Spending analysis

Spending Analysis

Purpose

Provides rules for analyzing categorized transaction data — aggregating spending by category, comparing month-over-month, detecting unusual transactions, and generating savings recommendations.

Category Aggregation

For a given month, aggregate all categorized transactions:

1. Group transactions by category
2. For each category, compute:
   • total: Sum of all transaction amounts (absolute values for expenses)
   • transaction_count: Number of transactions
   • avg_transaction: total / transaction_count
3. Calculate each category's percentage of total spending: category_total / total_spending × 100
4. Sort categories by total (highest first)

Write results to monthly-summary.csv.

Receipt-Level Breakdown

For a given merchant, aggregate spending at the receipt-item level instead of the transaction level. Used when the user asks about spending at a specific merchant (e.g., "Hvad bruger jeg hos Bilka?").

Data Source

1. Read receipts.csv, filter by merchant (case-insensitive match against normalized merchant name per skills/categorization/SKILL.md)
2. If month argument is provided, also filter by date column (YYYY-MM prefix match)
3. Collect all matching receipt_id values and their date values
4. Determine which monthly files to read: extract unique YYYY-MM values from the receipt dates
5. For each monthly file receipt-items-{YYYY-MM}.csv: read it if it exists, filter to rows where receipt_id is in the collected set. Skip missing files.
6. Combine all matching rows

Aggregation Rules

Group matching receipt items by subcategory:

1. For each subcategory, compute:
   • total: Sum of total_price for all items in that subcategory
   • item_count: Sum of quantity for all items in that subcategory
   • avg_price: total / item_count
2. Calculate each subcategory's percentage of total merchant spending: subcategory_total / merchant_total × 100
3. Sort subcategories by total (highest first)

Top Items per Subcategory

For each subcategory, identify the top 3 most purchased individual items:

1. Group items by item_name within the subcategory
2. For each item, compute:
   • purchase_count: Sum of quantity across all receipts
   • item_total: Sum of total_price across all receipts
3. Sort by purchase_count (highest first), break ties by item_total
4. Return top 3

Overall Top Items

Across all subcategories, identify the top 3 most frequently purchased items using the same logic as above but without the subcategory grouping.

Multi-Month Handling

• Month specified: Show totals for that month only. Header: {Merchant} — Indkøbsoversigt {month_name} {year}
• No month specified: Show all-time totals plus a per-month average. Header: {Merchant} — Indkøbsoversigt samlet
  • Per-month average: total / number_of_distinct_months (count distinct YYYY-MM values from matching receipts)

Output Format: Merchant Breakdown

## {Merchant} — Indkøbsoversigt {month or "samlet"}

**Antal kvitteringer**: {count}
**Samlet forbrug**: {total} kr

### Fordeling pr. varekategori
| Kategori | Beløb | Andel | Antal varer |
|----------|-------|-------|-------------|
| {subcategory} | {amount} kr | {pct}% | {item_count} |
| ... | ... | ... | ... |

### Hyppigst købte varer
1. {item_name} — {count}x — {total} kr
2. {item_name} — {count}x — {total} kr
3. {item_name} — {count}x — {total} kr

If no month is specified, append a summary line after the table:

**Gennemsnit pr. måned**: {avg} kr ({month_count} måneder)

Month-Over-Month Comparison

Compare the current month's spending to the previous month:

1. For each category present in either month:
   • vs_prev_month: current_total - previous_total (absolute difference in kr)
   • vs_prev_month_pct: (current_total - previous_total) / previous_total × 100
2. Flag categories with significant changes:
   • Increase > 20%: highlight as notable increase
   • Decrease > 20%: highlight as notable decrease
   • New category (not in previous month): flag as new spending area

If there is no previous month data (first month of tracking), skip the comparison.

