Scaling Lean Innovation and Emerging Directions

Introduction and context

Lean Startup practices are most powerful when they are treated not as a collection of workshop techniques, but as an organizational system for reducing uncertainty. Customer interviews, minimum viable products (MVPs), innovation accounting and pivot decisions are relatively easy to apply in a small venture. They become substantially more difficult when innovation is embedded in a corporation, a social venture, a public institution or a regulated industry.

Scaling creates a paradox. An organization must preserve disciplined learning while introducing more people, more capital, more customers, more dependencies and more formal controls. A process that is appropriate for discovering whether a problem exists may be inappropriate for scaling a validated solution. Conversely, premature standardization can suppress learning before the business model is sufficiently understood.

This chapter examines how Lean Startup principles change across organizational settings and growth stages. It develops an integrated view of innovation portfolios, organizational learning, continuous and remote discovery, generative artificial intelligence (AI), governance and experimentation beyond the traditional MVP.

Learning objectives

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

1. Adapt Lean Startup methods for corporations, social ventures and regulated sectors.
2. Distinguish discovery, validation, growth and scaling transitions.
3. Design an innovation portfolio and allocate resources under uncertainty.
4. Use evidence, learning metrics and decision rules without confusing activity with progress.
5. Evaluate remote research, continuous discovery and generative AI practices.
6. Identify ethical, legal, privacy, safety and governance requirements for experimentation.
7. Critically assess emerging models of experimentation and organizational learning.

1. From startup experimentation to scaling systems

A startup usually has a small team, a short communication chain and a limited number of assumptions. A large organization has multiple business units, risk owners, procurement processes, legacy systems and competing incentives. The core Lean logic remains useful, but the unit of analysis changes from an individual product team to a system of teams and decision makers.

The central learning loop is:

Hypothesis → experiment → evidence → interpretation → decision → revised hypothesis

The loop is not identical to simply building, measuring and learning. Measurement is valuable only when the organization has identified a decision that the evidence can inform. A dashboard with many indicators may create the appearance of rigor while leaving strategic uncertainty unresolved.

1.1 Four transitions in innovation

Innovation normally moves through four overlapping transitions:

The methods and thresholds should change at each transition. A founder may accept a manually delivered service during discovery. A scaled operation cannot rely indefinitely on hidden manual work, unverified data or an individual employee’s tacit knowledge.

A useful distinction is between innovation risk and execution risk. Innovation risk concerns whether a desirable, feasible and viable opportunity exists. Execution risk concerns whether a known model can be delivered consistently, efficiently and safely. Scaling requires a deliberate handover from learning primarily about the model to learning about the operating system.

1.2 The scaling trap

Organizations commonly commit one of two errors:

• Premature scaling: investing in infrastructure, sales or staffing before key assumptions are validated.
• Permanent experimentation: continuing small pilots without making a clear commitment to standardization, investment or termination.

The remedy is a transition charter. Before scaling, the team should specify the validated proposition, the remaining uncertainties, minimum operating capabilities, decision rights, risk controls and criteria for increasing investment. Scaling is therefore not the end of experimentation; it is a change in the object and governance of experimentation.

2. Corporate entrepreneurship and ambidextrous organization

Corporate entrepreneurship refers to entrepreneurial activity within an established organization. It includes new businesses, internal ventures, business-model renewal, strategic innovation and entrepreneurial renewal of existing operations.

Large organizations possess assets that startups often lack: distribution, brand, data, technical infrastructure, regulatory knowledge and capital. They also possess liabilities: bureaucracy, legacy incentives, risk aversion and decision processes designed for predictable operations. A corporate innovation team must therefore be protected from inappropriate performance expectations while remaining connected to assets and expertise that enable scale.

2.1 Exploration and exploitation

The distinction between exploration and exploitation is central. Exploration searches for new opportunities through experimentation and variation. Exploitation improves efficiency, reliability and returns from established activities. An organization that only exploits becomes vulnerable to disruption; one that only explores fails to convert learning into sustainable value.

An ambidextrous organization maintains both activities, either in structurally separate units or through carefully designed contextual arrangements. Separation can protect exploratory work from short-term operating targets. Integration can provide access to customers, capabilities and channels. Neither arrangement is universally superior.

