Evidence-Based Product Decisions and Governance

Introduction

Product discovery is often described as a search for evidence about customer problems, desirability, feasibility, viability, and strategic fit. Yet evidence does not automatically produce a good decision. Product teams must interpret incomplete and sometimes contradictory signals, distinguish observation from assumption, communicate uncertainty, allocate scarce resources, and create an accountable record of why a decision was made.

This chapter examines the managerial system through which discovery evidence becomes a product decision. It focuses on five linked capabilities:

1. synthesising qualitative, quantitative, behavioural, technical, and commercial evidence;
2. expressing confidence without pretending that uncertainty has disappeared;
3. making explicit persevere, pivot, pause, or stop recommendations;
4. maintaining stakeholder alignment through transparent decision records; and
5. governing discovery across products, teams, and investment portfolios.

The central argument is that evidence-based product management is not simply a matter of collecting more data. It is the disciplined design of decision processes that make evidence relevant, comparable, traceable, and proportionate to the decision at hand.

Learning Objectives

By the end of this chapter, you should be able to:

• distinguish evidence, assumptions, interpretations, and decisions;
• evaluate evidence quality using validity, reliability, relevance, recency, and independence;
• construct a weighted discovery scorecard and state its limitations;
• assign and communicate confidence levels responsibly;
• formulate persevere, pivot, pause, and stop recommendations;
• maintain a decision log that supports organisational learning;
• manage stakeholder alignment and disagreement;
• apply governance principles to discovery portfolios and funding decisions; and
• identify ethical, regulatory, and organisational risks in evidence-based decisions.

1. From Discovery Evidence to Product Decisions

A discovery decision is a commitment about what the organisation should do next. It may concern a problem, a customer segment, a proposed solution, an experiment, a market, or the continuation of investment. A useful decision statement contains four elements:

• Object: what is being assessed;
• Evidence: what has been observed or measured;
• Uncertainty: what remains unknown;
• Action: what will happen next, including ownership and timing.

For example, “Customers like the concept” is not a decision. A stronger statement is: “Among 24 target administrators interviewed, 17 described the reporting problem as frequent and costly; however, willingness to pay remains untested. We will run a paid pilot with five qualified accounts before requesting further engineering funding.”

1.1 Evidence, assumptions, interpretations, and decisions

These categories should not be conflated.

A common failure is to present an interpretation as if it were evidence. For example, “Users need artificial intelligence” may actually mean that interviewees described a slow manual task. The evidence supports investigating the task; it does not yet establish that AI is the appropriate solution.

1.2 The decision should determine the evidence standard

Evidence is decision-relative. A low-cost prototype may justify a reversible decision based on directional evidence. A regulated launch, major infrastructure commitment, or public safety feature requires stronger evidence and formal review. Product leaders should therefore ask:

• What decision is being made?
• What would be the cost of being wrong?
• Is the decision reversible?
• Which stakeholders bear the consequences?
• What evidence would materially change the decision?

This prevents two opposite errors: over-investigating trivial choices and under-investigating consequential ones.

2. Synthesising Mixed Evidence

Mixed evidence is evidence from different methods or sources that answer related questions. Typical sources include interviews, field observation, surveys, product analytics, usability tests, support records, sales data, experiments, technical spikes, legal reviews, and competitor research.

The purpose of synthesis is not to average all evidence into one number. It is to establish what each source can legitimately tell us and then compare the findings against a clearly stated decision criterion.

2.1 Evidence dimensions

A practical evidence assessment uses five dimensions:

• Relevance: Does the evidence address the decision question and target segment?
• Validity: Does the method support the claim being made?
• Reliability: Would a similar process produce reasonably consistent results?
• Recency: Does the evidence reflect current conditions?
• Independence: Is the source sufficiently independent of the team’s desired conclusion?

These dimensions are not interchangeable. A large survey can be reliable in measuring stated preferences but weak in predicting actual purchase. A small usability test may have limited statistical generalisability but strong relevance for identifying interaction failures.

2.2 Triangulation

Triangulation compares different kinds of evidence to test whether they converge, complement one another, or conflict. Four useful patterns are:

• Convergence: interviews, behavioural data, and experiments point in the same direction;
• Complementarity: each source illuminates a different part of the decision;
• Discrepancy: stated preferences and observed behaviour differ;
• Silence: no source adequately addresses a critical question.

