An Acumatica customization is a packaged application layer. The official documentation calls it a "customization project"; the community calls it a "CUSP" or simply "the package". Whatever the name, it is the unit of deployment for everything you build on top of Acumatica ERP. This guide is everything I have learned shipping customization projects across finance, distribution, and manufacturing clients in Kenya, Rwanda, Tanzania, Zambia, and South Africa.
1. What is in a customization
A customization is a ZIP that Acumatica publishes into a tenant. It can include any of:
- Source code (C# DLLs) — graph extensions, DAC extensions, business logic, event handlers, custom reports, services.
- Database changes — new DACs, new fields on existing DACs, new tables, new views.
- Screens — ASPX files for the classic UI, TypeScript/HTML for the Modern UI.
- Generic Inquiries — exported as XML.
- Reports — RDLC, schema, and parameter definitions.
- Dashboards, business events, import/export scenarios, webhooks, push notifications — also exportable as XML.
- Site map and access rights overrides — for the classic UI.
The crucial fact: a customization is a single atomic unit. You publish it; the entire bundle is applied. There is no partial publish (in production, anyway).
2. The Customization Project Editor
The Customization Project Editor is the Visual Studio-like IDE inside Acumatica. It is fine for the first 20 files. Past that, you should be in Visual Studio or Rider with the Acumatica add-in, and using source control. The CPE is for packaging, not authoring.
3. Project structure that scales
Beyond a few screens, a flat customization becomes unmaintainable. The structure I use for any project that will live past one release:
AcumaticaMyClient/
├── AcumaticaMyClient.sln
├── src/
│ ├── AcumaticaMyClient.Core/ // DACs, business logic, services
│ ├── AcumaticaMyClient.Graph/ // Graph extensions
│ ├── AcumaticaMyClient.Web/ // ASPX, Modern UI extensions
│ └── AcumaticaMyClient.Reports/ // RDLC and schemas
├── test/
│ └── AcumaticaMyClient.Tests/ // xUnit + Acumatica Unit Test Framework
└── db/
└── migrations/ // SQL change scripts, versioned
4. DACs — the heart of Acumatica
A DAC (Data Access Class) is a C# class that maps to a database table. The mapping is attribute-based:
[Serializable]
[PXCacheName("My Entity")]
public class MyEntity : IBqlTable
{
#region RecordID
[PXDBIdentity(IsKey = true)]
public virtual int? RecordID { get; set; }
public abstract class recordID : PX.Data.BQL.BqlInt.Field<recordID> { }
#endregion
#region Name
[PXDBString(60, IsUnicode = true)]
[PXUIField(DisplayName = "Name")]
public virtual string Name { get; set; }
public abstract class name : PX.Data.BQL.BqlString.Field<name> { }
#endregion
}
Notice the dual field pattern: a property and a matching abstract class that BQL uses for type-safe queries. This is the Acumatica way, and skipping it will bite you within a week.
5. Graph extensions vs new graphs
You almost always want to extend an existing graph rather than create a new one. A graph extension inherits the entire screen and lets you add fields, tabs, actions, and event handlers without re-implementing anything. The base graph does all the heavy lifting (CRUD, validation, persistence); your extension customises behaviour.
public class ARInvoiceExt : PXGraphExtension<ARInvoiceEntry>
{
[PXOverride]
public IEnumerable<ARInvoice> GetInvoices()
{
// additional logic before calling the base
return Base.GetInvoices();
}
}
6. Field-level customisation on existing DACs
To add a field to a built-in DAC (e.g. ARInvoice), declare a DAC extension. Acumatica uses the Usr prefix convention:
[PXTable(IsOptional = true)]
public class ARInvoiceExt : PXCacheExtension<ARInvoice>
{
#region UsrExternalRef
[PXDBString(40)]
[PXUIField(DisplayName = "External Ref")]
public virtual string UsrExternalRef { get; set; }
public abstract class usrExternalRef :
PX.Data.BQL.BqlString.Field<usrExternalRef> { }
#endregion
}
7. Events — the right tool for the job
Acumatica graphs raise dozens of row events. The ones you will use 90% of the time:
RowSelected— UI-level: enable/disable fields, set defaults, hide controls.RowInserting/RowUpdating— validation before write.RowInserted/RowUpdated— after-the-fact side effects.FieldUpdated— when a specific field changes.FieldDefaulting— when a field needs a calculated default.RowPersisting— last chance to set fields before commit.