Effective corporate Lean practice commonly requires:

• a separate evidence-based funding process for uncertain opportunities;
• senior sponsorship that removes organizational barriers rather than dictating solutions;
• access to real customers and operational data;
• explicit interfaces between venture teams and core functions;
• a transition path from venture governance to business-unit governance; and
• incentives that reward validated learning, not merely launch activity.

2.2 Corporate constraints and adaptation

A corporation cannot always run an unconstrained experiment on its customers. Brand risk, cybersecurity, employment obligations and contractual commitments may restrict randomization or public release. The team may use staged exposure, simulated environments, internal users, opt-in pilots, synthetic data or controlled feature flags instead.

Corporate entrepreneurship also requires political and relational work. A product hypothesis may be technically sound but fail because a business unit fears cannibalization, a sales force is not incentivized to sell it, or a data owner refuses access. The Lean canvas must therefore be complemented by a stakeholder and capability map identifying sponsors, gatekeepers, affected groups, dependencies and sources of resistance.

3. Social ventures: multiple forms of value

Social ventures pursue social or environmental outcomes through entrepreneurial means. Their value proposition is often directed toward a beneficiary, while revenue may come from a payer, donor, government agency or institutional partner. The user, beneficiary, customer and payer may consequently be different people.

This creates a multi-sided discovery problem. A service can be highly valued by beneficiaries but financially unsustainable for the payer. Alternatively, a financially attractive intervention can produce unintended harm or exclude the people it intends to serve.

3.1 A social venture theory of change

A theory of change connects activities to outputs, outcomes and longer-term impact:

Inputs → Activities → Outputs → Outcomes → Impact

The relationships are hypotheses, not guarantees. For example, providing digital financial education may produce course completion (an output), but completion does not necessarily produce improved financial resilience (an outcome). Impact evaluation must therefore distinguish participation from meaningful change.

Social ventures should test at least four dimensions:

1. Desirability: Do beneficiaries recognize and value the intervention?
2. Effectiveness: Does it improve the targeted outcome relative to a credible comparison?
3. Viability: Can the organization sustain delivery financially and operationally?
4. Equity and legitimacy: Who benefits, who bears costs and who is excluded?

A useful impact indicator should define the population, outcome, time frame and measurement method. Where appropriate, an impact rate may be expressed as:

This ratio is meaningful only when the numerator and denominator use the same eligibility rules, observation period and outcome definition. It does not by itself prove causality; comparison groups or other evaluation designs may be needed.

3.2 Ethical tensions in social experimentation

The language of experimentation can be uncomfortable when people face poverty, illness or displacement. Consent must be meaningful, participation must not be coercive and control groups must not be denied essential services. The organization should assess distributional effects rather than relying on an average treatment result. An intervention that improves the average outcome while worsening outcomes for a vulnerable subgroup may be unacceptable.

4. Lean innovation in regulated sectors

Regulated sectors include healthcare, financial services, aviation, energy, pharmaceuticals, education and critical infrastructure. Regulation is not simply an obstacle to speed. It often encodes public safety, rights, market integrity and accountability. The objective is not to bypass regulation but to design learning activities that are lawful, proportionate and auditable.

4.1 Discovery under regulatory constraints

Early discovery can usually focus on needs, workflows, failure points and non-sensitive problems without deploying an unapproved product. Later experiments require classification of the product, intended use, risk level, data types and responsible authority.

A regulated experiment should specify:

• the applicable laws, standards, licenses and professional duties;
• whether the experiment affects safety, treatment, financial decisions or essential access;
• data ownership, consent, retention and security controls;
• adverse-event or incident reporting procedures;
• qualified review and approval responsibilities;
• rollback and business-continuity arrangements; and
• an evidence trail sufficient for audit.

The minimum viable product in a regulated setting is often better understood as a minimum viable compliant system. It may contain more documentation, testing and human oversight than a consumer software MVP. Minimum does not mean careless; it means the smallest system that can responsibly test the relevant hypothesis.

4.2 Sandboxes and staged exposure

Regulatory sandboxes can allow limited testing under defined conditions. They are useful when the regulator and innovator agree on scope, safeguards, reporting and exit criteria. A sandbox is not a waiver of all obligations, and success in a constrained environment does not automatically establish readiness for general deployment.

Staged exposure can follow this sequence:

1. Concept and workflow research.
2. Non-production prototype.
3. Internal or professional-user testing.
4. Small, monitored pilot with informed participants.
5. Controlled expansion with predefined stop conditions.
6. Full deployment subject to continuing monitoring.