Discrepancy is not automatically a problem. Customers may express strong interest but fail to adopt because onboarding, pricing, procurement, or trust creates friction. The discrepancy identifies the next investigation.

A synthesis table should retain source-level detail rather than hide it:

2.3 Avoiding false precision

A score can support discussion, but it cannot eliminate judgement. Combining unlike evidence—such as a usability observation and a revenue forecast—into a single average may create a misleading appearance of scientific objectivity. A scorecard is best treated as a structured argument, not a mechanical substitute for professional reasoning.

3. Scorecards for Structured Judgement

A product scorecard makes criteria, weights, ratings, and evidence visible. A simple weighted score is:

where S is the total score, wᵢ is the weight assigned to criterion i, rᵢ is the rating for criterion i, and n is the number of criteria. If weights are expressed as proportions, then ∑wᵢ = 1. Ratings must use a defined scale, such as 1 to 5. The score is meaningful only when criteria are sufficiently distinct, ratings are supported by evidence, and the weighting reflects the decision context.

Typical criteria include:

• customer problem severity;
• strategic fit;
• expected customer value;
• commercial potential;
• technical feasibility;
• operational readiness;
• regulatory and ethical acceptability; and
• learning value relative to cost and time.

Worked example: a discovery scorecard

Suppose a team assesses a proposed workflow product using five criteria. Ratings use 1–5, and weights sum to 1.

The total is S = 3.70. If the organisation has defined 3.5 as a threshold for a controlled pilot, the proposal qualifies for a pilot. It does not qualify automatically for full-scale development because commercial evidence is weak and feasibility is only moderate.

The correct managerial interpretation is therefore conditional: “Proceed to a bounded learning investment,” not “The product is validated.”

Scorecard governance

A credible scorecard should record:

• the decision question and date;
• definitions for each criterion and rating level;
• evidence supporting every rating;
• the person responsible for the rating;
• confidence and unresolved risks;
• the funding or capacity requested; and
• the next review condition.

Weights should not be changed merely to make a favoured initiative pass. If the context changes, the team should version the scorecard and explain the change.

4. Confidence Levels and Uncertainty

Confidence expresses how strongly the available evidence supports a claim under stated conditions. It is not the same as probability of success unless a defensible probabilistic model exists.

A practical confidence scale may be:

Confidence should be attached to a specific claim, not to an entire initiative. A team may have high confidence that a problem exists, moderate confidence that users can complete a proposed workflow, and low confidence that the market will pay enough to support the business model.

4.1 Confidence is not sample size alone

Larger samples can reduce random error, but they do not correct a biased sample, a flawed measure, or an irrelevant population. A survey of thousands of non-buyers may be less useful for a business-to-business pricing decision than a small number of interviews with budget owners, provided the limitations are acknowledged.

When quantitative estimates are used, teams should distinguish:

• measurement uncertainty: limitations in how a variable was measured;
• sampling uncertainty: variation arising because only part of a population was observed;
• model uncertainty: uncertainty about whether the analytical model is appropriate; and
• implementation uncertainty: uncertainty about whether a solution can be delivered and adopted.

4.2 Confidence statements

A useful confidence statement has the form:

We are [confidence level] that [claim] for [population or context], because [evidence]. Confidence is limited by [specific uncertainty]. The next test is [action], and the decision will be revisited when [criterion].

This format prevents vague declarations such as “The data is positive.”

5. Decision Options: Persevere, Pivot, Pause, and Stop

These four options describe different responses to evidence. They should not be treated as moral judgements about the team.

Persevere

Persevere when the core problem and strategic rationale remain credible, evidence is improving, and the next investment is proportionate to the remaining risk. Perseverance may involve scaling, refining execution, or running the next planned experiment.

Pivot

Pivot when a material assumption is weakened but an adjacent opportunity remains attractive. A pivot changes a meaningful element of the product logic, such as the segment, problem framing, channel, business model, or solution mechanism. It should not be used as a rhetorical way to avoid acknowledging failure.

Pause

Pause when the opportunity may remain valuable but the organisation lacks a necessary dependency, decision, capability, or external condition. A pause must include a trigger and review date; otherwise it becomes unexamined continuation.

Stop

Stop when evidence indicates that the opportunity no longer justifies further investment, when unacceptable risk cannot be mitigated, or when strategic priorities have changed. Stopping an initiative is not equivalent to declaring all learning worthless. The team should preserve reusable insights, assets, and records.

