ACC 315 Module 8 Data Analytics Short Paper Example

Finding the Missing Gallons: Data Analytics and Continuous Monitoring at a Composite Propane Dealer

[Student Name]

Southern New Hampshire University

ACC 315: Accounting Information Systems

Module Eight Short Paper

[Instructor Name]

[Date]

The organization, setting and figures below are a composite written as a model document. No real employer, client, colleague or patient is described.

Finding the Missing Gallons: Data Analytics and Continuous Monitoring at a Composite Propane Dealer

Introduction

For years the composite Vermont dealer at the center of this course could not explain about 2.6 percent of the propane it bought, roughly 70,000 gallons a year. Module Three recommended monthly reconciliations, and Project Two recommended an integrated system that records every truck meter reading. With that system live, the dealer now holds data it never had: every delivery's meter start and end, every truck's morning fill and every tank's usage history. This paper explains how analytics can turn that data into control, designs one continuous monitoring test, reports what a month of results showed and sets out the limits.

Four Kinds of Analytics Applied

Analytics is often described in four types (Romney et al., 2021). Descriptive analytics reports what happened, such as gallons sold by town last month. Diagnostic analytics asks why, for example why gallons per delivery fell in one town, which might reveal a new competitor. Predictive analytics estimates what will happen, such as when a tank will reach 30 percent given forecast temperatures. Prescriptive analytics recommends an action, such as which deliveries to move forward before a storm. Vasarhelyi et al. (2015) argue that accounting has been slow to use the volume and variety of data now available and that its value lies in combining transaction data with nonfinancial data, such as weather and meter readings, which is exactly the combination this dealer now holds.

A Daily Monitoring Test

The most valuable test addresses the missing gallons directly. Each evening, the system compares, for each truck, the gallons pumped according to the meter readings recorded at each stop with the gallons on that day's invoices, and compares both with the truck's morning fill and evening gauge reading. Any truck-day where pumped and billed gallons differ by more than 0.5 percent is flagged and listed on a report that the controller reviews the next morning.

The threshold matters. Too tight, and normal meter rounding and temperature effects produce dozens of flags a week, which teaches the reviewer to ignore them. Too loose, and a driver could leave small deliveries off tickets without triggering a flag. Half a percent, about two gallons on a typical 400-gallon day, was chosen after reviewing two weeks of results in which nearly all legitimate differences fell below it.

Appelbaum et al. (2017) describe this kind of analytic as moving assurance from sampling after the fact toward testing whole populations as they occur, and they note that the hard work lies in deciding which exceptions deserve follow-up. That judgment remains the accountant's.

One Month of Results

Table 1. Daily Truck Test, First Full Month

Seven of the nine flags were timing issues, deliveries recorded after the tablet regained signal, and the system now matches these automatically. One revealed a meter that had drifted. One, a Saturday pattern on one route, was referred to the owner and turned out to be a driver delivering to a relative's farm without recording it, the kind of loss that had been hidden for years in the annual 2.6 percent.

Predictive Analytics for Delivery

The same data improves operations, and it is where predictive analytics earns its keep in this business. Each tank's usage factor, the gallons it uses per heating degree day, can be re-estimated after every delivery. Combined with the weather forecast, the system predicts when each tank will reach the reorder point and builds the next day's route. Better forecasts mean fewer emergency runouts and fewer partial deliveries, and they also keep budget plan payments accurate, the $41,000 problem identified in Project One. A runout is costly in several ways: a special trip, a leak check required by safety rules before the system is relit and an unhappy customer who may switch dealers. If better forecasts prevent even forty runouts a winter, the savings in overtime and leak checks alone would be several thousand dollars.

Prescriptive analytics takes one more step. Given the forecast, the system can propose the cheapest route that reaches every tank before its reorder point, and it can recommend moving deliveries forward when a storm will close back roads. The dispatcher still decides, but she starts the day with a plan built from data rather than memory.

Limits

Analytics depends on data quality. A driver who types the wrong tank number makes the forecast wrong for two tanks. False positives waste time; in the first month seven of nine flags were harmless. And a dashboard no one reviews controls nothing. In a five-person office, the daily report must be short enough to read in five minutes and assigned to one person, with the owner reviewing a monthly summary. Finally, analytics finds patterns, not intent: the Saturday deliveries were explained by a conversation, not by the data.

There are also costs that a vendor demonstration does not show. Someone must maintain the threshold as trucks and meters change, investigate each flag and record the outcome, and decide when a pattern justifies a conversation with an employee. Handled carelessly, a monitoring test can damage trust among drivers who have done nothing wrong. The dealer should tell staff that the test exists, explain that it protects honest drivers by clearing them quickly when gallons do not match, and apply it to every truck equally.

Conclusion

With an integrated system in place, the dealer can test every truck every day instead of comparing annual totals once a year. One month of a simple test found a faulty meter and an unrecorded delivery pattern that had cost money for years. The accountant's role shifts from finding the problem at year end to designing the test, choosing the threshold and judging which flags matter.

References

Appelbaum, D., Kogan, A., & Vasarhelyi, M. A. (2017). Big data and analytics in the modern audit engagement: Research needs. Auditing: A Journal of Practice & Theory, 36(4), 1-27. https://doi.org/10.2308/ajpt-51684

Romney, M. B., Steinbart, P. J., Summers, S. L., & Wood, D. A. (2021). Accounting information systems (15th ed.). Pearson.

Vasarhelyi, M. A., Kogan, A., & Tuttle, B. M. (2015). Big data in accounting: An overview. Accounting Horizons, 29(2), 381-396. https://doi.org/10.2308/acch-51071