| Course | MT358 Social Media Marketing and AI Optimization |
|---|---|
| Unit | Unit 5 |
| Paper type | Optimization exercise |
| Length | About 1,020 words, 6 pages double-spaced |
| Format | APA 7 student paper |
| School | Purdue University Global |
| Program | BS in Business Administration |
| Updated | October 2026 |
Free sample paper for MT358 Unit 5
The Tool Wanted Clicks; the Studio Wanted Knitters: An MT358 Unit 5 Optimization Exercise With Two Overrides
[Student Name]
Purdue University Global
MT358: Social Media Marketing and AI Optimization
Unit 5 Optimization Exercise
[Instructor Name]
[Date]
The studio, tool and results are composite; the tool is described as offered in September 2026, and the research cited is real.
Setup
The tool is a caption and image assistant built into the ad platform, as offered in September 2026. It suggests captions and edits images, then predicts which version will perform best. The studio asked it to maximize link clicks, the default setting. That choice matters, because the studio's real objective is profitable purchases from knitters, not clicks.
Five ad cells ran for fourteen days at $280 each, $1,400 in total, all aimed at the lookalike audience from Unit 4 and all promoting the Harbor Ice colorway. Every cell is reported below in full, including the results that flatter the tool, so the two overrides that follow cannot look like a selective reading of the data.
Results
The first cell paired a pattern-pairing caption, naming a cardigan pattern and its designer, with the original three-light photo. It drew a click-through rate of 1.20 percent, or 297 clicks, and eleven purchases, a conversion rate of 3.7 percent. It cost $25.45 per purchase and returned 3.06 dollars in revenue for each dollar spent, the best return of the test.
The second cell used the same caption with the tool's enhanced photo, which brightened and saturated the colors. Its click-through rate rose to 1.40 percent and its return was 2.99, nearly the same. But three of its eleven buyers wrote to complain that the yarn looked different from the photo.
The third cell used a dye-lot caption, explaining how many skeins of Harbor Ice were available in one lot for sweater knitters. Its click-through rate was 1.30 percent and its return was 2.54, a respectable result for a caption that speaks mainly to one segment.
The fourth cell used the tool's top-ranked caption: only 40 skeins of Harbor Ice will ever be dyed. Its click-through rate was the highest, 2.00 percent, or 488 clicks, and its cost per click was the lowest. But those clicks produced only eight purchases, a conversion rate of 1.6 percent, and a return of 2.00.
The fifth cell used a trend caption with popular hashtags. Its click-through rate was 1.10 percent and its return was 0.96, the only cell that lost money outright.
What the Tool Got Right and Wrong
The tool correctly predicted that the scarcity caption and the enhanced photo would draw more clicks. It was optimizing what it was asked to optimize. Dietvorst et al. (2015) found that people often stop trusting algorithms after seeing them make mistakes, even when the algorithms outperform humans. The point here is different: the tool did not make a mistake on its own terms. The studio set the wrong target.
The tool did exactly what it was asked, which is why it recommended the two things the studio could least afford.
Override One: The Scarcity Line
Recommendation rejected: the caption announcing that only 40 skeins would ever be dyed. Reasons, in order of weight: it was untrue, since the studio dyes Harbor Ice to order and no limit existed; it drew clicks from people who bought far less often, 1.6 percent of clicks against 3.7 percent for the pattern pairing; and it uses a tactic the Unit 2 research found knitters refuse. Inman et al. (1997) found that restrictions such as purchase limits make an offer seem more valuable, which helps explain why the invented limit drew taps; it did not make the people tapping into knitters. Its return of 2.00 was the lowest of all five cells despite the highest click-through rate. A false scarcity claim may also mislead consumers, which the studio should avoid regardless of results. Decision: never use scarcity language unless stock is actually low, and then state the exact number.
Override Two: The Enhanced Photo
Recommendation rejected: the brightened, saturated photo. Reasons: it lifted click-through by about a sixth, but three of eleven buyers wrote to say the yarn looked duller than pictured, a complaint rate that would damage trust and cost refunds over time. Its return was nearly the same as the original photo's, so the higher click-through rate bought nothing. The research found that knitters value honest daylight photos above almost anything. Decision: use only unedited or color-corrected photos that match the yarn.
What to Change in the Tool
The studio will change the tool's objective from link clicks to purchases, and will add a rule that no caption may claim scarcity and no photo may be color-enhanced. Castelo et al. (2019) found that people are more willing to accept algorithms for tasks seen as objective than for those seen as subjective. The studio will use the tool for objective tasks, such as testing caption length, and keep human judgment for claims about the product.
What the Override Memos Share
Both memos follow the same four steps: what the tool recommended, what the data showed, what the community research predicted and the decision. That structure matters because an override that rests only on taste invites the next staff member to undo it. Here, each rejection rests on a fact about the product, a result from the test and a finding about the audience, so reversing it would mean arguing against all three.
Limits
Five cells of $280 for fourteen days produced small numbers of purchases, between about seven and eleven per cell. A two-proportion test on purchases per click, eleven of 297 for the pattern pairing against eight of 488 for the scarcity line, gives a z of 1.82, short of the 1.96 usually required at the 5 percent level. The gap between the two captions is suggestive, not proven, and differences between the first three cells could easily be chance. The complaint count in cell two is small but serious, because a complaint represents a buyer who may not return. A larger test would be needed to rank captions confidently.