Decision matrix

The quality of a recommendation depends on the alternative considered. “Continue” is not a neutral choice: it consumes funds, capacity, and opportunity cost.

6. Decision Logs and Organisational Learning

A decision log is a durable record of what was decided, why, by whom, under which assumptions, and when the decision should be revisited. It differs from meeting minutes because it captures reasoning and consequences rather than conversation chronology.

A practical decision-log entry includes:

A decision log should be written close to the decision date. Retrospective reconstruction is vulnerable to hindsight bias: once outcomes are known, people tend to remember the original evidence as more decisive than it was.

Decision logs also support “decision quality” reviews. A poor outcome does not necessarily mean a poor decision; a sound decision can produce a bad result because of random events. Conversely, a favourable outcome may result from luck despite weak reasoning. Reviewing the evidence and reasoning separately from the outcome encourages learning rather than blame.

7. Stakeholder Alignment and Product Governance

Stakeholder alignment is not universal agreement. It is shared understanding of the decision, the evidence, the unresolved disagreement, and the authority to act. Product managers should make disagreement visible rather than seeking superficial consensus.

7.1 Stakeholder mapping

Relevant stakeholders may include customers, users, product and engineering teams, sales, marketing, operations, finance, legal, security, accessibility specialists, executives, and partners. Their interests differ:

• customers care about outcomes, trust, and switching costs;
• engineering evaluates feasibility, reliability, architecture, and maintainability;
• finance evaluates economic logic and resource trade-offs;
• legal and security assess exposure and controls;
• executives evaluate strategy, timing, and portfolio consequences.

A strong alignment session separates three questions:

1. What do we know?
2. What do we believe but still need to test?
3. What are we choosing to do, given the current uncertainty?

7.2 Disagreement protocols

When stakeholders disagree, the product leader should:

• restate the decision question;
• identify whether the disagreement concerns facts, interpretation, values, or risk tolerance;
• ask what evidence would change each person’s view;
• distinguish reversible from irreversible commitments;
• assign owners to close critical gaps; and
• record the final decision and dissenting concerns.

Escalation should be based on decision rights and materiality, not on seniority alone. A senior sponsor may decide the investment level, but cannot erase evidence of a security or accessibility risk.

7.3 Governance layers

Effective discovery governance operates at three levels:

• Team level: hypotheses, experiments, evidence quality, and weekly or fortnightly decisions;
• Product level: opportunity sequencing, cross-functional dependencies, release readiness, and risk review;
• Portfolio level: strategic fit, funding allocation, capacity balancing, and termination or continuation decisions.

Governance should be lightweight for low-risk discovery and more formal for initiatives involving sensitive data, regulated claims, vulnerable users, safety, or substantial irreversible expenditure.

8. Portfolio Funding and Resource Allocation

A portfolio view recognises that products compete for scarce engineering capacity, research access, design capability, capital, and executive attention. Funding should therefore be staged rather than treated as a single approval event.

8.1 Stage-gated funding

A staged model may include:

1. Explore: small budget to understand the problem and target segment;
2. Validate: budget for prototypes, experiments, technical spikes, and commercial tests;
3. Pilot: controlled delivery with explicit success and safety criteria;
4. Scale: broader investment after critical risks are addressed; and
5. Renew, pivot, pause, or stop: periodic portfolio review.

Each gate should specify the evidence required for the next commitment. The purpose is not bureaucratic approval but protection against escalating investment merely because money has already been spent.

8.2 Real-options reasoning

Early discovery investment can be viewed as purchasing the option to learn more before making a larger commitment. This perspective favours small, informative, reversible tests when uncertainty is high. It also highlights the value of stopping early: a small failed experiment may prevent a much larger loss.

However, option logic should not justify endless experimentation. A test has value only if its result can change an action. Teams should define the decision threshold before collecting evidence whenever practical.

8.3 Portfolio balance

A healthy portfolio may contain:

• core improvements with comparatively predictable returns;
• adjacent opportunities that extend existing capabilities; and
• transformational or exploratory bets with high uncertainty.

The appropriate balance depends on strategy, cash position, competitive conditions, and organisational capability. A numerical portfolio score should not conceal correlated risk. Five initiatives may all depend on the same platform, regulation, supplier, or market assumption.

Useful portfolio indicators include:

• funding by discovery stage;
• percentage of initiatives with explicit hypotheses and decision dates;
• time from evidence collection to decision;
• proportion of initiatives stopped or pivoted for evidence-based reasons;
• concentration of technical, commercial, or regulatory risk; and
• learning achieved per unit of time or funding.

