Receipt Matching

LUMA's inbox captures receipts and invoices, then uses AI to match them to your bank transactions. No more manual filing or searching for receipts at tax time.

#How it works

1. A receipt arrives in your inbox (via email forward, connected email, or upload)
2. AI extracts the merchant, amount, and date
3. The system searches your transactions for matches
4. When confident, it attaches the receipt automatically
5. Uncertain matches are flagged for your review

#Matching intelligence

The AI uses multiple signals to find the right match:

• Amount: Receipt total vs transaction amount (exact matches score highest)
• Date: Purchase date vs transaction clearing date (accounting for banking delays)
• Merchant: Company name on receipt vs bank description
• Semantic understanding: AI understands that "AMZN" means "Amazon.com"
• Your patterns: Past matching decisions for similar merchants

#How scoring works

Each potential match gets a confidence score based on weighted factors:

#Confidence levels

Based on the confidence score, matches fall into three categories:

#Auto-matched (95%+ confidence)

The AI is highly confident this is correct. The receipt is attached automatically without requiring your review.

This happens when:

• Amount matches exactly
• Merchant name is clearly the same
• Date is within expected range
• Similar matches have been confirmed before

#Suggested match (70-95% confidence)

The AI thinks this is likely correct but wants your confirmation. You'll see these in your inbox with a "Suggested match" indicator.

Review and either:

• Confirm: The receipt is attached and the AI learns from your confirmation
• Reject: The AI learns this wasn't correct and won't suggest it again

#No match (below 70% confidence)

No transaction was found that meets the confidence threshold. The receipt stays in your inbox as "Pending."

You can:

• Wait for the transaction to sync (it might not have cleared yet)
• Manually match to any transaction
• Leave it unmatched if it doesn't need matching

#How the AI learns

LUMA's matching improves over time through several learning mechanisms:

#Merchant pattern learning

When you confirm matches, the AI learns:

• "This merchant name on receipts matches this description in bank statements"
• After 3+ confirmed matches with 90%+ accuracy, similar merchants can auto-match

#Team-specific calibration

The system adapts to your team's patterns:

• If you confirm many suggested matches, thresholds adjust to auto-match more
• If you reject suggestions, thresholds become more conservative

#Dismissed match memory

When you reject a match, the AI remembers:

• That specific receipt-transaction pair is never suggested again
• This prevents the same incorrect suggestion from reappearing

#Post-match learning

Even after matching, the AI learns from your actions:

• If you later unmatch a receipt, that counts as negative feedback
• High-confidence matches that were wrong trigger more conservative thresholds

#Getting receipts into LUMA

#Email forwarding

Every LUMA account has a unique inbox email address:

1. Go to Inbox → Settings
2. Copy your inbox email address
3. Forward receipts to this address

Pro tip: Set up automatic forwarding rules in your email client to send receipts automatically.

#Connected email

Connect Gmail or Outlook to automatically capture receipts:

1. Go to Inbox → Settings
2. Click Connect email
3. Authorize your email account
4. LUMA scans for receipts and invoices automatically

The AI identifies receipt emails and ignores newsletters, marketing, and other non-receipt content.

Set up Gmail → Set up Outlook →

#Slack integration

Share receipts directly from Slack:

1. Connect your Slack workspace in Inbox → Settings
2. Share any file to the connected channel
3. It appears in your LUMA inbox automatically

Great for teams where receipts come through Slack.

Set up Slack →

#Manual upload

Drag and drop files directly:

1. Go to Inbox
2. Click Upload or drag files into the window
3. PDFs and images are processed automatically

#Review matches

Items needing your review appear in the inbox:

1. Go to Inbox
2. Items with suggested matches show a confidence indicator
3. Click to see the suggested transaction
4. Review the match:
   • Confirm if it's correct
   • Select different to match to another transaction
   • Dismiss if it shouldn't match anything

#What you're confirming

When you confirm a match, you're saying:

• "This receipt documents this transaction"
• The receipt attaches to the transaction permanently
• The AI learns from your confirmation

#Unmatched items

If a receipt doesn't match any transaction:

#Transaction hasn't cleared yet

Bank transactions can take 1-3 days to appear. The receipt will match once the transaction syncs.

#Transaction is in a different account

The purchase might be on an account you haven't connected. Check your connected accounts in Settings → Accounts.

#Manual matching

Match any receipt to any transaction:

1. Open the unmatched receipt
2. Click Match manually
3. Search for the transaction
4. Select and confirm

#Inbox settings

#Blocklist

Block senders whose emails you don't want in your inbox:

1. Go to Inbox → Settings
2. Add email addresses or domains to the blocklist
3. Future emails from these senders are ignored

Useful for blocking:

• Marketing emails that slip through
• Newsletters
• Automated notifications that aren't receipts

#Sync settings

Control how connected email accounts are scanned:

• Frequency of scanning
• How far back to look
• Which folders to scan

#Tips for reliable matching

#Best practices

• Forward receipts promptly: Same-day is best, while the transaction is fresh
• Include the full receipt: Not cropped photos or partial screenshots
• Keep bank connections healthy: Transactions need to be available for matching
• Review suggestions regularly: Don't let the inbox pile up

#What makes a good receipt

The AI works best with receipts that clearly show:

• Merchant name
• Total amount
• Date of purchase
• (Optional) Line items

Poor quality images, partial receipts, or handwritten notes are harder to match.

#When matching fails

If matching consistently fails for certain merchants:

1. Check if transactions are syncing from that account
2. Try manually matching a few times—the AI learns from this
3. Ensure receipt images are clear and complete

#Privacy and security

• Receipts are processed securely and stored encrypted
• Only your team can access your inbox
• Connected email access is read-only—LUMA can't send emails
• You can disconnect email accounts anytime

Learn about data security →