| Course | BU 443 Marketing Analytics |
|---|---|
| Unit | Unit 6 |
| Paper type | Test design and reading |
| Length | About 940 words, 6 pages |
| Format | APA 7 student paper |
| School | Herzing University |
| Program | BS in Marketing |
| Updated | September 2026 |
Free sample paper for BU 443 Unit 6
Fifty Dollars or Sixty-Five for Free Shipping: A Test Design and Reading for a Composite Online Houseplant Retailer, Including What It Means When Nothing Clearly Moves
Student Name
Business, Herzing University
BU 443: Marketing Analytics
Instructor Name
July 8, 2026
Fifty Dollars or Sixty-Five for Free Shipping: A Test Design and Reading for a Composite Online Houseplant Retailer, Including What It Means When Nothing Clearly Moves
The Question and What Is Varied
Greenline Plant Co., the composite houseplant retailer in this course, offers free shipping on orders over $65. The owners believed lowering the threshold to $50 would lift sales enough to cover the extra shipping cost. Plants are heavy and fragile to ship, so free shipping is expensive, and the question could not be settled by opinion. This paper designs a test that could answer it, commits to its size and duration in advance, and reads the result, which turned out to be the case the owners least expected: no clear change.
One thing is varied: the free-shipping threshold, shown in the site banner, on product pages and in the cart. Visitors in version A see $65; visitors in version B see $50. Everything else is held still: prices, the product range, promotions, email and advertising, and the shipping charge below the threshold. Visitors are assigned at random, half to each version, by the store platform on their first visit, and stay in their version for the whole test, even if they return weeks later. Kohavi et al. (2009) set out the requirements for a trustworthy online test, including random assignment, a single change and an outcome chosen before the test begins, and the design follows them.
What Counts as a Win
The primary measure is gross profit per visitor: product revenue less product cost and shipping cost paid by Greenline, divided by visitors assigned to the version. Conversion rate alone would be the wrong measure, because a lower threshold could raise conversion while shrinking orders and adding shipping costs, making the business worse off while its headline number improved. Three guardrail measures are also tracked, since a version could win on the primary measure while damaging something else: conversion rate, average order value and shipping cost per order. The decision rule was written down on May 20: adopt $50 only if gross profit per visitor is higher in version B and the difference is unlikely to be chance.
How Many Visitors, and for How Long
Before the test, the team reasoned about how many visitors would be needed to see an effect. Greenline's visitor conversion rate was about 3.1 percent. The owners said a lift smaller than 10 percent, to about 3.4 percent, would not be worth the extra shipping cost, so the test was sized to detect a lift of that size. A common rule of thumb for 80 percent power at the usual 5 percent significance level is that each version needs about 16 times the variance of the measure divided by the square of the difference to be detected (Kohavi et al., 2009). For a conversion rate of 3.1 percent, the variance is about 0.030, and the difference is about 0.0031, so each version needs about 16 x 0.030 / 0.0031 squared, roughly 50,000 visitors.
Summer traffic runs about 25,000 visitors a week, so 100,000 visitors would take four weeks. The test was committed to run from June 1 to June 28, four full weeks covering every day of the week equally, with no early stopping. Kohavi et al. (2020) warn that checking results daily and stopping when a difference looks significant greatly increases the chance of a false win. The team also noted in advance that profit per visitor varies more than conversion, so the test would be less sensitive for the primary measure than for conversion.
The Result
The test ran as planned, with 50,120 visitors in version A and 49,880 in version B. The near-even split was checked first, since a lopsided split can signal a fault in assignment; the difference was well within what random assignment produces, so the comparison was read.
| Measure | A: $65 threshold | B: $50 threshold | Difference | Reading |
|---|---|---|---|---|
| Conversion rate | 3.12% | 3.29% | +0.17 points (+5.4%) | Not clearly different; 95% interval about -0.05 to +0.39 points |
| Average order value | $63.40 | $58.10 | -$5.30 | Lower in B |
| Shipping cost paid by Greenline per order | $6.10 | $8.90 | +$2.80 | Higher in B |
| Gross profit per visitor | $1.09 | $1.05 | -$0.04 (-3.7%) | Not clearly different; interval includes zero |
Version B converted slightly more visitors, but the difference could easily be chance: a two-proportion test gives a p-value of about 0.13, and the interval for the difference runs from slightly negative to about 12 percent above the control. Orders in B were smaller, as people bought just enough to reach $50 rather than $65, and Greenline paid more shipping per order. On the primary measure, profit per visitor was four cents lower in B, a difference also within the range chance could produce.
Reading a Result Where Nothing Clearly Moved
A null result is not a finding that the threshold does not matter. It means the test found no effect as large as the one it was designed to detect. The interval for conversion leaves room for a real lift of up to about 12 percent, and the interval for profit leaves room for a small gain or a small loss. What the test does show is that the large gain the owners hoped for did not happen, and that the mechanism which would erode profit, smaller orders and higher shipping costs, clearly did.
Under the decision rule written in May, the threshold stays at $65. The result also answers a question the owners had not asked: customers respond to a free-shipping threshold by adjusting their basket to it, which suggests a higher threshold, say $75, might raise average order value without much loss of conversion. That idea is a hypothesis for a future test, not a conclusion from this one. Farris et al. (2010) caution that a test's value lies as much in ruling out a costly change as in finding a profitable one, and this test prevented an expensive policy on the strength of a belief. Greenline, its visitors and every figure here belong to the composite case.
