- 01Data Profiling
- 02Strategy
- 03Execution
- 04Validation
Strategy & Planning
In this module, aha presents its proposed harmonization plan for every variable, and you can review, confirm, or edit the proposals before execution.
Overview
For each source variable, aha proposes a decision, a target standard variable, and a sequence of transformations. Each proposal carries a plan confidence score, and you set a threshold that determines which variables require human review before the plan can be approved.
Decision types
Every variable in the plan is assigned one of five decision types, reflecting how the source variable relates to the standard:
| Decision | Definition |
|---|---|
| Matched | The source variable is identical to the standard variable. No transformation is needed; it is used as-is. |
| Mapped | The source directly corresponds to a standard variable with consistent meaning. Full consistency is achieved after applying transformations. |
| Partially Mapped | Source and standard capture the same concept but differ in operationalization. A partial relationship exists, but full alignment is not achievable. |
| Calculated | The standard variable does not exist in the source but can be derived from one or more source variables using a formula or logic. |
| Not Mapped | The source variable cannot be aligned with any standard variable — directly, partially, or through calculation. |
Plan confidence
Each row in the plan table includes a Plan Confidence score — aha’s overall certainty that its proposed plan for a variable is correct.
Human intervention threshold
At the top of the module, a colour-coded slider sets the confidence threshold below which variables are flagged for human review. The recommended default is 60%.
You can return the slider to the recommended value at any time using the Reset to recommended button below it.
The plan table
The main table lists every source variable and its proposed plan:
| Column | Description |
|---|---|
| Source | The variable name in the unharmonized dataset. |
| Target | The corresponding variable in the standard dataset. Empty for Not Mapped variables. |
| Decision | The proposed decision type. |
| Plan Confidence | The overall confidence score for the proposed plan. |
| Human Intervention | Whether the variable requires review before proceeding, based on the threshold. |
| Review | A button that opens the individual variable review panel. |
Use the sort button to order variables by confidence, and the filter button to show only a specific decision type.
Reviewing variables
Review variables individually with the Review button on any row, or in bulk from the top of the module:
- Review all — opens the panel at the first variable and steps through all of them in sequence.
- Review only those that require human intervention — opens the panel filtered to the flagged variables.
The review panel
The panel shows the source-to-target mapping at the top (e.g. RID → NACCID) and a progress indicator. It is divided into four sections:
- Source & Target Info — each variable’s data type, unique values, missing rate, and a plain-language explanation of how the two relate.
- Decision — the five decision types, aha’s pick is highlighted. Select a different one if appropriate.
- Target — ranked candidate standard variables, aha’s pick at the top. Click the info icon on any candidate for its full metadata.
- Transform — ranked transformation sequences, each with a description and a before/after preview. Confirm one with the Selected button.
Where to go next
Once you approve the plan, aha moves to Execution, which applies the transformations to the data. See Transformation types for what each transform does.
