ACC 430 Module 4 Project One Example

Stops, Miles or Pounds? A Regression of Daily Delivery Cost at a Composite Janitorial Supply Distributor

[Student Name]

Southern New Hampshire University

ACC 430: Data Analytics for Financial Professionals

Project One

[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.

Stops, Miles or Pounds? A Regression of Daily Delivery Cost at a Composite Janitorial Supply Distributor

The Question

The CFO's margin question led to a narrower one: how much does it cost to deliver to a customer? Delivery is the distributor's largest cost after product, about $7.4 million a year across 34 routes. If cost depends mainly on stops, small accounts that need weekly visits may cost more to serve than they earn; if it depends mainly on miles or weight, the answer changes. This project uses regression to estimate how daily route cost responds to stops, miles and weight, so that cost can be assigned to customers in the profitability project (Datar & Rajan, 2021).

Data and Variables

The data cover 520 route-days sampled across twelve months and all 34 routes. Daily route cost, the dependent variable, includes driver wages and overtime, fuel, and a per-day allocation of truck lease and maintenance. Explanatory variables are the number of stops, miles driven from the routing system and weight delivered in thousands of pounds. Stops range from 8 to 38, miles from 40 to 260 and weight from 2,000 to 21,000 pounds, which defines the range in which the model can be used.

Results

Table 1. Regression of Daily Route Cost

All three variables are statistically significant, with t statistics well above 2. Together they explain about 81 percent of the variation in daily cost across the sample.

Interpretation

Holding miles and weight constant, each additional stop adds about $14.20 to the day's cost, reflecting the driver's time to park, unload, get a signature and handle paperwork. Each additional mile adds about $1.65, mostly fuel and wear. Each additional thousand pounds adds about $18, reflecting unloading time and fuel for heavier loads. The intercept of $210 represents costs that do not vary with these activities within the range of the data, such as part of the daily truck allocation, but it should not be read as the cost of a day with no deliveries, because no such days are in the data.

A typical route-day with 22 stops, 140 miles and 9,000 pounds is predicted to cost $210 plus $312 for stops, $231 for miles and $162 for weight, about $915. Of that, about a third is driven by stops, which is the figure that matters most for small accounts.

Diagnostics

Residuals plotted against predicted cost show no curve or funnel shape, suggesting the linear form and constant variance are reasonable. Two large positive residuals trace to a truck breakdown that required a second truck and to the day before Thanksgiving, when overtime was paid; both are genuine but unusual, and removing them changes coefficients by less than 3 percent, so they were kept. Stops and miles are moderately correlated at 0.48, because routes with more stops tend to cover more ground, but the variance inflation factors are about 1.3, well below levels that would make the separate coefficients unstable. Richardson et al. (2021) stress that these checks are what separate a model a manager can rely on from one that merely produced output.

How the Data Were Sampled

The 520 route-days were drawn as a stratified random sample: about 15 days from each of the 34 routes, spread across all twelve months. Stratifying by route ensured that rural routes with long miles and few stops were represented alongside dense urban routes, which matters because the model's purpose is to compare different kinds of customers. Days with partial data, such as those when a truck's telematics unit failed, were excluded before sampling, and the number excluded, 23, is recorded. A random sample of this size gives coefficients that are stable enough for planning; rerunning the regression on a second, independent sample of 520 days produced coefficients within 8 percent of the first.

Limits

The model describes cost behavior within the observed range: 8 to 38 stops, 40 to 260 miles. It should not be used to estimate the cost of a 60-stop route or a new territory twice as far away. It estimates associations, not causes; adding a stop might cost less on a dense route than on a rural one, which the model averages. And costs such as dispatcher salaries are excluded, so the per-stop figure understates the full cost of serving a customer.

Alternative Specifications

Two alternatives were tested before settling on the three-variable model. A model with stops alone explained 62 percent of the variation, showing that stops matter most but that miles and weight add real information. A model adding a dummy variable for summer months, when trucks idle with air conditioning running, raised R-squared only to 0.82, and its coefficient of about $24 per day was small relative to daily cost, so it was left out for simplicity. A logarithmic model, in which costs rise less than proportionally with stops, fit no better than the linear one within the observed range. The chosen model is therefore the simplest that explains the data well, which also makes it easier for operations managers to understand and challenge.

Use in Decisions

The estimates give a defensible way to assign delivery cost to customers: each customer's annual stops, its share of route miles and the weight it receives. A small restaurant receiving a weekly delivery of 150 pounds, on a route where it adds about 4 miles, costs roughly $14.20 plus $6.60 plus $2.70, about $23.50 per delivery, or $1,220 a year, before warehouse and order-handling costs. Davenport and Harris (2017) argue that analytics creates value when it changes specific operating decisions; this figure goes directly into the customer profitability project and the decision about minimum order sizes.

Conclusion

Daily delivery cost at the distributor is driven mainly by stops, about $14.20 each, with miles at $1.65 and weight at $18 per thousand pounds; the model explains 81 percent of the variation and passes basic diagnostic checks. Within its range, it provides a sound basis for estimating what each customer costs to deliver to.

References

Datar, S. M., & Rajan, M. V. (2021). Horngren's cost accounting: A managerial emphasis (17th ed.). Pearson.

Davenport, T. H., & Harris, J. G. (2017). Competing on analytics: The new science of winning (Updated ed.). Harvard Business Review Press.

Richardson, V. J., Teeter, R. A., & Terrell, K. L. (2021). Data analytics for accounting (2nd ed.). McGraw Hill.