Deluge is Zoho's scripting language and the layer where most real customisation happens. This is a non-technical guide to what it does well, where it stops, and the failure modes worth knowing about before you rely on it.
What Deluge is for
| Trigger | Typical use |
|---|---|
| On record create or edit | Validation, field calculation, related record updates |
| On a schedule | Nightly sync, periodic cleanup, reminder generation |
| On button click | Custom action a user invokes deliberately |
| On workflow rule | Logic too complex for the standard actions |
| Via webhook | Responding to an event in another system |
What it handles well
- Cross-module logic — updating related records when something changes.
- Calculations beyond what formula fields express.
- Validation with real conditions — checking against other records or external data.
- API calls to external systems, in both directions.
- Scheduled processing — anything that should happen nightly or weekly.
- Data transformation between formats when integrating.
Where it hits limits
| Limit | What it means practically |
|---|---|
| Execution time per invocation | Long-running processing must be chunked |
| Statement count | Large loops fail; batch instead |
| API calls per execution | Per-record external calls do not scale |
| Limited debugging | Diagnosing failures is harder than in normal code |
| No version control natively | Changes are not tracked unless you manage it |
| Silent failure by default | Errors do not surface unless you make them |
Silent failure is the problem that matters most. A Deluge function that errors typically does so without notifying anyone. We have seen setups where a nightly sync stopped working a month earlier and was discovered only when someone questioned a report. Error handling that alerts someone is not optional — it is what makes the automation trustworthy.
The bulk problem
Logic that works perfectly in testing frequently fails on a real data load, for a specific reason: it was written to handle one record at a time.
- A function making one API call per record works for ten records and exhausts limits at five hundred.
- A loop over related records hits statement limits on large datasets.
- Scheduled functions processing everything at once time out as data grows.
Test with production-scale data, not a handful of sample records. The behaviour is genuinely different, and discovering that during a migration is a bad time.
The pattern that avoids most of this: process in batches with a scheduled function that picks up where it left off, rather than attempting everything in one execution. It is slightly more code and it scales indefinitely rather than failing at an unpredictable threshold.
When to move logic out of Deluge
- Heavy processing that repeatedly hits execution limits.
- Complex integration needing retry logic, queuing and reconciliation.
- Logic you need to test properly — Deluge testing is limited.
- Anything business-critical where silent failure is unacceptable.
- Code that must be version-controlled and reviewed.
The alternative is an external service that Zoho calls, or that polls Zoho — giving you proper error handling, logging, testing and version control while Zoho remains the system of record.
Making Deluge maintainable
- Add error handling everywhere, with notification on failure. This is the highest-value practice.
- Log what happened — a custom module recording executions is crude and effective.
- Keep a copy in version control outside Zoho, even manually.
- Comment the business reason, not the syntax. Six months later, why matters more than what.
- Test in a sandbox with realistic data volumes.
- Document what exists — a list of functions and what triggers them prevents the "nobody knows why this happens" state.
Warning signs in an inherited setup
- Nobody has a list of what automation exists.
- No error handling anywhere.
- Functions with names like "test2" still running in production.
- Logic duplicated across several functions.
- Nobody is sure whether a function is still needed.
Inherited a Zoho setup nobody understands, or hitting limits on automation? Tell us what is failing. See our Zoho extension service and the customisation ladder.