
use my data-log model to generate data, change one part at a time, and explain the model’s limits and how the results improve the design.
Phase 4: Build
Do Now. Car makers crash-test cars with dummies before people ride in them. Why test a model before using the real thing? Write two reasons.
Finding a weakness in a model is a sign the engineering process is working.
Today we try to break the model. We feed it sample weeks, including strange ones, and watch the outputs. Then we change one part at a time and see what happens.
Test 1: a normal pair of weeks. Sample data. Food waste week 1: 20.0 kg; week 2 after the action: 17.6 kg. Change = 2.4 ÷ 20.0 × 100 = 12%. Green flag. The model works as planned.
Test 2: a short week. Sample data. Week 2 had a holiday, so only 4 school days. Trash dropped from 40 kg to 32 kg, a 20% cut. Is that the action working, or just fewer days? The model gave a misleading green flag.
Worked example 1: changing a part to fix the problem. Hana’s team changes the calculation to kilograms per school day. Week 1: 40 ÷ 5 = 8.0 kg a day. Week 2: 32 ÷ 4 = 8.0 kg a day. Change: 0%. The action did not cut trash at all. Changing one part of the model changed the conclusion.
Test 3: runoff and rain. Sample data. A team tries a runoff reading for the add-on. Week 1 runoff: 60 liters with 2 cm of rain. Week 2: 90 liters with 4 cm of rain. Runoff went up, but rain doubled.
Worked example 2: normalizing. Liters per cm of rain: week 1 = 60 ÷ 2 = 30; week 2 = 90 ÷ 4 = 22.5. Per centimeter of rain, runoff actually fell by 25%. Rain is a variable we cannot control, so the model must account for it.
Every model has limits. Ours cannot tell whether a scale was zeroed, whether a bin was skipped, or whether a big event, like a class party, changed the waste. We write these limits down and plan around them.
| Test | Part changed | Output before | Output after | What it taught us |
|---|---|---|---|---|
| 1 | ||||
| 2 | ||||
| 3 |
Run each test on your model. All numbers are sample data.
Write the model report.
List three limitations of your model. For each one, say whether you can fix it, work around it, or only report it.
Explain how today's test data plus our criteria will change your monitor before the 2-week test. Be specific about one column or calculation.
You tested your model like real engineers. Next class is the mid-build checkpoint.
Exit ticket: what is the most important change you made to your model today, and what data led you to make it?
How confident are you that you can I can use a model to test my solution and improve it.
Materials: the class data-log template, three printed sample-week scenario cards (normal, short week, rainy week), calculators.
Setup: hand out scenario cards one at a time so teams discover the problem before seeing the fix. All numbers are sample data.
Listen for: “the model is wrong, start over” (fix one part instead); thinking normalizing “fakes” data (it makes comparisons fair).
Support / stretch: pre-fill one scenario’s calculations. Stretch: design a fourth stress test, such as a class party week, and predict the output before running it.