Cleaning verification audit - ATP results crossed with cleaning logs in four quadrants: is it a method problem or a discipline problem

You are the cleaning supervisor at a central kitchen. I
will give you two datasets: cleaning execution records
(zone, date, time, who cleaned) and verification results
(ATP readings or micro swab results). Cross-audit them
and separate, for every sampling point, a method problem
from a discipline problem.

The four quadrants (every sampled point lands in one):
- Logged + passing: normal; does not enter the report
  body
- Logged + failing: method problem - the cleaning is
  being done wrong or insufficiently; reclean and run
  the fixed investigation sequence
- Not logged + passing: discipline problem - the
  cleaning itself may work, but the record system has
  failed; on the day of an incident there is no way to
  prove it was done
- Not logged + failing: the worst cell - handled the
  same day: reclean the point, and the record-keeping
  gap is corrected the same day

Judging standards:
- ATP thresholds by surface class: direct food-contact /
  non-contact / hands - three bands entered into [the
  threshold table]. A point without a threshold cannot
  be judged: mark "threshold not set"; "reading seems
  high" is not a verdict
- A first failure gets one re-swab (sampling technique
  contamination is common). Re-swab passing is logged
  "pass on retest, initial suspect"; failing again
  counts as a true failure, both readings recorded
- Micro swab points carry fixed numbers and fixed
  locations - once points drift, before-and-after data
  is not comparable and trend analysis is void

Method-problem investigation (fixed order; record how
far it went and what was found): contact time →
chemical concentration → the cleaning tool itself (is
the cloth already contaminated) → task order (soil-last
or clean-last reversed) → training, last. No skipping
steps; no opening with "retrain everyone"

Output:
- Table 1: four-quadrant distribution (point counts and
  shares, split by zone)
- Table 2: investigation log for method-problem points
  (point / step reached / what was found / re-test
  result after reclean)
- Table 3: discipline-problem and worst-cell points with
  same-day disposition
- Table 4: fixed-point historical trend (when history
  exists) - flag only two classes: persistently rising,
  and consistently passing; no commentary on the rest

Constraints:
- Points with no threshold are not judged; say plainly
  "cannot be judged" - that inability is itself a
  finding
- No "raise cleaning awareness" - the quadrants have
  already located the problem as method or discipline;
  vague language stirs the two back together
- No trend conclusions from a single week's data

Why it is written this way

The four quadrants are the design's center of gravity. Read separately, cleaning logs and verification results each tell their own story - all boxes ticked, readings over limit, two documents at war. Crossed, they distinguish "cannot do" from "does not record": a method problem walks the five-step investigation, a discipline problem needs a records fix, and treating them as one blob cures neither. The forgivable-looking cell - not logged but passing - is the most insidious from an audit standpoint: it means that on the day something goes wrong, you will hold no proof the cleaning ever happened.

The one-retest rule respects the physics of sampling. ATP testing is sensitive to technique - swab pressure, a centimeter of position drift, someone talking over the plate - any of which manufactures a false positive on the first pass. Skip the retest and the team guts a perfectly sound cleaning procedure over a contaminated swab; run it, and after two weeks nobody takes real failures seriously either. Suspect-first-pass results are still retained - accumulated, they become the evidence that the sampler needs retraining.

Locking "retraining" into the last position of the sequence runs directly against management reflex. The instinctive response to a failed verification is a training session, but "logged and failing" is mostly caused by dilution ratios, contaminated cloths, and reversed task order - jumping straight to training announces that nobody intends to find the real cause, and the same failure returns next week. The fixed five-step sequence, logged as "reached step three, found X," makes the investigation itself auditable.

Fixed point numbering is the foundation under the trend table. Half the value of verification data is the week-over-week curve - a fixed point trending up predicts system decay earlier than any single reading - and the moment points drift, the history is void. Table 4 flags only "persistently rising" and "consistently passing" on purpose: the middle says nothing worth meeting about, and the two ends do.

Variables