These are governance indicators, not universal performance targets. Optimising the number of experiments can encourage low-value activity and gaming.

9. Ethical, Regulatory, and Responsible Evidence Governance

Evidence-based decisions can still cause harm. Governance must therefore address not only whether a proposition is likely to succeed, but also whether it should be pursued and under what safeguards.

Important questions include:

• Was consent obtained for research and data use?
• Are participants represented fairly, including people with disabilities and vulnerable groups?
• Could the product discriminate through data, design, or automated decisions?
• Is the evidence being used outside the context for which it was collected?
• Are privacy, security, retention, and data-minimisation requirements satisfied?
• Can customers understand and contest consequential automated outcomes?
• Are incentives causing researchers or teams to suppress negative findings?

In regulated or high-impact domains, legal review should not be postponed until launch. Product teams should maintain an evidence chain showing data provenance, analysis decisions, model limitations, approvals, and corrective actions.

AI-assisted research and analytics introduce additional risks. Generative systems may fabricate summaries, omit minority views, reproduce bias, or expose confidential material. AI can support transcription, clustering, coding, and scenario generation, but human review is required for interpretation and consequential decisions. Teams should label synthetic or AI-generated evidence clearly and never treat model output as direct customer evidence.

10. Contemporary Developments

Several developments are changing discovery governance.

Continuous discovery systems connect research repositories, product analytics, support data, and experiment results. Their value depends on data quality, consistent definitions, and access controls; a repository does not automatically create insight.

Causal experimentation is increasingly used to estimate whether a change caused an outcome rather than merely correlated with it. Randomised experiments can support stronger causal claims when assignment, exposure, measurement, and interference assumptions are acceptable. Observational data remains useful but requires more cautious interpretation.

Product operating models increasingly shift from project delivery toward persistent teams accountable for outcomes. This increases the importance of decision logs and outcome measures because funding is renewed according to evidence rather than completion of a predetermined feature list.

Responsible innovation expands governance beyond compliance. Teams are expected to consider social effects, environmental cost, accessibility, worker impact, and long-term trust during discovery rather than after public release.

11. Applied Case: The Atlas Compliance Workspace

The following is a fictional illustrative case.

Background

Atlas, a mid-sized software company, considered building a workspace that would help regulated businesses prepare compliance evidence. Sales teams reported frequent requests, while engineering leaders warned that integrations with customer systems could be expensive.

Problem

The executive committee had to decide whether to fund a six-month build. Stakeholders held conflicting views: sales considered the opportunity strategically important; engineering questioned the integration model; legal warned that inaccurate compliance claims could create liability.

Analysis

The discovery team created three claims:

1. Compliance managers experience a frequent, costly evidence-collection problem.
2. They would adopt a central workspace if it reduced preparation time without weakening control.
3. Atlas could deliver a secure, auditable product within the proposed investment.

Evidence included 22 interviews, analysis of support tickets, observation of four preparation cycles, a clickable prototype, a technical integration spike, and five paid design-partner commitments. The evidence showed strong problem frequency and credible interest. The prototype reduced task time in observed sessions, but the sample was small. The integration spike revealed that one major data source lacked a stable interface.

The team’s scorecard was:

The score exceeded the pilot threshold but not the organisation’s scale threshold of 4.2. More importantly, the low feasibility rating represented a concentrated risk rather than an ordinary weakness.

Approach and decision

The team recommended pause with a bounded validation investment, not full development and not immediate termination. The next investment would:

• secure a data-access agreement with the major source system;
• run a controlled pilot with three design partners;
• test auditability and permission controls with legal and security review;
• measure preparation time, error rate, adoption, and willingness to renew; and
• return to the portfolio committee after eight weeks.

The decision log recorded that the initiative would pivot toward an evidence-collection workflow if deep integration remained unavailable. It also recorded a stop condition: no production launch without validated audit trails and appropriate customer controls.

Outcome

One integration agreement was secured, but the second remained unavailable. The pilot demonstrated strong value when customers uploaded structured evidence manually, but the proposed automated integration could not meet the security standard within the funding envelope. Atlas stopped the automated-integration version and launched a narrower document-control product using a safer architecture.