References
Farris, P. W., Bendle, N. T., Pfeifer, P. E., & Reibstein, D. J. (2010). Marketing metrics: The definitive guide to measuring marketing performance (2nd ed.). Pearson Education.
Kohavi, R., Longbotham, R., Sommerfield, D., & Henne, R. M. (2009). Controlled experiments on the web: Survey and practical guide. Data Mining and Knowledge Discovery, 18(1), 140-181. https://doi.org/10.1007/s10618-008-0114-1
Kohavi, R., Tang, D., & Xu, Y. (2020). Trustworthy online controlled experiments: A practical guide to A/B testing. Cambridge University Press.
Reading the BU 443 Unit 6 instructions
The sixth unit of BU 443 asks students to design a controlled test and interpret its result. Common requirements include stating the hypothesis, what is varied and what is held constant, how subjects are assigned, the primary and guardrail measures, the sample size and duration with reasoning, and a reading of the result including the possibility of no effect. Many instructors ask how to reason about sample size without advanced statistics; a documented rule of thumb is usually accepted. Tests need not involve web pages; an email subject line, a price or a shipping policy can be tested the same way. Composite results are acceptable if clearly labeled. A design section followed by a results table and interpretation in APA 7 is typical.
How this BU 443 Unit 6 example is built
The paper opens with the owners' belief and why it needed testing, then states what is varied, what is held still and how visitors are assigned. A section explains the choice of profit per visitor over conversion and lists three guardrails with the decision rule. The sample size section works through a rule of thumb with the numbers and commits to four full weeks, citing warnings against early stopping. The results section checks the split, then presents a four-measure table with differences and readings. The final section explains what the null result rules out and what it leaves open, applies the decision rule and notes a new hypothesis for a future test.
Where the points sit in the BU 443 Unit 6 rubric
Test design rubrics generally reward a single clear change, sound randomization, a well-chosen primary measure, reasoned sample size and duration, and careful interpretation, plus presentation and sources. The change criterion earns credit when exactly one variable differs between versions. Randomization scores highest when the unit of assignment and consistency over time are addressed. The measure criterion rewards choosing the outcome that reflects the business result and adding guardrails. Sample size is judged on explicit reasoning with numbers, and duration on commitment in advance. Interpretation earns full points when a null result is explained in terms of what effect sizes it rules out, rather than dismissed or overclaimed.
BU 443 Unit 6 help: mistakes that cost points
Test designs often lose points by changing several things at once, making any result impossible to attribute. Another common problem is choosing conversion rate as the only measure when the change affects order size or costs. Students frequently skip sample size reasoning, or pick a duration based on convenience. Some plan to stop the test when results look good, which inflates false positives. Others report a non-significant difference as proof of no effect, or as a small win. Many forget guardrail measures that reveal harm elsewhere. Finally, reading results without first checking that the split between versions was balanced can miss a broken test. Writing the decision rule before launch, as the sample does, removes the temptation to reinterpret the result afterward.
Get a BU 443 Unit 6 example written to your instructions
Forward the Unit 6 instructions with the BU 443 rubric from your classroom and any scenario the assignment fixes. We write a custom example that varies one element, explains assignment before results, commits sample and duration in advance, reports counts beside outcomes and reads the null case. First custom sample free, returned in 24 to 48 hours.
More BU 443 papers and marketing samples
- BU 443 Unit 1: Metrics definition exercise
- BU 443 Unit 2: Data collection and tracking audit
- BU 443 Unit 3: Channel performance analysis
- BU 443 Unit 4: Customer and cohort analysis
- BU 443 Unit 5: Campaign evaluation
- BU 443 Unit 7: Dashboard or visualization
- BU 443 Unit 8: Analytics recommendation
- BU 245 Unit 6: Proposal
- BU 302 Unit 6: Influence tactic comparison
- BU 321 Unit 6: Two sample comparison paper
- BU 380 Unit 6: Strategic commitment paper
BU 443 Unit 6 questions, answered
Do I need real data to write this?
No, and most sections supply a scenario or expect a constructed one. Say clearly that the figures belong to the case, then treat them exactly as you would treat observed ones. The grade sits in the design reasoning and in how carefully you read the result, not in where the numbers came from. Constructed figures are only a problem when presented as measured.
How do I reason about sample size without a formula I trust?
Reason in the open and follow whatever your course materials supply. State the baseline rate you start from, the size of change that would actually be worth acting on, and the fact that smaller differences require far more observations. Where your section teaches a calculation or a tool, use it and cite it. What fails is a number with nothing behind it.
Can I test something other than a web page?
Usually yes, and subject lines, advertising creative, landing content and offer wording all fit the same structure. Check your instructions first, since some sections fix the setting. Whatever you choose, the requirements do not change: one element varied, comparable groups, a duration set in advance and a result read against a difference you decided beforehand was worth finding.
Where can I find a free BU 443 Unit 6 sample paper?
This page offers the complete BU 443 Unit 6 test design and reading for free, including the sample size reasoning, the results table and margin notes. For a different test, forward the assignment and we prepare a custom design free, typically within 48 hours.
What if nothing changes in the BU 443 Unit 6 test?
Explain what the test can rule out. In the sample, the interval for conversion shows no large lift, profit per visitor did not improve, and the decision rule written in advance keeps the current threshold.