Evidence Plan, Interviews, and Bias
10 min read
Lesson objectives
- Analyze a professional situation by specifying claims, required evidence, reliability, and alternative explanations before interviews and analysis and justify a reviewable decision.
Please sign in to use all the tools
Once you sign in, the reading tools are yours: larger or smaller text, typeface, line spacing, focus mode, highlighting and bold, and notes saved to your account and gathered in My study.
You arrive at a factory with an initial margin explanation and find production and sales blaming each other. First build a plan that makes disagreement testable. Record each manager's claim precisely. 'Sales sells at a loss' is broad; 'discounting on the first product rose over the last two months' can be tested against invoices. That translation identifies the needed records and reduces the influence of seniority or a loud voice on the conclusion.
An evidence plan specifies the claim, its test source, and the comparison before collecting large files. Request net price, quantity, unit cost, and the same period. Margin analysis fails when current prices are combined with old costs. Also distinguish an announced discount from a rebate recognized later through a credit note. Matching definitions matters more than having a large apparently complete table. Ask the accountant how each field is produced, then trace one row to its supporting record.
Begin an interview with the most recent discount decision rather than asking whether policies are good. Ask the employee to describe the customer, approval, and available alternative. Then ask what would have happened if the discount had been refused. This reveals decision logic but remains an account to compare with records. Use neutral follow-ups such as 'What information was missing?' They uncover constraints without making the speaker feel that you are collecting an admission against them.
People remember one large deal more readily than dozens of small ones. Compare the salient story with the full order distribution. A manager may cite a heavily discounted customer whose total sales are small. Calculate its contribution to overall revenue before judging importance, then check whether the same policy applies elsewhere. This prevents a true episode from becoming an explanation broader than the evidence supports.
Keep quotations, observations, and inferences in separate fields. A quotation proves someone said something, not that it is true. An observation describes what you saw at a defined time and place. An inference connects evidence to a hypothesis. If you write 'centralization delayed discount approval,' show waiting time and approval sequence. Missing order information may explain the delay instead. This separation lets the client challenge the reasoning without denying the observation or attacking the speaker.
Stop interviewing when another interview repeats the evidence without closing a decision-relevant gap. Repeated management accounts are insufficient if line workers or departed customers are absent. Review coverage by where knowledge sits, not just by head count. If an important group remains unrepresented, state the limitation and propose a safe way to reach it. This keeps research proportionate and prevents interview volume alone from being treated as proof of a strong conclusion.
Build a compact evidence register containing the claim, supporting evidence, disconfirming evidence, source quality, time period, and affected decision. An internal source close to an event is not necessarily neutral, while an authoritative external source may not transfer to the organization. Triangulate quantitative data, operating evidence, and accounts from different stakeholders. When they conflict, do not hide the conflict in an artificial average; explain why they differ and what could resolve it. Apply it here to govern execution and review.
Test discounting by calculating net price for the same product rather than average price across all products. The overall average may fall because customers buy more of a cheaper product even though each product's price is unchanged. Hold mix constant, then compare the remaining difference with discount policy. Recalculate a sample from original documents. If it differs from the extract, repair extraction before recommending pricing changes or judging the sales team.
Immediately banning discounts may lose orders that cover variable cost and use spare capacity. Examine order contribution, credit limits, and available capacity before imposing one rule. A balanced recommendation addresses unjustified discounts while allowing a controlled exception with a recorded rationale. Track margin, volume, and collection together. The policy should improve the factory's economics rather than one ratio at the expense of cash or the customer relationship.
A manufacturer's margin fell from 31% to 24%. A margin bridge attributes three points to discounting, two to product mix, one and a half to waste, and leaves half a point within measurement error. Use order and line data from the same period rather than interview blame. Stop unauthorized discounts, review the highest-waste line, and reprice the low-margin product. Because waste records cover only eight weeks, describe that contribution as probable rather than settled until a longer production cycle is observed. To verify the reasoning, the learner explains which case evidence changed the recommendation in Evidence Plan, Interviews, and Bias and what evidence could reverse it.
Weight a source by its relationship to the claim. An operator knows the stoppage they observed but may not know final product margin. Finance knows aggregation but may miss a daily machine-setup fault. Match each question to the relevant knowledge holder. Also ask about incentives: does the speaker's evaluation depend on the reported number? An incentive does not make the account false; it calls for another source. Protect employee names in presentation notes and use role descriptions when sufficient. Review purpose and consent before sharing a sensitive quotation, distinguishing protection of identity from hiding a material operating problem from the decision maker.
Lesson summary
Evidence Plan, Interviews, and Bias separates evidence from assumptions, compares genuine alternatives, and links the recommendation to conditions and review indicators.