Chapter 8 — Independent review and defense
Lesson map
Lesson 08 published. Map, example, EX045.
flowchart TD
N0["Learning outcomes"]
N1["8.1 Independent review"]
N2["8.2 Viva preparation"]
N3["8.3 The viva"]
N4["8.4 Follow-up"]
N5["Answering a hostile question"]
N6["Lessons for future students"]
N7["Synthesis"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5 --> N6 --> N7
N7 --> EX["Worked example"]
EX --> Q["Exercise EX045"]
Logical drawing
flowchart LR
N0["Learning outcomes"]
N1["8.1 Independent review"]
N2["8.2 Viva preparation"]
N3["8.3 The viva"]
N4["8.4 Follow-up"]
N5["Answering a hostile question"]
N6["Lessons for future students"]
N7["Synthesis"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5 --> N6 --> N7 --> Q["EX045"]
[Learning outcomes]
|
v
[8.1 Independent review]
|
v
[8.2 Viva preparation]
|
v
[8.3 The viva]
|
v
[8.4 Follow-up]
|
v
[Answering a hostile question]
|
v
[Lessons for future students]
|
v
[Synthesis]
|
v
[Worked example] --> [Exercise EX045]
Learning outcomes
You will (1) obtain independent review; (2) prepare the viva; (3) plan follow-up.
8.1 Independent review
A second analyst at Corvane, not involved in tuning, reviews a random sample of fifty changes against the register's justifications. Forty-eight are judged appropriate; two exceptions are tightened.
Before submission, an independent reviewer who was not involved in the study reads the dissertation and checks whether conclusions follow from the evidence. Amara records each comment and her response, revising where the reviewer found claims stronger than the data supported. This step catches blind spots that close involvement makes hard to see, and it prepares her for examination.
8.2 Viva preparation
Amara expects questions on causation, generalisability and the lost detection. She prepares a slide showing the normalised series and the explanations table.
Preparing for the viva, Amara rereads the dissertation, lists likely questions about design, measures, alternatives and the missed detection, and prepares short evidence-based answers. She practises explaining her contribution in a few sentences. A mock viva with colleagues helps her handle unexpected questions calmly and admit uncertainty where evidence is limited, which examiners respect more than overconfident answers.
8.3 The viva
Examiners ask why recall was measured on only forty events. Amara explains the trade-off between label quality and quantity and proposes growing the set over time.
In the viva, examiners test whether Amara understands her work and its limits. She listens carefully, answers the question asked, and refers to specific evidence in the dissertation. When she does not know something, she says so and explains how it could be found out. She notes any corrections the examiners request, so they can be addressed promptly and accurately afterwards.
8.4 Follow-up
Corvane adopts quarterly exception reviews. Amara plans a short paper with Dr Brandt.
After the viva, Amara makes required corrections and records each change. She sends Corvane the practitioner summary, returns or deletes company data as the agreement requires, and archives her code and synthetic data. She also considers publishing a short paper, with the partner's agreement, so other small teams can benefit from the findings and method.
She thanks the analysts who took part, and offers a short session explaining results to the team.
Answering a hostile question
One examiner asks whether the study simply proved that switching off rules reduces alerts. Amara answers with evidence rather than defensiveness: only rules for absent services were disabled, each with proof; most reduction came from narrow exceptions and thresholds; and recall was checked on a fixed labelled set, with the one lost detection investigated and fixed. She adds that the concern is legitimate, which is exactly why the governance and recall checks were built into the design from the start. The examiner accepts the answer and asks for the governance section to be summarised in the abstract.
Lessons for future students
Amara writes a short note for future students on the programme. Agree data access and governance before the start date, because approvals took three weeks. Build the labelled set early, since it limits every later analysis. Measure normalised volume from the first day. And treat practitioners as co-investigators: Corvane's analysts suggested two of the most effective tuning changes. Dr Brandt adds the note to the programme's guidance for industry-partnered dissertations.
Worked example
The examiners accept the dissertation with minor corrections.
Synthetic case. The viva is synthetic.
Errors and limits
Independent review loses value if the reviewer helped design the changes.
Exercises
- EX045 — Review sample. What did the independent review find? Rubric: 3 = 48 of 50 appropriate, 2 tightened;; 2 = mostly fine;; 1 = vague;; 0 = none.
- EX046 — Reviewer independence. Why must the reviewer be uninvolved? Rubric: 3 = avoid bias;; 2 = independence;; 1 = vague;; 0 = none.
- EX047 — Small label set. Defend the 40-event label set. Rubric: 3 = quality over quantity, growth plan;; 2 = small;; 1 = vague;; 0 = none.
- EX048 — Viva slide. Which slide best answers causation questions? Rubric: 3 = normalised series and explanations table;; 2 = series;; 1 = vague;; 0 = none.
- EX049 — Adoption. What did Corvane adopt? Rubric: 3 = quarterly exception reviews;; 2 = reviews;; 1 = vague;; 0 = none.
- EX050 — Outcome. What was the viva outcome? Rubric: 3 = minor corrections;; 2 = pass;; 1 = vague;; 0 = none.
Synthesis
Independent review and an honest viva complete the study.
Continue studying — chapter 08 of 08
← Chapter 07 (draft) · Course chapter index · Worked results ↓ · Next course via chapter index →