3.1.2 · 35–55 minutes · investigate, calculate, discuss, review
What you will learn
- Assess provenance, relevance, measurement and incentives.
- Separate facts, estimates and inferences.
- Create a reproducible evidence register.
3.1 · The Investment Case
Check the evidence
The Investment Case · the work you will do
- 1Check the evidence
- 2Build the opposing case
- 3Test what matters
- 4Reach a judgement
Before you begin
A founder says loyalty is rising. An online poll says customers are unhappy. A payment file shows repeat visits. Which source answers which question?
Record your first answer before continuing.
Video coming soon · about 3 minutes
What Counts as Evidence?
Use the narration below while the recording is being made, then try the decisions in this lesson.
Read the full narration
Welcome to On Judgement. A founder says loyalty is rising. An online poll says customers are unhappy. A payment file shows repeat visits. Which source answers which question? Take a first view. You can change it when you have considered the evidence.
Evidence quality depends on the claim being tested. A payment record can help establish realised transactions, but not why a customer returned. An interview can reveal motives without measuring their prevalence. An annual report contains reported figures and estimates; it is not free of judgement. The task is to match the evidence to the question rather than declare one source type universally decisive.
Record the original source, date, population, period, unit and definition. Check gross versus net sales, visits versus unique customers, and menu prices versus realised receipts. A growing sample can still be selected badly. An average can hide a changing mix. Reconcile discrepancies before combining figures. Independent reproduction is stronger than repeating the same claim across several websites.
Management, competitors, short sellers and journalists may have incentives. That calls for checking their claims, not automatically rejecting them. Corroborate material facts and apply the same test to supportive and unwelcome evidence. A confident AI response can be a research lead; verify its citations and numbers against original material, and retain responsibility for the argument.
For each material claim, record source details, what it establishes, limitations, relevance and the next verification step. Mark missing evidence openly. Revise the interpretation when better data arrives, preserving the earlier entry. Sarah Ivory’s public critical-thinking outline informs the emphasis on arguments and evidence; these investing tasks and source cards are original On Judgement material.
Let's work through A loyalty measure with a denominator. DailyCup’s fictional payment sample contains 400 identifiable customers from one loyalty scheme; 240 return within 30 days. A separate voluntary social poll has 20 respondents, of whom 12 complain. Neither sample is stated to represent all customers. Pause here and try the calculation or reasoning before continuing.
Observed return share in the loyalty sample = 240/400 = 60%. Complaint share among poll respondents = 12/20 = 60%. The identical percentages describe different populations, questions and selection processes.
Neither establishes that 60% of all customers are loyal or dissatisfied. The results can coexist; investigate representativeness, definitions and timing.
Now try the choices in this lesson. If your answer changes, identify the reason. Create a reproducible evidence register. Use the case activities to put the idea into your own words, and return to the evidence when more than one conclusion is defensible.
Begin with the question
Evidence quality depends on the claim being tested. A payment record can help establish realised transactions, but not why a customer returned. An interview can reveal motives without measuring their prevalence. An annual report contains reported figures and estimates; it is not free of judgement. The task is to match the evidence to the question rather than declare one source type universally decisive.
Inspect how a number was made
Record the original source, date, population, period, unit and definition. Check gross versus net sales, visits versus unique customers, and menu prices versus realised receipts. A growing sample can still be selected badly. An average can hide a changing mix. Reconcile discrepancies before combining figures. Independent reproduction is stronger than repeating the same claim across several websites.
Consider incentives without dismissing facts
Management, competitors, short sellers and journalists may have incentives. That calls for checking their claims, not automatically rejecting them. Corroborate material facts and apply the same test to supportive and unwelcome evidence. A confident AI response can be a research lead; verify its citations and numbers against original material, and retain responsibility for the argument.
Use a living register
For each material claim, record source details, what it establishes, limitations, relevance and the next verification step. Mark missing evidence openly. Revise the interpretation when better data arrives, preserving the earlier entry. Sarah Ivory’s public critical-thinking outline informs the emphasis on arguments and evidence; these investing tasks and source cards are original On Judgement material.
A loyalty measure with a denominator
DailyCup’s fictional payment sample contains 400 identifiable customers from one loyalty scheme; 240 return within 30 days. A separate voluntary social poll has 20 respondents, of whom 12 complain. Neither sample is stated to represent all customers.
- Observed return share in the loyalty sample = 240/400 = 60%.
- Complaint share among poll respondents = 12/20 = 60%.
- The identical percentages describe different populations, questions and selection processes.
Neither establishes that 60% of all customers are loyal or dissatisfied. The results can coexist; investigate representativeness, definitions and timing.
Put it into practice
1. Compare six source cards
Assess: a till export; founder slide; loyalty sample; voluntary poll; competitor estimate; AI summary with an unverified link. State one use and one limitation for each.
Write your response in your workbook before opening feedback.
Compare with a model response
Till data needs reconciliation; a slide identifies management’s claims; a loyalty sample describes its members; a poll reveals respondents’ views; competitor estimates need methods; AI summaries need source verification. Use original evidence and question-specific tests.
2. Build three register rows
For transactions, repeat demand and sustainable cash, record a source, definition, uncertainty and next check.
Write your response in your workbook before opening feedback.
Compare with a model response
For transactions, reconcile till to receipts and refunds. For repeat demand, inspect cohorts and excluded customers. For sustainable cash, connect accounts to capex, working capital and tax assumptions. Do not invent unavailable documents.
3. Check an AI number
An AI answer says revenue is £500,000 and cites a page you cannot open. What should your memo say?
Write your response in your workbook before opening feedback.
Compare with a model response
Mark the figure unverified. Seek the original case ledger or report, check units/period and reproduce the calculation. Do not convert a plausible generated answer into a confirmed fact.
Jargon buster
Provenance
Where evidence came from and how it was produced.
Relevance
How directly evidence bears on the claim.
Selection
The process determining which observations enter a sample.
Corroboration
Checking a claim with additional appropriate evidence.
Inference
An interpretation drawn from observations and assumptions.
Evidence register
A dated record of claims, sources, limits and verification.
Knowledge check
Make each choice on the page. Feedback explains the reasoning. Use Try again after revisiting an idea.
AQA connection and further reading
AQA 7138: AO 2 application, AO 3 analysis and AO 4 evaluation (4.2), with command-word practice (4.4). Case connections include 3.1.4 finance, 3.3.3 strategy and 3.3.4 uncertainty. These original investing tasks and rubrics are not official AQA questions or mark schemes.
AQA 7138 specification
Primary curriculum reference; links and original classroom rubrics are interpretive, not endorsement.
Sarah Ivory · public course outline
Public outline only: quality arguments, evidence and guided practice informed the pedagogy. No paid course content accessed or reproduced.
Take forward
Assess provenance, relevance, measurement and incentives. Separate facts, estimates and inferences. Create a reproducible evidence register.