RowSelected fires for every row on every UI render. Anything heavy there (loops, remote calls) will make the screen feel sluggish. Push heavy work to RowPersisting or to a graph action.
8. The Modern UI layer
Since 2025 R2, the Modern UI is the default. Customizing it is a different craft from the classic UI:
- Screens are TypeScript/HTML files, not ASPX.
- You write to the Screen Configuration model for layout changes.
- For deep customization, the mobile framework (SM204510) is your friend.
- User-Defined Fields (UDFs) are first-class — they no longer live in a separate tab.
9. Source control and packaging
Customizations should live in Git. The binary ZIP should be buildable from source. The CPE is for one-off inspection; CI/CD should produce the package.
10. Testing
Acumatica ships a Unit Test Framework. Use it. A customization with even minimal coverage is dramatically easier to upgrade. At minimum:
- Tests for every graph action that has business logic.
- Tests for every DAC field that has a default or validation rule.
- Tests for every import scenario.
11. Upgrade survival
Three habits keep your customization upgrade-safe:
- Wrap everything in extensions. Never re-implement a base graph.
- Wrap DAC additions in
PXCacheExtension. Never edit base DACs. - Use the
Usrprefix on all new fields. Acumatica will not collide.
12. ISV solutions — when to go commercial
If your customization is genuinely a product (multi-tenant, distributed, versioned, supportable), you publish it as an ISV solution on the Acumatica Marketplace. Different packaging, different pricing, different support obligations. Worth the path when you have customers beyond one.
13. Common production issues
- Forgetting to publish the database script. The code deploys, the field does not exist, the tenant 500s.
- Modifying a base DAC. Your change gets blown away on upgrade.
- Heavy logic in RowSelected. The screen becomes a slideshow at 10k records.
- Hard-coded tenant IDs. The next tenant cannot use your customization.
- No source control on the customization ZIP. A laptop dies, a year of work dies with it.
Related reading
Going deeper: production-grade patterns
The patterns above cover the basics. In production, the same patterns have to survive three things: scale, edge cases, and the next Acumatica upgrade. Here are the patterns that distinguish a working customisation from a great one — the ones I have applied to every client project in East and Southern Africa, and the ones that make the difference between a customisation the user trusts and a customisation they curse.
Defensive coding for the unexpected
Production is where the assumption dies. Every customisation that "works in test" fails in production the first time a customer name has a special character, an invoice is in a foreign currency, or a record has a null in a field you thought was required. The defensive habit is to explicitly handle the null, the empty, the special character, and the foreign currency in every event handler and every code path. The cost is 20% more code. The payoff is 95% fewer production tickets.
Three patterns I apply everywhere:
- Null-safe property access. Use
?.on every property access; the alternative is a NullReferenceException at 2 AM. - Explicit value handling. If a field can be empty, treat it as empty. Do not assume the default value.
- Defensive database reads. A
PXSelectthat returns null is a valid result, not an error. Handle it.
public class DefensiveExt : PXGraphExtension<BaseGraph>
{
protected void _(Events.RowSelected<MyDAC> e)
{
var row = e.Row;
if (row == null) return; // null-safe
var ext = row.GetExtension<MyDACExt>();
if (ext == null) return; // null-safe extension
var value = ext.UsrField ?? "DEFAULT"; // null-coalesce
var ok = decimal.TryParse(value, out var n); // try-parse
if (!ok) { /* handle */ }
}
}
Performance: the patterns that scale
Five performance patterns I apply on every customisation, in order of impact:
- Move heavy logic out of
RowSelected. Push validation toRowPersisting, side effects to a graph action triggered by a button.RowSelectedfires for every row on every render. - Index the join columns. Every BQL
Where<>filter needs an index. Check the execution plan before you ship. - Filter at the GI, not the UI. A GI that returns 5 million rows and filters in the presentation layer will time out. Push filters into the Conditions tab.