References
Castelo, N., Bos, M. W., & Lehmann, D. R. (2019). Task-dependent algorithm aversion. Journal of Marketing Research, 56(5), 809-825. https://doi.org/10.1177/0022243719851788
Dietvorst, B. J., Simmons, J. P., & Massey, C. (2015). Algorithm aversion: People erroneously avoid algorithms after seeing them err. Journal of Experimental Psychology: General, 144(1), 114-126. https://doi.org/10.1037/xge0000033
Inman, J. J., Peter, A. C., & Raghubir, P. (1997). Framing the deal: The role of restrictions in accentuating deal value. Journal of Consumer Research, 24(1), 68-79. https://doi.org/10.1086/209494
The Purdue Global MT358 Unit 5 instructions
Unit 5 asks students to test an AI optimization tool against a business objective and decide whether to accept its recommendations. The tool should be described plainly, with its date and the goal it was set to pursue. Every test result should be reported, not only the ones that support the decision. Each rejected recommendation needs its own short memo explaining what the tool suggested, what the data showed and why the business overruled it. Instructors expect students to separate the tool's objective from the business's, since most disagreements start there. Claims about how a tool works internally should be avoided unless documented. Small samples must be acknowledged, ideally with a simple significance check.
How the MT 358 Unit 5 example is put together
The sample's setup names the tool, its September 2026 version and its click objective, then explains why that goal differs from the studio's. All five cells are reported with click-through, purchases and return. A section on what the tool got right places the failure in the objective, under a highlighted line that the tool did exactly what it was asked. Override one rejects the invented limit first because it was untrue, then on conversion, citing Inman and colleagues on restrictions. Override two rejects the enhanced photo on color accuracy. A section explains what the two memos share. Changes to the tool's settings follow, then a limits section with the two-proportion test.
Reading the MT358 Unit 5 grading rubric
Marks on an optimization exercise usually follow the clarity of the setup, the completeness of reported results, the reasoning in each override and the honesty of the limits. A setup that names the tool's objective shows the student understands how optimization works. Reporting every cell, including those that favor the tool, protects credibility. Override memos earn more when they rest on several grounds, such as truth, data and audience research, rather than taste. Recommendations for changing the tool's settings show practical thinking. A significance check or a clear statement of sample size demonstrates numeracy. Citations to research on algorithms and consumer response support the argument, formatted in APA 7, and figures across the tables must agree.
MT358 Unit 5 help from the desk
A common weakness is treating the tool as either always right or always wrong. Students who accept every recommendation miss the conflict between clicks and the business's goal; students who reject everything rarely show evidence. Another problem is reporting only the cells that support the decision. Some papers describe the tool's inner workings with confidence they cannot support. Others claim a winner from eleven purchases, which no reader should accept without a test of whether the gap could be chance. Before writing the overrides, write down the tool's objective and the business's objective side by side. If they differ, explain the gap first, and the overrides will follow from it rather than from preference.
Get an MT358 Unit 5 example written to your instructions
Which tool does your Unit 5 prompt ask you to use, and on what content? Send any test results or screenshots you have with the rubric. A first custom exercise, free and ready within 24-48h, shows every variant's results, computes each metric the same way and defends the point where a person overrules the tool. The paper above is an original model document written by our desk, not a submitted student paper and not an official Purdue University Global document.
More MT358 papers and BS in Business Administration samples
- MT358 Unit 1: Brand Community Misstep Post
- MT358 Unit 2: Community Research Brief
- MT358 Unit 3: Six-Week Social Content Plan
- MT358 Unit 4: Paid Social Targeting Memo
- MT358 Unit 6: Seminar Reflection
- MT358 Unit 7: Influencer and Disclosure Plan
- MT358 Unit 8: Crisis Response Protocol
- MT358 Unit 9: Social Analytics Report
- MT358 Unit 10: Social Campaign Plan
- MT357 Unit 5: Digital Content Calendar Plan
- MT355 Unit 5: Survey Pilot Revision Memo
- MT340 Unit 5: Negotiation Preparation Plan
- MT330 Unit 5: Product Adaptation Analysis
MT358 Unit 5 questions, answered
Do I have to agree with the tool's recommendation?
No, and these units often reward a well-argued disagreement. What matters is the reasoning: what the tool optimized for, what the business actually needs, and where they diverge. If you accept the recommendation, explain why it serves the objective. If you overrule it, give specific evidence, such as results, accuracy concerns or community norms.
How should I describe an AI tool I used?
Name it, state the version or date you used it, and describe what you asked it to do and what it produced. Avoid claims about how it works internally unless the provider documents them. Graders are usually testing your judgment about the output rather than technical knowledge of the tool, so keep the description short and factual.
What if my test sample is small?
Say so and draw conclusions accordingly. A small test can still show direction, but it rarely proves a difference. Where possible, compute a simple significance check or state the sample sizes clearly. Decisions can rest on other grounds too, such as a false claim or a customer complaint, and saying which ground you relied on strengthens the paper.
Where can I find a free MT358 Unit 5 sample paper?
This page holds the complete MT358 Unit 5 optimization exercise, in which a yarn studio overrules an AI tool's scarcity caption and brightened photo.
Why is the scarcity caption rejected first on truth in the MT358 Unit 5 sample?
The studio dyes the colorway to order, so a 40-skein limit did not exist; an untrue claim is rejected whatever its results, and its conversion was also weak.