Trainer error-mapping analysis - every error sorted into not-trained, trained-not-tested, trained-but-erred, or trained-on-old-version

You are the trainer at a central kitchen. The owner has
read the quarterly error log and said "training isn't
working." I will give you the quarter's error records and
the matching training records. Run the cross-analysis:
assign every error to one of four buckets and identify
which link of the training system the gap sits in.

Four buckets (every error lands in exactly one, assigned
on records, not impressions):
1. Coverage gap: the person involved never received
   training on that operation - new hire, transferred,
   or on leave during the session. Evidence: name absent
   from the attendance list
2. Verification gap: trained, but no assessment record -
   or the assessment type doesn't match the operation (a
   batching error by someone who passed a written test).
   Evidence: training record present, assessment record
   absent or type-mismatched
3. Execution gap: assessed and passed, erred anyway.
   This bucket is not a training failure - it points at
   SOP design (too many steps, not enough time) or
   floor management (steps skipped under output
   pressure). Evidence: a passing assessment predates
   the error
4. Version gap: trained on the old SOP; the change never
   triggered retraining. Evidence: the training
   content's version number predates the SOP effective
   date

Input:
- Error records: date / workstation / description
  (wrong-batch ingredient addition, mislabeled packs,
  skipped temperature log, wrong sanitizer) / employee
  ID / consequence level
- Training records: session / content version number /
  attendance list / assessment method and result
- SOP change log: changed item / effective date /
  retraining coverage record

Output:
- Table 1: distribution and share across the four
  buckets plus "needs interview," broken down by error
  type
- Table 2: the largest bucket expanded - which
  workstation, which operations, which people
- Table 3: next quarter's training-plan revisions. Every
  line lands on four elements: audience (which group) /
  format (demo, hands-on, micro-lesson) / hours /
  assessment method. The version-gap fix does not
  belong in the training plan - it belongs in the change
  process; name who it goes to and what gate gets added
- Table 4: the execution-gap list (passed to floor
  management as-is) - facts and the SOP clauses involved
  only, no remediation advice; that is the floor's call

Constraints:
- "Strengthen training" and "raise awareness" are
  banned - nothing that cannot be checked for having
  been done
- The execution gap must not absorb blame that belongs
  to floor management: state plainly that training does
  not fix this bucket, and that piling hours onto it is
  waste
- Classification follows the records; anything that
  does not reconcile goes to "needs interview" with a
  note on which record is missing and who can supply it.
  "Probably" is not a bucket
- If "needs interview" exceeds three in ten items,
  output the record-completion list first and suspend
  the bucket distribution - no conclusions yet

Why it is written this way

The four buckets exist to dismantle "training isn't working" - a conclusion that can be neither rebutted nor acted on. A coverage gap is cured by scheduling sessions; a verification gap calls for changing the assessment (a written test does not predict a batching error - more hours will not help); an execution gap is not training's problem at all; a version gap belongs to the change process. Without the split, the trainer facing the accusation has exactly one move ("more sessions"), and it is the one move that fails against three of the four buckets - next quarter's numbers get worse, not better.

Calling out the execution gap as "not ours to fix" is the line most likely to be cut in review and least safe to cut. Pouring training hours onto execution gaps is the training department's largest recurring waste - an SOP with too many steps stays broken after a hundred repetitions. Assigning the bucket openly to the floor is what makes the analysis acceptable to floor managers in the first place; the data now backs a boundary that previously ran on argument. Table 4 passes facts without advice precisely to hold that boundary - hand the floor recommendations and the responsibility flows right back.

"Needs interview" with a three-in-ten circuit breaker protects the distribution's credibility. Force classification to completeness and you get "probably" assignments - get the shares wrong and next quarter's plan is wrong. The breaker says: past this level, the problem is no longer a training gap but a records gap; fix the records first. Admitting you cannot classify is both more honest and more useful than pretending you did.

The four elements for every plan revision (audience, format, hours, assessment) are a lock on the training plan itself. Drop any one and the plan degrades at execution into "gather everyone for a talk" - the talk happens, the sign-in sheet exists, the errors continue, and the "training isn't working" cycle starts its next lap.

Variables