Learning points

• Strong customer evidence does not cancel a critical feasibility or compliance risk.
• A pause is useful only when it has a defined test, owner, and date.
• A pivot can preserve validated value while changing the delivery mechanism.
• Stopping one version does not mean discarding all learning.
• The decision log preserved why the original build was not approved and prevented later claims that the failure had been unforeseeable.

12. Best-Practice Checklist

Before making a significant discovery decision, ask:

• Is the decision question explicit?
• Are evidence and interpretation separated?
• Does each source have a clear scope and limitation?
• Have contradictory findings been investigated rather than averaged away?
• Are confidence statements claim-specific?
• Are scorecard weights and rating definitions visible?
• Is the recommended investment reversible and proportionate to uncertainty?
• Are stop, pivot, pause, or persevere conditions explicit?
• Are dissenting stakeholder views recorded?
• Are legal, privacy, security, accessibility, and ethical risks addressed?
• Is there a named owner, review date, and decision trigger?
• Will the decision log support later learning?

Challenges and Limitations

Evidence-based governance has several limitations. Evidence may be delayed, expensive, inaccessible, or strategically manipulated. Quantitative measures can create false confidence, while qualitative findings can be dismissed unfairly as “anecdotal.” Scorecards can institutionalise existing biases through their criteria and weights. Governance can also become a bottleneck when every experiment requires excessive approval.

Organisations must therefore balance consistency with judgement. Templates should establish minimum quality, not eliminate contextual expertise. Leaders should reward accurate learning and responsible stopping, not only successful launches. Finally, evidence cannot resolve value conflicts by itself. Decisions about fairness, acceptable risk, and strategic purpose require explicit judgement and accountable leadership.

Chapter Summary

Evidence-based product decisions convert mixed discovery findings into explicit, accountable action. High-quality synthesis distinguishes evidence from assumptions and interpretations, assesses relevance and limitations, and uses triangulation to identify convergence and discrepancy. Scorecards can structure judgement through weighted criteria, but they must not create false precision or replace reasoning.

Confidence should be attached to specific claims and expressed with clear limitations. Persevere, pivot, pause, and stop are legitimate responses to different evidence patterns; each requires explicit conditions and follow-up. Decision logs preserve reasoning, uncertainty, alternatives, ownership, and outcomes, enabling organisational learning and reducing hindsight bias.

Discovery governance connects team-level experimentation with product-level prioritisation and portfolio-level funding. Staged investment, real-options reasoning, and portfolio risk analysis help organisations commit progressively while protecting scarce resources. Ethical, legal, privacy, accessibility, security, and responsible-AI considerations must be integrated into evidence governance, especially where products affect vulnerable people or consequential decisions. The ultimate purpose of governance is not bureaucracy but better judgement under uncertainty.

Key Terms

Evidence: An observation, measurement, record, or credible finding used to support a claim.

Assumption: A proposition treated as true but not yet adequately tested.

Triangulation: Comparison of different evidence sources or methods to assess convergence and discrepancy.

Validity: The extent to which evidence or a method supports the claim it is intended to address.

Reliability: The degree to which a method would produce reasonably consistent results under similar conditions.

Confidence level: A qualitative or quantitative expression of how strongly evidence supports a specified claim.

Scorecard: A structured assessment using defined criteria, ratings, and sometimes weights.

Weighted score: A combined score calculated by multiplying each criterion rating by its weight and summing the contributions.

Persevere: Continue investment because the core rationale and evidence remain sufficiently credible.

Pivot: Change a material element of the product opportunity while preserving a credible source of value.

Pause: Temporarily withhold major investment while a specified dependency or uncertainty is resolved.

Stop: End further investment because expected value, safety, feasibility, or strategic fit is inadequate.

Decision log: A durable record of a decision, its reasoning, evidence, uncertainty, alternatives, and follow-up.

Decision rights: Defined authority to make or approve a particular decision.

Stage-gated funding: Progressive investment released when specified evidence or conditions are met.

Portfolio governance: Coordination of investment, capacity, risk, and strategic fit across initiatives.

Opportunity cost: The value forgone by allocating resources to one option rather than another.

Responsible innovation: Development that considers ethical, social, environmental, accessibility, and long-term trust implications.

Review Questions

5 Short Answer Questions

1. Distinguish evidence from an interpretation and provide one product example of each.
2. What are the five dimensions commonly used to assess evidence quality?
3. Why should confidence be attached to a claim rather than to an entire initiative?
4. What information should a decision log contain?
5. How does a pause differ from a stop decision?