Unusual Transaction Detection

Flag individual transactions that stand out:

By Amount

• Transaction amount > 3× the average for its category
• Single transaction > 5.000 kr (unless the category is Bolig or Indkomst)

By Frequency

• Multiple transactions to the same merchant on the same day (possible duplicate charges)
• Weekend transactions to work-related merchants

By Category Mix

• A merchant that usually falls in one category appearing in a different one

Savings Recommendations

Generate actionable suggestions based on analysis:

Subscription Savings

• Unused subscriptions: Subscriptions where status: active but no other transactions to that merchant's category (e.g., Viaplay active but no streaming usage signals)
• Duplicate services: Multiple subscriptions in the same subcategory (e.g., Netflix + HBO Max + Viaplay = 3 streaming services)
• Price increases: Subscription amount increased > 10% from the average

Category Savings

• High dining spend: If Restauranter > 15% of total, suggest cooking more
• Transport optimization: If Taxi subcategory > Transport total × 30%, suggest alternatives
• Impulse shopping: If Shopping has many small transactions (> 10 in a month, avg < 200 kr), flag as potential impulse purchases

Formatting Suggestions

Present each suggestion as:

{action} — Spar {amount} kr/år

Example: "Opsig Viaplay — Spar 1.188 kr/år"

Danish Currency Formatting

All user-facing monetary values must use Danish formatting:

• Thousands separator: period (.)
• No decimal places for whole amounts in casual output
• Suffix: kr (with space before)
• Examples:
  • 1.847 kr (one thousand eight hundred forty-seven)
  • 28.450 kr (twenty-eight thousand four hundred fifty)
  • 149 kr (no thousands separator needed)
  • 4.200 kr (four thousand two hundred)

Output Format: Spending Overview

When presenting a spending overview, use this structure:

## {Month} {Year} — Overblik

**Samlet forbrug**: {total} kr

### Fordeling pr. kategori
| Kategori | Beløb | Andel | Ændr. |
|----------|-------|-------|-------|
| {category} | {amount} kr | {pct}% | {change} |
| ... | ... | ... | ... |

### Abonnementer ({count} aktive)
| Tjeneste | Månedlig | Årlig | Status |
|----------|----------|-------|--------|
| {service} | {amount} kr | {annual} kr | {status} |
| ... | ... | ... | ... |
**Total abonnementer**: {total} kr/måned ({annual_total} kr/år)

### Forslag til besparelser
1. {suggestion} — Spar {amount} kr/år
2. {suggestion} — Spar {amount} kr/år

Output Format: Monthly Report

When generating a detailed report, include additional sections:

## {Month} {Year} — Månedlig rapport

### Resumé
{2-3 sentence summary of the month's spending}

### Forbrug pr. kategori
{Category breakdown table as above}

### Sammenlignet med {previous_month}
{Highlight notable changes — increases, decreases, new categories}

### Usædvanlige transaktioner
{List any flagged transactions with amounts and reasons}

### Abonnementer
{Subscription table as above}
{Note any changes: new subscriptions, cancellations, price changes}

### Handlinger i denne måned
{List actions taken from the Action Log: syncs, cancellations, etc.}

### Anbefalinger
{Savings suggestions with estimated annual impact}

Examples

Category Aggregation Example

Given 5 transactions in January 2026:

Netto        -347.50  Dagligvarer
Føtex        -289.00  Dagligvarer
Netflix      -149.00  Abonnementer
DSB          -72.00   Transport
Wolt         -198.00  Restauranter

Monthly Summary rows:

month: 2026-01, category: Dagligvarer, total: 636.50, count: 2, avg: 318.25
month: 2026-01, category: Restauranter, total: 198.00, count: 1, avg: 198.00
month: 2026-01, category: Abonnementer, total: 149.00, count: 1, avg: 149.00
month: 2026-01, category: Transport, total: 72.00, count: 1, avg: 72.00

Savings Recommendation Example

Detected 3 streaming subscriptions:
- Netflix: 149 kr/måned (1.788 kr/år)
- HBO Max: 149 kr/måned (1.788 kr/år)
- Viaplay: 99 kr/måned (1.188 kr/år)
Total: 397 kr/måned (4.764 kr/år)

Suggestion: "Overvej at konsolidere streamingtjenester? Netflix + HBO Max + Viaplay = 397 kr/måned"