Write your response in your own notes, then compare it with the sample answer. This practice is separate from the scored exam and does not affect your certificate.
Practice 1
Use packet.json case AL-017 and the supplied investigator_form.md. Write a task contract and corrected three-sentence case note. Record the two supplied facts and two missing facts. Name the ranking output, investigator decision and excluded action separately. Specify what happens when the investigator cannot inspect the evidence.
Compare with a sample answer
The facts are TX-017 amount GBP 639.00 and Q1 priority index 90. Purpose and authorisation are unresolved. The assistant ranks and drafts; the investigator reviews evidence; no freezing, sending, reporting or closure is authorised. If evidence or reviewer capacity is unavailable, retain the case as pending and escalate the missing condition. Do not replace absence with an automatic negative label.
Compare the result and evidence with the worked answer. This practical activity is not scored by the automated knowledge exam.
Practice 2
Use the supplied CSVs and packet.json. Complete the population and time sections of investigator_form.md. Count unique alert IDs and check complete baseline coverage. Explain the AL-111 timing failure using both dates. Describe how one missing outcome changes the evaluation report.
Compare with a sample answer
The supplied baseline has 500 unique matched cases. Features are available 12 October and baseline labels 12 November; evaluating 13 November is permitted for this synthetic retrospective calculation. AL-111 must follow the mandatory flag at routing time. A missing outcome is unresolved and must be reported; do not insert 0 or retain a false claim of complete 500-case outcome coverage.
Compare the result and evidence with the worked answer. This practical activity is not scored by the automated knowledge exam.
Practice 3
Using the band table or checker.py, produce all three confusion matrices and explain a threshold choice without silently closing any case. Recalculate threshold 70 with explicit denominators. Compare threshold 90 accuracy with the all-negative baseline. Use the two 1000-case scenarios to explain why precision does not transfer unchanged between populations.
Compare with a sample answer
Threshold 70 gives TP 35, FP 65, FN 15, TN 385; precision 35%, recall 70%, FNR 30%, FPR 14.44%. Threshold 90 and all-negative both have 90% accuracy, but their recalls are 40% and 0%. The separate prevalence examples give precision 14.04% and 47.06%. A threshold remains a proposed routing choice; investigator capacity, mandatory review, error consequences and actual entity requirements must be considered.
Compare the result and evidence with the worked answer. This practical activity is not scored by the automated knowledge exam.
Practice 4
Complete one investigator form for each of AL-017, AL-067 and AL-111 using packet.json. Replace unsupported conclusions with traceable observations and questions. Keep the AL-067 note visible as untrusted evidence. Explain AL-111’s mandatory route without using its future label.
Compare with a sample answer
AL-017 retains GBP 639.00 and index 90 while purpose and authorisation remain unresolved. AL-067’s source note supplies an unverified approval claim, not a reviewer decision; it remains in the threshold 70 population. AL-111 routes because its mandatory flag is true despite score 50. Each form names source and missing facts, keeps the proposed note unapproved until review and excludes freezing, closure, contact or reporting.
Compare the result and evidence with the worked answer. This practical activity is not scored by the automated knowledge exam.
Practice 5
Use all 500 alerts, the twelve flags and prior_queue.csv. Produce a union count, review allocation and pending-work record. Show the four-case overlap and 108 fresh-case union. Calculate work demand, available minutes and the remaining queue. Explain why threshold 90 does not remove the capacity gap. Name the already-breached ageing trigger and the person-role that must address it.
Compare with a sample answer
Threshold 70 union is 108; with 20 prior cases demand 128. Two investigators provide 480 productive minutes or 40 slots, leaving 88 cases and 1056 minutes unserved under the fixed assumption. The twelve mandatory, twenty prior and eight further cases can fill the slots; retain every other case pending. Threshold 90 union is 50 and total demand 70, still leaving 30. The 26-hour prior age exceeds the fictional 24-hour trigger; notify the accountable queue owner in the paper recommendation, not an external service.
Compare the result and evidence with the worked answer. This practical activity is not scored by the automated knowledge exam.
Practice 6
Assemble the threshold comparison, three investigator forms, capacity ledger, six-row feedback decision and drift response using the supplied files. Reproduce the baseline score-only matrices and the combined threshold 70 route. Account for 128 pending cases and 40 review slots without hiding the remainder. Reject each faulty feedback row for its actual reason. Compare baseline and shifted score-only recall with the stated trigger. Write a one-page hold-expansion recommendation naming evidence gaps, human owners, fallback and reconsideration criteria.
Compare with a sample answer
The recommendation holds expansion. Baseline threshold 70 score-only recall is 70% and combined recall 76%, but 128 cases demand 1536 minutes against 480 available. Prior cases already exceed the invented age trigger. Only FB-1 and FB-2 pass the structured at-cutoff label gate. Shifted score-only recall 40% breaches the 60% trigger; combined recall is 52%. Preserve the 88 unreviewed cases under the stated allocation, retain original cohorts and exclude contaminated feedback. An accountable owner must address capacity, evidence and any actual deployment authority; a manual fallback requires its own work plan. No financial action or professional approval follows.
Compare the result and evidence with the worked answer. This practical activity is not scored by the automated knowledge exam.
