When you start with Parseur, one of the most common questions is:
“Do I need one mailbox per document layout/vendor, or can everything go into one mailbox?”
In Parseur, a mailbox defines a single parsing configuration:
the data schema you want to extract (fields / table data)
the parsing instructions used by AI (and/or templates)
the export configuration (destination, mappings, etc.)
Because of that, the best number of mailboxes depends less on how many vendors you have, and more on whether you need different outputs (schema) or different destinations (exports).
The recommendation for most users: fewer mailboxes = more automation
Most customers should aim for as few mailboxes as possible, because multiple mailboxes add ongoing maintenance:
If you want to change a field, export mapping, or setting later, you must update it in every mailbox.
Having many mailboxes with the same schema and same destination can become hard to manage at scale.
You can scale to multiple mailboxes when needed, but we recommend exhausting simpler approaches first.
Step 1: Always separate by document type
Different document types should always go into different mailboxes, because they usually require different fields and instructions.
For example, these should be separated:
Receipts
Purchase orders
Invoices
Bank statements
Tax documents
Shipping labels / bills of lading
Even if two document types look similar, they usually require different logic, different fields, and different parsing behavior - mixing them tends to reduce accuracy.
organize different document types into different mailboxes as the schema will always differ between them
Step 2: Decide based on output schema and destination
Use these recommendations in order:
1) Same data schema + same destination? Use one mailbox (recommended)
Use one mailbox when:
You extract the same fields/table data from multiple layouts
You send the data to the same endpoint/export
You want the easiest setup and the most automation
This is the setup we recommend for most users.
Example: If you receive invoices from 25 vendors with different layouts, but you always extract the same fields and send them to the same system, use one “Invoices” mailbox.
2) Different data schemas? Create one mailbox per schema
Use multiple mailboxes when:
Different layouts require extracting different sets of fields
You need a different table structure (e.g., line items vs no line items)
The downstream system expects different outputs
Example: Some invoices require line items and others only need totals, in that case you may want separate “Invoices (Summary)” and “Invoices (Line Items)” mailboxes.
3) Different destinations/exports? Create one mailbox per destination
Use multiple mailboxes when:
Documents need to be routed to different endpoints (different webhooks, different apps, different exports)
Each destination requires different export settings, mapping, or logic
Example: If one set of documents goes to NetSuite, and another goes to a data warehouse, separating mailboxes can simplify export configuration.
4) Only if accuracy isn’t acceptable, consider one mailbox per layout (last resort)
Sometimes, multiple layouts in one mailbox can create conflicting instructions (especially after you add more field-level customization over time). If you can’t achieve acceptable results with one mailbox + templates, then splitting can help.
Consider splitting by layout (or subsets of layouts) when:
You’ve tried AI instructions + templates for outliers, but parsing is still unreliable
Layouts are extremely different and require very specific instructions
You need maximum consistency for critical workflows
Important caveat:
Creating many mailboxes that all have the same schema and same destination should generally be avoided unless it’s truly needed for accuracy. It increases maintenance (fields/settings changes must be repeated everywhere).
5) Processing documents on behalf of customers? Create one mailbox per customer
This is a common “power user” pattern when you need any of the following:
Give each customer access to their own mailbox
Track credit usage per customer
Use different export destinations/settings per customer
Example: A service provider creates one mailbox per client so each client can log in and see their own documents, and exports can be customized per client.
(For advanced setups like this, see our article on master mailboxes.)
Where “one mailbox per vendor” fits in
A common misconception is that the “best” setup is one mailbox per vendor. In practice:
If you extract the same schema and export to the same destination, one mailbox per vendor often becomes unnecessary overhead.
It’s usually better to keep a single mailbox and handle differences with AI + templates for outliers.
Only split by vendor/layout when you’ve confirmed that combined layouts produce instruction conflicts you can’t resolve.
A practical path to get started
Create one mailbox per document type.
Start with one mailbox for that document type if schema + destination are the same.
Add templates for outliers as you encounter them.
Split into additional mailboxes only when you need:
a different schema, or
a different destination, or
a last-resort accuracy improvement because layouts conflict.
If you have any questions or want confirmation on best practices for your use case, please feel free to reach out.
Related articles: