Testing, Troubleshooting and Analyzing Performance

What is measured (and units), how it is measured (method and conditions), number of trials, and the pass condition.

Mean = 48 ÷ 4 = 12 s; range = 15 − 9 = 6 s.

One result could be a fluke; repeated trials show whether results are consistent and give a more reliable mean.

Mean = 13.2 ÷ 3 = 4.4 N, which is less than 4.5 N, so it fails.

The symptom is what you see (long waits); the root cause is why it happens (one person serving and making change).

So you know which change caused the improvement (or made it worse). Changing several at once hides the cause.

Parts substitution.

A test where only one variable changes and everything else is kept the same.

Add all the values and divide by how many there are.

The way something fails (tips, bends, leaks, crushes, overheats).

A step-by-step search for the cause of a problem and a fix.

8 − 3 = 5.

Sample: Why burnt? Baked too long. Why? Timer not set. Why? The baker was serving customers. Why? No one was assigned to the oven. Why? Roles were not planned. Root cause: no role assignment in the plan. Fix: assign an oven person and use a timer.

Mean = 20 ÷ 5 = 4 min. No: one customer waited 8 min, so the "each ≤ 5 min" criterion fails for 20% of customers. Look at individual results, not just the mean.

Test sections of the string separately (or check the fuse, then the plug, then each section) to narrow down where the fault is, instead of checking every bulb one by one.

They do not know which change fixed it. One change might not matter, or one might cause new problems. They should test one change at a time.

Sample: give the instructions to 3 people who have never seen the product; time them and count how many questions they ask. Pass: each person completes setup in under 3 minutes without asking for help, and rates clarity ≥ 4/5.

Both means = 18 ÷ 3 = 6.0 N. A's range is 0.4 N; B's range is 3.5 N. A is better: it is consistent and never fails the 5 N criterion, while B failed one trial (4.0 N). Consistency matters for safety.

Sum = 38; mean = 3.8 min (down from 4.7). Range = 5 − 3 = 2 min (down from 4). Customers over 5 min: 0, so the criterion is met 100% of the time. The fix worked and made waits more consistent.

Plan: load the centre with masses in 100 g steps, 3 bridges per design, record the load at failure and where it failed. Failure mode: note whether it buckled at the top, tore at the bottom, or crushed at the supports. Strategy: root-cause analysis (5 whys) on the failure point, e.g., it buckled because the top surface was thin; then change one variable (add a laminated top strip) and retest to compare means.