In MT358's Unit 5 optimization exercise, an AI caption tool picks a false scarcity line and a color-shifted photo, and the studio's own test results support overruling both. Searches like "mt 358 unit 5 assignment example", "mt358 unit 5 sample" and "mt358 unit 5 example" land here.
What a finished MT358 Unit 5 optimization exercise looks like
Five pages: a setup, a results table, two override memos and a limits note. The setup describes the tool as a caption and image assistant inside the ad platform, as offered in September 2026, and states that it was asked to maximize link clicks. Five cells of $280 each ran for fourteen days. The table reports, by script: the pattern-pairing caption with the original photo at 1.20 percent click-through, $25.45 per purchase and 3.06 return on ad spend; the same caption with the tool's enhanced photo at 1.40 percent and 2.99, plus three color complaints from eleven buyers; a dye-lot caption at 2.54; the tool's scarcity pick at 2.00 percent click-through but 2.00 return; and a trend caption at 0.96. Override one rejects the scarcity line; override two rejects the enhanced photo.
How a MT358 Unit 5 example is structured
The tool's objective is stated before any result, because the exercise turns on it: the assistant optimized clicks, while the studio needed profitable purchases from knitters who would keep the yarn. All five cells are reported in full, including those that flatter the tool, so the overrides cannot look selective. Each override is a short memo with the same parts: what the tool recommended, what the data showed, what the community research predicted, and the decision. The scarcity line is rejected first on truth, since no 40-skein limit existed, and only second on return. The enhanced photo is rejected on color accuracy, the complaint knitters raised most. A limits note closes the paper: with 11 and 8 purchases, the gap between the leading captions is suggestive, not proven.
What the tool was told to want
The assistant was set to maximize link clicks, the default goal offered when the test was built. The exercise names that setting first, because a tool optimizing clicks will favor whatever makes people tap, true or not.
Five cells, all reported
Each cell spent $280 over fourteen days. Click-through ran from 1.10 to 2.00 percent and return on ad spend from 0.96 to 3.06, and all five rows appear, including the scarcity line's lead on clicks and cost per click.
Override one: a limit that did not exist
The scarcity caption drew 488 clicks but eight purchases, a 1.6 percent conversion rate against 3.7 for the pattern-pairing caption. The memo rejects it first because the studio dyes to order, making the line untrue.
Override two: a brighter skein than the real one
The tool's enhanced photo raised click-through by 16.4 percent, but three of eleven buyers wrote to say the yarn looked duller than pictured. Color mismatch was the community's top complaint, so the original photo stays.
What eleven and eight purchases can show
A two-proportion test on purchases per click gives z of 1.82, short of the conventional threshold. That limit is stated outright, and the decisions rest on the false claim and the complaints rather than on sales counts.
Where marks go in MT358 Unit 5
MT358 optimization work is judged on the person, not the tool, so reports that list what an optimizer recommended and adopt it earn little however polished the tables look. What scores is a stated objective for the tool, results shown for every variant, and a defended decision where the tool's choice and the business's interest diverge. Overrides grounded in truthfulness, community norms or returns usually outscore those grounded in taste. Metrics should be computed consistently, with click-through, cost per click, cost per purchase and return on ad spend each defined once. Small samples presented as conclusive draw comment. Descriptions of a platform's AI features should carry a date and avoid claiming how the feature works internally, since that is rarely published.
Get a 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.
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.