Automatic Categorization

How LUMA uses AI to categorize your transactions and how categories affect your metrics.

Every transaction that flows into LUMA is automatically categorized using AI. This saves hours of manual work and keeps your books organized for accounting.

#How it works

When a new transaction arrives, LUMA's AI:

  1. Analyzes the merchant: Identifies the company from the transaction description
  2. Checks your history: Looks at how you've categorized similar transactions
  3. Considers patterns: Factors in amount, frequency, and context
  4. Assigns a category: Picks the most likely category

The AI learns from your corrections. The more you use LUMA, the more accurate it becomes for your specific business.

#Category structure

LUMA uses a two-level category system designed for business accounting:

#Parent categories (14 total)

  1. Revenue - Income from your business
  2. Cost of Goods Sold - Direct costs of products sold
  3. Sales & Marketing - Customer acquisition costs
  4. Operations - Day-to-day running costs
  5. Professional Services - External expertise
  6. Human Resources - Team costs
  7. Travel & Entertainment - Business travel
  8. Technology - Software and subscriptions
  9. Banking & Finance - Financial transactions
  10. Assets - Capital purchases
  11. Liabilities & Debt - Obligations
  12. Taxes & Government - Tax payments
  13. Owner / Equity - Founder transactions
  14. System - Uncategorized and transfers

#Subcategories

Each parent category contains specific subcategories. For example, Technology includes:

  • Software
  • Non-Software Subscriptions

See all categories and subcategories →

#Excluded categories

Some categories are "excluded" from burn rate calculations to prevent double-counting:

CategoryWhy it's excluded
Credit Card PaymentThe purchases on your card already count; the payment would be double-counting
Internal TransferMoving money between accounts isn't spending

This happens automatically. Your burn rate and runway calculations stay accurate.

Learn more about how categories affect metrics →

#Changing a category

If the AI gets it wrong:

  1. Click on the transaction
  2. Select the correct category from the dropdown
  3. Save

#What happens when you correct

  • The transaction is recategorized immediately
  • The AI learns from your correction
  • Future similar transactions are more likely to be categorized correctly
  • Your financial metrics update to reflect the change

#Bulk categorization

For multiple similar transactions:

  1. Go to Transactions
  2. Use filters to find the transactions (e.g., filter by merchant name)
  3. Select multiple transactions using checkboxes
  4. Choose Categorize from the bulk actions menu
  5. Select the category to apply
  6. Confirm

This is useful for:

  • Fixing multiple miscategorized transactions at once
  • Categorizing a backlog of transactions
  • Applying consistent categories to recurring expenses

#How the AI learns

#Merchant patterns

The AI builds patterns based on merchant names:

  • "SPOTIFY" → Technology > Software
  • "UBER" → Travel & Entertainment > Travel
  • "AMAZON" → Depends on your history (could be office supplies, software, etc.)

#Your corrections

When you correct a category:

  • That specific merchant pattern is updated
  • Similar merchant names benefit from the learning
  • The AI weighs your choices higher over time

#Team patterns

The AI learns from your entire team's categorization choices, building a model specific to your business.

#Categories and financial metrics

How you categorize transactions directly affects your reports:

MetricAffected by
RevenueTransactions in Revenue categories
Burn RateAll expense categories except excluded ones
ProfitRevenue minus expenses
Cash FlowAll income and expense categories

Accurate categorization means accurate metrics.

#Categories and accounting

#Export to CSV

When you export transactions, categories are included. Your accountant can map them to their chart of accounts.

#Export to accounting software

If you connect Xero, QuickBooks, or Fortnox:

  • Categories map to account codes
  • You configure the mapping during setup
  • Transactions export with the correct account mappings

#Tips for better categorization

#Review regularly

  • Check new transactions weekly
  • Correct any miscategorized items promptly
  • The AI learns faster with quick feedback

#Be consistent

  • Use the same category for the same type of expense
  • Don't switch between similar categories (pick one)
  • Consistency helps both AI learning and your reports

#Use tags for detail

Categories are intentionally broad (for accounting). Use tags for additional detail:

  • Project names
  • Client names
  • Tax deduction status
  • Department codes

#Handle ambiguous merchants

Some merchants (like Amazon) could be many categories:

  • Office supplies
  • Software
  • Equipment
  • Personal (shouldn't be here!)

Correct these carefully. The AI will learn that Amazon transactions from your account might be different from another business.

#Common questions

#Can I create custom categories?

LUMA uses a standard category set that maps to accounting charts of accounts. This ensures consistency for exports and integrations. Use tags for custom groupings.

#What if a transaction fits multiple categories?

Choose the primary purpose. For example:

  • A laptop is Equipment, not Technology
  • Conference travel is Travel, not Marketing

#How do I handle personal transactions?

If a personal transaction accidentally appears:

  1. Categorize it appropriately (or use "Other")
  2. Consider excluding it from reports
  3. Discuss with your accountant for proper handling

#Why did the AI categorize this wrong?

Common reasons:

  • New merchant the AI hasn't seen
  • Ambiguous merchant name
  • Similar merchant to one you categorized differently

Correct it once, and the AI will learn.

Full category reference → How categories affect metrics →