DSC 330 Week 4 Validating Data for Visualization Example

Is the Dashboard Telling the Truth? Validating the Data Behind a Hotel Company's Visuals

[Student Name]

University of Phoenix

DSC/330: Data Communication and Visualization for Business

Week 4 Assignment

[Instructor Name]

[Date]

Music Row Hospitality, its hotels, data and figures are composites written for a model paper.

In its first week of use, the daily dashboard built in Week 3 for the fictional hotel operator drew a phone call. A general manager reported that it showed her hotel $41,000 below budget for the prior month, while the accounting system showed the hotel on budget. If that kind of discrepancy went unexplained, managers would stop trusting the dashboard. This paper describes how the data were validated.

What Data Quality Means

Pipino et al. (2002) argued that data quality has both subjective dimensions, how users perceive the data, and objective dimensions, measurable properties such as free-of-error rates, completeness and timeliness, and proposed simple ratio measures for objective assessment, such as the share of records without errors. Batini et al. (2009) compared methodologies for assessing and improving data quality and found that most share common steps: understanding data and processes, measuring quality on chosen dimensions and choosing improvement actions that address causes rather than only correcting data. Validation for the dashboards followed those steps.

Designing the Checks

The first full validation run took about three days, most of it spent tracing differences to their source. Four kinds of checks were designed.

Reconciliation. Monthly room revenue by hotel in the dashboard compared with the accounting system, and rooms sold compared with each property management system's own monthly report.

Completeness. Every hotel must have a nightly load for every day; every survey response must have a hotel and date.

Validity rules. Occupancy between 0 and 100 percent; average daily rate within a plausible range for each hotel; no negative revenue except recorded refunds.

Timeliness. Each nightly load must arrive by 5 a.m.; survey files weekly by Tuesday.

What the Checks Found

Double-counted group bookings. At the general manager's hotel, the property management export listed group blocks both as a group record and as individual room nights when guests checked in, inflating rooms sold but assigning revenue to only one. The dashboard's revenue per available room was therefore low, and its occupancy high. Correcting the extract to exclude block records removed the $41,000 discrepancy.

Missing day. A system upgrade at one hotel caused its export to fail for one night. The dashboard showed zero revenue for that day, which pulled down weekly figures and triggered a false below-target flag.

Misassigned survey responses. Guests of two hotels with similar names in the survey system were sometimes assigned to the wrong one; about 4 percent of responses for each were affected.

Time zone. The Kentucky hotel is in the Central time zone, but its export timestamps were treated as Eastern, shifting some late-night check-ins to the next day. Daily figures were affected; monthly totals were not.

Implausible rates. Eleven records across the portfolio showed average daily rates under $10, all employee rate bookings coded incorrectly.

The dashboard was not wrong about the hotel; it was faithfully reporting a data export that counted some guests twice.

Why the Problems Mattered

Each problem would have led to a wrong decision if left in place. The double-counted group bookings made one hotel's revenue per room look weak, which could have prompted a rate cut that was not needed. The missing day flagged a hotel as below target, which could have triggered a needless call from the chief operating officer. The misassigned surveys moved satisfaction scores for two hotels in opposite directions. Redman (1998) argued that poor data quality raises costs and weakens decisions across an enterprise, often in ways no one traces back to the data. Validation is how those hidden costs are avoided.

The General Manager's Response

When the analyst called the general manager back with the explanation and the corrected figure, she asked whether other numbers might be wrong. The analyst walked her through the new checks and the incomplete-data marker. A week later she began using the dashboard in her morning meeting. Her first question, and the analyst's answer, became the model for how the company responds to data questions: take every challenge seriously, trace it to a cause and explain the fix.

Measuring Quality

Using Pipino et al.'s ratio approach, before fixes, 96.1 percent of hotel-days passed all checks; after fixes, 99.8 percent did. Monthly revenue reconciled with accounting within 0.2 percent for all nine hotels.

Fixing Causes

Each problem was traced to its cause. The group booking issue was fixed in the extract rule. The missing day was reloaded, and the load process now retries automatically. Survey hotel assignments now use a hotel code rather than a name. The Kentucky time zone is now set in the data model. Employee rate bookings are excluded from rate calculations by code, and front desk staff received a reminder about coding.

Checks That Found Nothing

Some checks found no problems, which is also useful information. Rooms sold matched each property management system's own monthly report at seven of nine hotels once group blocks were fixed, and payroll hours matched timekeeping totals everywhere. Knowing which data are reliable lets the analyst focus attention on the sources that need watching.

Ongoing Monitoring

Checks now run automatically every night. If any fails, the analyst receives an alert before managers open the dashboard, and the affected figure is marked. A monthly reconciliation report compares dashboard revenue with accounting and is reviewed by the chief financial officer.

Telling Users About Uncertainty

The dashboard now marks any hotel-day with a missing or failed load with a gray symbol and the note "data incomplete, update expected by noon." Satisfaction scores with fewer than 30 responses are shown in lighter gray with the count. This policy tells users when not to rely on a number instead of hiding problems.

Conclusion

A single phone call exposed several data problems behind Music Row's dashboards. Systematic checks of reconciliation, completeness, validity and timeliness found them, tracing showed their causes and fixes addressed causes rather than symptoms. Nightly automated checks and clear marking of uncertain data mean users can trust what they see, which prepares the company to present its data story in Week 5.

References

Batini, C., Cappiello, C., Francalanci, C., & Maurino, A. (2009). Methodologies for data quality assessment and improvement. ACM Computing Surveys, 41(3), Article 16. https://doi.org/10.1145/1541880.1541883

Pipino, L. L., Lee, Y. W., & Wang, R. Y. (2002). Data quality assessment. Communications of the ACM, 45(4), 211-218. https://doi.org/10.1145/505248.506010

Redman, T. C. (1998). The impact of poor data quality on the typical enterprise. Communications of the ACM, 41(2), 79-82. https://doi.org/10.1145/269012.269025