, utility bill — and upload it. Household's OCR engine reads every line item individually: the product name, the quantity, the price, and any VAT.

• Splitting a supermarket run into food vs. household supplies
• Separating business meals from personal dining (for ZZP tax deductions)
• Verifying that what appeared on your bank statement matches what's on the receipt

Uploaded receipts are automatically matched to bank transactions within a 7-day window (±€0.05 tolerance). Manual linking is also supported if you prefer to do it yourself.

• Supports JPG, PNG, and PDF receipt/invoice uploads.
• Extracts: merchant/store name, transaction date, total amount, VAT breakdown, individual line items with quantities and unit prices.
• Each extracted line item receives full 3D categorization (expense_type, essentiality, classification).
• Auto-reconciliation: (±€0.05, ±7 days) against bank transactions.
• Scope tagging: line items inherit the upload scope (Household or Personal).

Household uses zero-based envelope budgeting — the same method used by YNAB and professional financial planners. The core idea is simple: every euro you earn gets assigned a job before you spend it. You create budget envelopes for each category (Rent: €1,200, Groceries: €600, Entertainment: €200, etc.) and Household tracks your progress in real time.

If this sounds like work, don't worry — the AI auto-budget feature fills in your envelope limits based on your actual spending history. One click, and your budget is set. You can tweak from there. You can also create multiple budget scenarios: a standard month, a frugal month, a vacation month, a holiday month — and switch between them instantly.

The wealth projection chart shows where your net wealth is heading over the next 12-36 months based on your current savings rate, with compound growth on any investments you've configured (ETFs at 7% expected return, savings accounts at 2.5%).

• Envelope allocation: Zero-based budgeting with per-category monthly limits. Real-time progress bars with color-coded alerts at 80% (amber) and 100% (red) of cap.
• Rollover: Unspent envelope balances roll forward month-to-month within the calendar year. Overspending is tracked as negative carry-forward.
• Auto-budget: AI derives initial limits from trailing 3-month average spending per category. Copy-last-month and Reset-all quick actions available.
• Multi-scenario: Create and switch between named budget scenarios. Each scenario maintains independent envelope allocations.
• Wealth projection: Linear forward projection based on trailing 3-month average net savings rate, compounded with current asset balances. Investment assets use configurable expected APY (default: 2.5% savings, 7.0% equity ETFs). Does not account for market volatility, inflation, or one-off life events — use the Scenario Simulator agent for stochastic modeling.
• NIBUD benchmark overlay: Budget categories are mapped to NIBUD reference categories for comparison against official Dutch household spending norms.

The Transactions page works like a search engine for your money. Type a merchant name, a category, or even an amount, and results appear instantly. Filter by date range, by account, by income/expense, by scope (household or personal). One-click CSV export when you need the data outside the app.

• Real-time keyword search across merchant, description, and category fields.
• Type tabs: All, Income, Expenses, Fixed, Household, Occasional, Transfers.
• 3D category popover for dimension-aware filtering.
• Account selector, date range presets (this month, last month, last 3 months, YTD, custom).
• 12 badge filters: Household, Personal, Shared, Recurring, Uncategorized, VAT Deductible, Flagged, Large, Tikkie, Cash Withdrawal, Refund, Internal Transfer.
• Active filter counter with one-click clear. Bulk select, recategorize, and delete. Export to CSV or JSON with all dimensions preserved.

If you share finances with a partner or family, Household gives you three visibility levels for every transaction and account:

Household scope: Visible to everyone in the family. Use this for joint accounts, shared expenses, and family budgeting. | Personal scope: Private to you only. Use this for your personal account, your side-hustle income, or anything you want to keep separate. | Shared/Both: Visible in both scopes. Use this for joint accounts that both partners need to see.

Switching between scopes is instant — there's a toggle at the top of every page. Separate dashboards, separate budgets, separate category trees. The database enforces isolation so personal transactions never accidentally leak into the household view.

• Scopes are enforced at the database level via PostgreSQL Row-Level Security policies.
• Transactions, categories, budgets, accounts, and invoices are all scope-tagged.
• Scope is confirmed during upload — the same statement file can be uploaded to Household scope (shared) or Personal scope (private).
• Scope badges render inline on every transaction row, category tree node, and account card.
• Category assignments are scope-specific: the same merchant can map to different categories in different scopes.

Household comes with sensible default categories (Rent, Groceries, Transport, Entertainment, etc.) but you can customize everything. Create parent-child category trees (e.g., Transport → Train, Car, Bike). Create custom dimensions to track things your way — tag expenses by project (Home Renovation 2026), by tax class, or by any other lens that matters to you.

The multi-month comparison view shows how your spending in each category changes month over month in a clean, scannable table. Spot trends immediately: Our dining-out spending doubled since April — let's rein that in.

• Parent-child hierarchy with drag-and-drop reparenting.
• Custom dimensions: independent orthogonal tagging axes beyond the built-in 3D system.
• Category merge: reassigns all transactions, budgets, and subcategories with impact preview.
• AI-suggested subcategories based on merchant names in your transaction history.
• Recategorize-all: apply default rules and AI to all transactions in bulk.
• Multi-month comparison table: reverse-chronological month columns with smart baseline-to-latest net delta calculations. Monthly and yearly aggregation views.