Platform: Workday Adaptive Planning
Scope: a net zero sweep on one revenue account, by project
The business problem
Finance knew that certain projects on one revenue account were swept to net zero. They did not know how the model chose those projects, or how to add and remove projects each year. Each change needed outside help.
Existing architecture
We traced the sweep from end to end and found five parts.
A mapping sheet lists the eligible combinations of project, account and fund.
A data source joins income statement actuals to that sheet, and a SQL column skips any row whose project is not on the list.
Loaders write trigger values for eligible projects; clearing stale triggers depends on the replace setting of each loader.
A task runs those loaders together.
A formula runs the sweep only where the trigger has a value.
Why it happened
Nothing was broken. The design was not documented, so the single control point, the mapping sheet, was not visible to the people who needed it.
We also found a second, separate sweep on the same account. It is controlled by an attribute on the project dimension and needs no loader. That explained why some projects moved without appearing on the mapping sheet.
Design decision
We changed no formula, integration or dimension. We documented the control points and wrote the add and remove steps for finance.
Implementation approach
- To add a project, update the approved project/account/fund mapping, run the trigger task and verify the expected trigger intersection.
- To remove a project, remove its mapping and rerun the applicable task. Check whether the specific loader mode clears old trigger data; do not assume that behaviour is the same in every version.
- Validate the sweep formula and relevant project/dimension attributes independently, including any second sweep mechanism.
Controls and validation
We traced the mapping, data source, loader and formula, and exercised add/remove scenarios in the sandbox. The actuals trigger behaviour was reviewed. The budget loader had a different replace setting, so budget trigger clearing could not be assumed from the actuals test and required its own configuration check.
Outcome
We delivered an understandable maintenance procedure and identified the separate dimension-attribute mechanism. We documented the budget-trigger clearing gap as a follow-up item.
Lessons learned
- Before changing a model that seems to behave oddly, trace it end to end. The behaviour is often intended and undocumented.
- Two mechanisms can act on the same account. Check dimension attributes as well as sheets.
- Check that actuals and budget loaders use the same replace setting.
Scope of the outcome
Sweep logic documented and selected sandbox checks performed; budget-trigger clearing needs separate configuration.
EPMLogic | Implementation insight based on an anonymized scenario. Exact operational amounts and client details are intentionally omitted.