Data integration, dashboards, predictive models, and analytics risk.
This chapter connects data preparation to analytics and decision support. The important ISC skill is to evaluate whether analytical outputs are complete, reliable, and appropriately controlled before they are used for assurance or management purposes.
Analytics questions should be answered from source to conclusion. A dashboard, model, or monitoring routine is only reliable if the data was complete, transformed correctly, governed appropriately, and interpreted with its limits in mind.
| Analytics area | What to verify first | Common ISC trap |
|---|---|---|
| Data integration | Whether source data is complete, consistent, mapped, and reconciled. | Trusting combined data without testing transformations and duplicates. |
| Dashboards and visualization | Whether the display accurately represents the underlying data and assumptions. | Treating a polished dashboard as reliable evidence. |
| Predictive analytics and AI | Whether model inputs, logic, training, and limitations are understood. | Accepting model output without evaluating bias, drift, or explainability. |
| Continuous monitoring | Whether analytics are governed, reviewed, and tied to control responses. | Assuming automated alerts are effective without follow-up procedures. |
| Step | ISC question to ask | Control implication |
|---|---|---|
| 1. Trace source data | Which systems, files, feeds, or manual inputs supply the analysis? | Source reliability sets the ceiling for analytics reliability. |
| 2. Test integration logic | How are fields mapped, transformed, deduplicated, reconciled, and refreshed? | Integration errors can make accurate source data unreliable. |
| 3. Evaluate dashboard or model design | Do visualizations, thresholds, model inputs, and assumptions match the decision need? | Presentation choices can distort interpretation even when data is correct. |
| 4. Assess monitoring and governance | Who reviews exceptions, model changes, access, and performance drift? | Analytics used in controls need ownership and follow-up. |
| 5. Decide reliance level | Can the output support assurance, management monitoring, or only preliminary analysis? | The degree of reliance should match control quality and evidence. |
| Checkpoint | Ask before relying on output | ISC evidence effect |
|---|---|---|
| Source completeness | Did the analysis include the full population, period, systems, and manual inputs needed? | Missing or stale data can make a useful dashboard unreliable as evidence. |
| Transformation accuracy | Were mappings, joins, filters, deduplication, and calculations tested or reconciled? | Integration errors can create misleading trends even when source systems are accurate. |
| Model limitation | Are assumptions, bias, drift, explainability, and training limits understood? | Predictive output should not be treated as objective fact without model governance. |
| Dashboard design | Do labels, scales, thresholds, and visual emphasis match the underlying data? | Visualization choices can distort user judgment. |
| Follow-up control | Are exceptions reviewed, assigned, resolved, and documented? | Continuous monitoring is weak if alerts do not lead to accountable response. |