Why it matters

A WNS that improves from -0.10 to -0.04 looks like progress. If part of it came from lowering a margin, the design has not improved by that part. Normalising each run to the baseline margin shows the real trend.

Worked example

1
2
3
4
5
6
7
8
9
10

runs = [("r1", 0.20, -0.10), ("r2", 0.18, -0.05), ("r3", 0.18, -0.04)]     # (run, total margin ns, WNS ns)
base_margin = runs[0][1]
norm = {r: round(w + (m - base_margin), 3) for r, m, w in runs}      # WNS as if the baseline margin had been kept
raw_gain = runs[2][2] - runs[0][2]
real_gain = norm["r3"] - norm["r1"]
assert norm == {"r1": -0.10, "r2": -0.07, "r3": -0.06}
assert round(raw_gain, 3) == 0.06 and round(real_gain, 3) == 0.04
margin_part = round(raw_gain - real_gain, 3)
assert margin_part == 0.02
print("PASS: raw WNS gain", round(raw_gain, 3), "ns; real gain", round(real_gain, 3), "ns; from margin cuts", margin_part, "ns")

From r1 to r3 the reported WNS improved by 0.06 ns. Of that, 0.02 ns is simply the margin being reduced from 0.20 to 0.18 ns; the real improvement from design work is 0.04 ns. Neither number is wrong: the first is what the signoff will see if the new margin is approved, the second is what the design did. A dashboard that shows both prevents arguments.

What the review looks for

• Margin changes appear as events on the trend, with the owner who approved them.

• WNS and TNS trends are shown at the baseline margin as well as at the current one.

• Margin increases are treated as risk reduction, and decreases as a decision needing approval, not as progress.

Related topics

• What to Track

• Next: Comparing Runs Honestly