Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.dualentry.com/llms.txt

Use this file to discover all available pages before exploring further.

​
How to Use Automated Categorization

When transactions arrive in DualEntry - through a bank feed, OCR upload, or bulk import - the system suggests a GL account and dimensions for each line. Suggestions draw from rules you define, patterns learned from your prior corrections, and the transaction description or memo. High-confidence suggestions apply automatically so you spend your time on the exceptions rather than the routine.

​
How the Suggestion Engine Works

DualEntry evaluates each incoming transaction against three layers of logic, applied in order of priority:
  1. Explicit rules - You create rules that map a specific condition to an account and dimension set. For example, “any transaction where the vendor is Acme Corp maps to account 6200, department Marketing.” Rules take absolute precedence; if a rule matches, the suggestion engine uses it and stops.
  2. Learned patterns - When no rule matches, DualEntry looks at how you categorized similar transactions in the past. The model weighs vendor name, memo text, amount range, and your prior account choices.
  3. Description analysis - For completely new vendors or unusual memos, the engine parses the transaction description and matches keywords to your chart of accounts names and aliases.
Each suggestion carries a confidence score. You decide what happens at each confidence level through the configuration described below.

​
Configuring Categorization Settings

Navigate to Settings → AI → Categorization to control how automated suggestions behave. The key options are:
  • Auto-apply threshold - Set a confidence percentage between 70% and 95%. Suggestions at or above this threshold apply without manual review. The default is 85%.
  • Enable/disable auto-apply - Toggle auto-apply off entirely if you prefer to review every suggestion before it takes effect.
  • Dimension suggestions - Choose which dimensions the engine suggests alongside the GL account. You can enable or disable suggestions for department, class, location, or any custom dimension.
Changes to these settings apply to all new incoming transactions. Existing transactions already in the review queue are not retroactively affected. If you lower the auto-apply threshold after initial setup, previously queued transactions that now meet the new threshold remain in the queue - you approve or reject them manually. You can test different threshold levels by reviewing the confidence scores shown in the queue; if most queued items have confidence above your target threshold, raising the auto-apply level is safe.

​
Reviewing Suggestions in the Review Queue

Transactions that fall below the auto-apply threshold land in the Review Queue. Navigate to Transactions → Review Queue to see them. Each row shows the transaction details, the suggested account and dimensions, the confidence score, and the current (or blank) categorization. You approve a suggestion with a single click or override it by selecting a different account. The queue supports bulk approval - select multiple rows and approve them all at once when the suggestions look correct. You can also filter the queue by confidence range, vendor, or date to focus on a specific batch. Sorting by confidence (lowest first) is a practical approach: address the uncertain items first and then bulk-approve the rest.
Overrides train the model. When you correct a suggestion, DualEntry records the correction and weighs it heavily for future transactions with similar attributes. Over time, the engine becomes more accurate for your specific data patterns.

​
Working with Bank Feed Rules and Templates

Automated categorization runs alongside bank feed rules and recurring transaction templates. If a bank feed rule already matches an incoming transaction, that rule takes priority and the categorization engine does not override it. For transactions without a bank feed rule or template match, the AI suggestion engine fills the gap. This layered approach means you do not need to choose between manual rules and AI suggestions - they work together, with explicit rules always winning. When you create a new bank feed rule for a vendor that the AI previously handled, the rule takes over immediately on the next import. The AI model retains its learned pattern in case the rule is later removed, so there is no loss of training data when you switch between approaches.
Last modified on May 28, 2026