- Batch the work. Loop with 1,000 calls is slow; loop with 10 calls of 100 records is fast. Batch where you can.
- Cache the static. Tax schedules, account lists, and other static reference data can be cached for the lifetime of the app pool. Reduce the database load.
For the full performance playbook, see the performance tuning guide and the SQL Server indexing guide.
Upgrade survival
The customisation that breaks on the next Acumatica upgrade is the one that took a shortcut. The patterns that survive:
- Extend, never modify.
PXCacheExtension<T>over editing the base DAC.PXGraphExtension<T>over editing the base graph. - Usr prefix on every field. Acumatica uses this to separate your fields from base fields. Without it, your field collides with a base field on the next upgrade.
- Source control with a clear branch strategy. Main is production; develop is next release; feature branches are work in flight. Tag every release.
- Test on production data. The staging tenant is a copy of production. The data shape is the same. The bugs are the same.
// Base field — Acumatica owns this
[PXDBString(40)]
public string RefNbr { get; set; }
// Your field — always Usr prefix, never collides
[PXDBString(40)]
[PXUIField(DisplayName = "External Ref")]
public string UsrExternalRef { get; set; }
// Your DAC extension — soft extension, survives table drops
[PXTable(IsOptional = true)]
public class MyDACExt : PXCacheExtension<MyDAC>
{
#region UsrCustomField
[PXDBString(60)]
public string UsrCustomField { get; set; }
public abstract class usrCustomField :
PX.Data.BQL.BqlString.Field<usrCustomField> { }
#endregion
}
Testing: the habit that pays for itself
If you are not testing your customisation with the Acumatica Unit Test Framework, you are running blind. The framework ships with every installation, costs nothing, and pays for itself the first time an upgrade changes a method signature on you. The minimum coverage:
- Every graph action with business logic — happy path + the most common error path.
- Every DAC field with a defaulting or validation rule.
- Every workflow transition — that the right state is reached from the right source state.
- Every import scenario with a sample CSV that exercises the validation rules.
For the full test framework walkthrough, see the unit test framework guide.
Operations: what to do after the customisation is live
A customisation is not "done" when it ships. It is "done" when it has run in production for a quarter without a critical incident. The operational habits that get you there:
- Monitor the slow queries. Acumatica's System Monitor has a slow-query log. Review weekly; the slow query you ship is the production incident in two months.
- Track the licence headroom. Every active session counts. A leaking integration can lock out real users within an hour.
- Review the audit log. Not for compliance — for understanding how the system is used. The audit log tells you where to optimise next.
- Document the runbook. When the customisation fails (and it will), the runbook is what saves the on-call. Document the failure modes, the diagnostic flow, the fix.
For the broader operational patterns, see the monitoring guide and the licence concurrency guide.
The migration off the old customisation
Every customisation is eventually replaced. Plan for that day from the start. The patterns:
- Wrap external dependencies. If your customisation talks to an external API, wrap the call in your own service. When the API changes, you change one place.
- Tag the version. Every customisation has a version. The version is in the database, in the metadata, in the package. When you upgrade, you know what you are upgrading from.
- Document the data model. Every custom field, every new table, every relationship. The next person who has to read the customisation should be able to start with the data model.
- Test the migration path. A customisation that ships and cannot be removed is a liability. The migration path off should be tested before the customisation is in production.
For the broader migration patterns, see the data migration guide.
Wrapping up
That is the working approach I use on Acumatica projects. The same patterns show up whether you are in Nairobi, Johannesburg, Kigali, Lusaka or Harare — and they are the things that keep work moving when an upgrade lands at 6 PM on a Friday. If you are stuck on something specific, reach out or keep reading through the rest of the Acumatica blog.