5 Long Answer Questions

1. Explain how triangulation can reveal both convergence and discrepancy in mixed discovery evidence.
2. Describe how to construct and govern a weighted product discovery scorecard. Discuss at least three limitations.
3. Compare persevere, pivot, pause, and stop recommendations, including the evidence patterns that support each.
4. Explain how decision logs contribute to stakeholder alignment and organisational learning.
5. Discuss how staged funding and portfolio governance can reduce product investment risk without eliminating innovation.

3 Analytical / Application Questions

1. A proposed product scores highly on problem severity and strategic fit but poorly on feasibility and privacy readiness. Recommend a course of action and specify the next evidence required.
2. Interviewees report strong interest in a feature, but usage data shows little engagement with the related workflow. Analyse at least four possible explanations and design a discriminating test.
3. Design a decision-log entry for an initiative that should pivot. Include the decision statement, evidence, confidence, alternatives, risks, owner, review date, and stop condition.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes evidence from interpretation?

A. Evidence is always quantitative, while interpretation is always qualitative.

B. Evidence is an observation or record, while interpretation is an explanation of what it may mean.

C. Evidence is a decision, while interpretation is a funding request.

D. Evidence is a stakeholder opinion, while interpretation is a legal requirement.

Correct Answer: B. Evidence is an observation or record, while interpretation is an explanation of what it may mean.

Question 2

What is the principal purpose of triangulation in product discovery?

A. To replace qualitative research with analytics

B. To ensure every evidence source receives the same weight

C. To compare sources and identify convergence, complementarity, or discrepancy

D. To eliminate the need for professional judgement

Correct Answer: C. To compare sources and identify convergence, complementarity, or discrepancy

Question 3

A scorecard has ratings of 4, 5, 4, 3, and 2 with weights 0.25, 0.20, 0.20, 0.20, and 0.15 respectively. What is the weighted score?

A. 3.20

B. 3.50

C. 3.70

D. 4.00

Correct Answer: C. 3.70

Question 4

Which situation most strongly supports a pivot recommendation?

A. The core problem is credible and execution is improving.

B. A dependency is temporarily unavailable but has a defined resolution date.

C. A critical assumption is weakened, but an adjacent customer problem remains attractive.

D. The initiative has unacceptable risk with no credible mitigation.

Correct Answer: C. A critical assumption is weakened, but an adjacent customer problem remains attractive.

Question 5

Which statement about confidence levels is most accurate?

A. Confidence is determined solely by sample size.

B. Confidence should be attached to a specific claim and include its limitations.

C. High confidence means success is guaranteed.

D. Confidence removes the need for additional testing.

Correct Answer: B. Confidence should be attached to a specific claim and include its limitations.

Question 6

What makes a decision log different from ordinary meeting minutes?

A. It records only attendance and discussion time.

B. It records reasoning, uncertainty, alternatives, ownership, and follow-up.

C. It excludes dissenting opinions to preserve alignment.

D. It is created only after the outcome is known.

Correct Answer: B. It records reasoning, uncertainty, alternatives, ownership, and follow-up.

Question 7

What is the primary purpose of stage-gated funding?

A. To approve the entire product budget at the beginning

B. To release progressive investment when specified evidence or conditions are met

C. To prevent teams from conducting experiments

D. To rank employees by delivery speed

Correct Answer: B. To release progressive investment when specified evidence or conditions are met

Question 8

Which is an example of a portfolio-level governance concern?

A. Whether a prototype button is understandable

B. Whether one interview question is leading

C. Whether several initiatives depend on the same constrained platform

D. Whether a researcher records an interview accurately

Correct Answer: C. Whether several initiatives depend on the same constrained platform

Question 9

Why is stopping an initiative not necessarily equivalent to losing all value?

A. Stopped initiatives always become successful later.

B. Learning, reusable assets, and decision records may remain valuable.

C. Stopping removes the need to document outcomes.

D. A stopped initiative does not consume any resources.

Correct Answer: B. Learning, reusable assets, and decision records may remain valuable.

Question 10

Which practice is most appropriate when using AI to support discovery analysis?

A. Treat AI-generated summaries as direct customer evidence.

B. Upload confidential research without access controls.

C. Use AI for support tasks while requiring human review of interpretation and consequential decisions.

D. Replace participant consent with automated transcription.

Correct Answer: C. Use AI for support tasks while requiring human review of interpretation and consequential decisions.