Five Numbers Riverbend Was Reading Wrong: A Plain-Language Memo on a Term of Delivery, Returns, Checkout and Staffing Data
Student Name
American College of Education
STAT5003: Business Statistics: Data-Informed Decision-Making
Module 6 Assignment
Instructor Name
June 23, 2025
Purpose and Bottom Line
This memo is written for Riverbend's leadership team and summarizes five analyses completed this spring on delivery times, store return rates, the cost of delivery damage, the redesigned online checkout and sales floor staffing. It uses no statistical terms that are not explained where they appear. The bottom line is that in four of the five cases, the number the business had been using was not wrong, but it answered a different question from the one managers were asking, and in each case a better number leads to a different decision.
Gigerenzer et al. (2007) found that professionals as well as the public misread statistics far less often when figures are expressed as counts of real cases, such as 81 customers out of 1,240, than when they are given as percentages or probabilities alone. This memo follows that advice wherever it can. Every number below comes with the question it answers, because a correct number attached to the wrong question is the most expensive kind of error.
Finding One: Delivery Times
What we found: the monthly report says deliveries take 3.4 days on average, and a typical delivery does take about three days. But in the first quarter, 81 of 1,240 customers waited more than a week, and 74 of those 81 were rural customers of the South distribution center whose orders went to an outside carrier, which averaged about 10 days. How sure we are: very sure, because this is every delivery in the quarter, not a sample. What to do: report the typical delivery time, the share of customers waiting more than a week and each route separately, and review the rural carrier arrangement, since the website promises every one of those customers three to five days.
Finding Two: Ranking Stores by Return Rate
What we found: the monthly ranking of stores by return rate mostly rewards and punishes store size. Small stores swing widely from month to month by chance; at the smallest store, a month's rate can move anywhere between about 2 and 10 percent without anything changing. Last month's best and worst stores were the two smallest, and both results were ordinary for their size. How sure we are: very sure about the pattern, which follows from how rates behave when they are based on few orders. What to do: stop ranking stores on a single month, compare each store with the range expected for its size and judge small stores over six months rather than one.
Finding Three: The Cost of Delivery Damage
What we found: an audit of 400 randomly chosen returns found that about three in ten are caused by damage in delivery, and each costs about $186 in a second trip, crew time and markdowns. How sure we are: the true share is very likely between 27 and 35 in every 100 returns, and the average cost between about $160 and $212. Put together, damage costs Riverbend between about $147,000 and $253,000 a year, most likely about $196,000. What to do: the proposed $60,000 packaging and crew-training program pays for itself if it cuts damage by about two-fifths, even at the low end of the estimate. Start it at the South center and repeat the audit in six months to see whether the share has fallen.
Finding Four: The Redesigned Checkout
What we found: online orders through the redesigned checkout averaged $9.43 more than orders through the old one, based on 812 orders from March to May. How sure we are: the increase is probably real, but its size is uncertain; it could plausibly be as small as about 75 cents or as large as about $18 per order. At the low end, the gain would not repay the $46,000 cost of building the redesign into the mobile app; at the middle, it would repay it several times over. What to do: release the redesign to half of mobile customers for 60 days, keep the other half as a comparison and commit to the full build only when the evidence shows the gain clearly exceeds its cost. A result can pass a test of statistical significance and still leave the business decision open, which is the point Wasserstein and Lazar (2016) made on behalf of the American Statistical Association when they warned that such a result does not measure the size or importance of an effect.
Finding Five: Adding Sales Floor Hours
What we found: the chart showing that each staffed hour brings about $445 in weekly sales mostly reflects the fact that stores add staff in weeks when more customers come in. Comparing weeks with the same number of customers, each added hour went with about $196 in sales. How sure we are: not very. The true figure could be anywhere from about $104 to $287 an hour, and because busy weeks also have promotions and events the data cannot separate out, even $196 may be too high. What to do: the proposal to add 40 hours a week at every store may pay, but test it first. Add the hours in alternating weeks at four randomly chosen stores for eight weeks, at a labor cost of about $13,800, and compare.
What to Do First
The five actions differ in cost and urgency, so they are ranked here. First, change the monthly delivery and returns reports; it costs nothing, and it stops managers from acting on misleading numbers next month. Second, start the damage-prevention program at the South center, since it is the largest dollar amount in this memo and pays for itself across the whole estimated range. Third, review the rural carrier arrangement, because the customers it serves are waiting twice as long as the website promises and some of them are likely among the damage and return cases as well. Fourth and fifth, run the staged tests for the checkout redesign and the added staffing hours, each of which costs little and protects a larger decision from an estimate that is still too uncertain to commit on.
Four Habits for the Next Number
The five findings share four lessons that apply beyond this term. First, ask what question a number answers before asking whether it is good; an average can be accurate and still hide the customers who complain. Second, ask how many cases a number rests on, since small groups produce extreme results by chance. Third, ask for a range, not a single figure, and make the decision work across the whole range. Gigerenzer (2004) described the habit of reporting whether a result is significant and stopping there as a ritual that replaces thinking; a range forces the thinking back in. Fourth, when a number seems to show that one thing causes another, ask what else moved at the same time, and where the stakes are high, test the change on part of the business before making it everywhere.
None of these habits requires statistical training. Each requires only that someone in the room ask one more question before the number is used.
References
Gigerenzer, G. (2004). Mindless statistics. The Journal of Socio-Economics, 33(5), 587-606. https://doi.org/10.1016/j.socec.2004.09.033
Gigerenzer, G., Gaissmaier, W., Kurz-Milcke, E., Schwartz, L. M., & Woloshin, S. (2007). Helping doctors and patients make sense of health statistics. Psychological Science in the Public Interest, 8(2), 53-96. https://doi.org/10.1111/j.1539-6053.2008.00033.x
Wasserstein, R. L., & Lazar, N. A. (2016). The ASA statement on p-values: Context, process, and purpose. The American Statistician, 70(2), 129-133. https://doi.org/10.1080/00031305.2016.1154108
How this STAT 5003 Module 6 example is structured
STAT 5003 Module 6 usually turns the term's analysis into a short memo written for non-specialists; your classroom's instructions decide the length and whether technical appendices are allowed. This example opens with the bottom line, gives each finding the same three parts, what we found, how sure we are and what to do, and keeps every technical term out of the body or explains it in the sentence where it appears. The closing section turns the term's lessons into habits rather than repeating the findings.
STAT5003 Module 6 questions, answered
What does STAT5003 Module 6 usually ask for?
STAT5003 Module 6 usually asks for a memo or report that turns the course's statistical analysis into findings and recommendations for managers without statistical training. Your classroom's instructions decide the length and whether technical detail may appear in an appendix.
How do I explain uncertainty to non-specialists?
Give a range in everyday units, say what the decision looks like at each end of it and use counts of real cases where you can. Avoid terms such as p-value in the body unless you explain them in the same sentence.
How long should a statistics memo for managers be?
Long enough to give each finding its size, its uncertainty and its action, and no longer. A consistent structure for every finding helps busy readers find what concerns them.
Write yours, or have the desk draft it
This paper is an original model document written by our desk, not a submitted student paper and not an official American College of Education document. Read it for the moves, then write your own to the instructions in your classroom. If you want one built to your exact prompt and rubric, the first custom sample is free and arrives in 24 to 48 hours.