The catch test is the source of truth. Spread over a measured strip, weigh what landed, and we back-calculate the rate your machine actually delivers — then compare it to your target and tell you which way to turn the dial.

nitrogen basis → product rate (Penn State):
  product rate = target N ÷ (N% ÷ 100)

how much should land on the strip:
  pass area    = swath × strip length
  target weight = product rate × pass area ÷ 1,000

catch-test back-calculation (TheYieldGrid):
  actual rate  = caught weight × 1,000 ÷ pass area

deviation + band:
  deviation %  = (actual − target) ÷ target × 100
  |dev| ≤ 5%   → calibrated   (safe)
  |dev| ≤ 10%  → caution
  |dev| ≤ 15%  → caution (upper edge)
  |dev| > 15%  → danger
  actual < target → increase the setting · actual > target → decrease

total product + bags (when area given):
  total lb     = product rate × (area ÷ 1,000)
  bags         = ceil(total lb ÷ bag lb)
  leftover     = bags × bag lb − total lb

Worked example: caught 0.75 lb over a 10 ft × 100 ft strip (1,000 sq ft) against a 2.5 lb / 1,000 target → actual 0.75 lb / 1,000, that's −70% (DANGER), so increase the setting and re-test. And the nitrogen front end: 0.5 lb N / 1,000 from a 20% bag → 0.5 ÷ 0.20 = 2.5 lb product / 1,000 (a 20-0-0 bag at 1.0 lb N / 1,000 → 5 lb product, matching LSU AgCenter).

The ±5% "calibrated" window (4.75–5.25 lb / 1,000 for a 5.00 target) follows UMass; the 20-0-0 → 5 lb worked example follows LSU AgCenter, whose page uses exact-match wording with no stated tolerance — the two are not conflated. The ±5 / ±10 / ±15% banded verdict and the increase / decrease / OK instruction follow TheYieldGrid; the lb N ÷ N% front end follows Penn State Extension. See Sources below.