5. Managing an innovation portfolio

An innovation portfolio contains initiatives at different stages and levels of uncertainty. Managing it requires more than ranking projects by projected return. Early forecasts are fragile, and apparently precise financial estimates can conceal major assumptions.

5.1 Portfolio categories

A practical portfolio may classify initiatives as:

• Core: improvements to existing customers, products and processes;
• Adjacent: extensions into related markets, channels or capabilities; and
• Transformational: new business models, technologies or customer systems.

The categories are not a universal allocation rule. Their purpose is to make risk, time horizon and capability requirements visible. Portfolio review should ask whether the organization has enough experiments in the future-oriented categories, not merely whether every current project has a positive forecast.

5.2 Options logic and staged funding

An uncertain innovation resembles a real option: a small investment can purchase information and preserve the choice to invest more later. Funding should therefore be released in tranches tied to evidence and capability milestones rather than to elapsed time alone.

A simple expected monetary value model is:

where p_i is the probability assigned to outcome i, V_i is the value of that outcome, and C is the cost of the initiative. The probabilities and values must use the same time horizon and monetary units. EMV is a decision aid, not a forecast of certainty; it is highly sensitive to assumptions and may omit strategic, social or regulatory effects.

For a staged initiative, management can compare the value of continuing with the cost of the next learning step. A decision rule might be: continue when the expected value of information and the plausible future value exceed the next-stage cost and risk; pivot when evidence invalidates a central assumption but suggests a credible alternative; stop when the opportunity is unattractive or the risk cannot be responsibly controlled.

5.3 Innovation accounting

Innovation accounting connects learning to decisions through three layers:

1. Leading indicators: activation, task completion, qualified interviews, prototype use or pilot adherence.
2. Behavioral indicators: retention, repeat use, conversion, referral or outcome improvement.
3. Business and impact indicators: contribution margin, cash generation, risk reduction, beneficiary outcomes or environmental effect.

Metrics must be interpreted in context. High registration with low repeated use may indicate weak value. High engagement may represent harmful dependence rather than success. A metric becomes useful when its definition, population, time window, data source and decision consequence are explicit.

6. Organizational learning and the scaling transition

Organizational learning occurs when experience changes the organization’s knowledge, routines, assumptions or capabilities. Individual insight is not enough. If learning remains in one employee’s memory, the organization has not developed a reliable capability.

6.1 Single-loop and double-loop learning

Single-loop learning corrects performance within existing goals and assumptions—for example, improving response time while retaining the same service model. Double-loop learning questions the goals, governing assumptions or problem framing—for example, asking whether response time matters more than preventing the underlying customer problem.

Scaling can unintentionally reduce double-loop learning. Standard operating procedures improve consistency but may make anomalies invisible. Mature organizations should preserve channels for frontline feedback, customer complaints, post-incident analysis and periodic assumption reviews.

6.2 Knowledge codification

Before scaling, teams should convert tacit learning into explicit artifacts:

• decision records showing why major choices were made;
• experiment registers linking hypotheses to evidence;
• service blueprints showing customer and backstage processes;
• data dictionaries and metric definitions;
• risk registers and control ownership;
• reusable research repositories; and
• playbooks identifying what must be standardized and what may remain adaptive.

Codification should not become bureaucratic accumulation. A document is valuable when it improves future decisions, onboarding, quality or accountability.

7. Continuous discovery and remote research

Continuous discovery is a recurring practice of engaging customers, observing use and testing assumptions throughout product development rather than conducting research only at project initiation. It is especially useful when customer needs, competitors, technologies or policies change quickly.

A robust continuous-discovery cadence combines:

• regular conversations with current, potential and lost customers;
• observation of real tasks and service contexts;
• analysis of behavioral and support data;
• prototype or concept tests;
• cross-functional synthesis; and
• explicit decisions about what changes as a result.

Remote research increases reach and speed through video interviews, asynchronous diaries, online communities, remote usability tests and digital analytics. However, it can exclude people with poor connectivity, limited digital literacy, disabilities or privacy concerns. Researchers should offer accessible formats, verify identity where necessary, protect recordings and avoid treating online convenience as representative sampling.

Remote evidence also has methodological limitations. Participants may perform differently in an observed video call than in a natural setting. Digital trace data show behavior but often do not explain motivation. Self-selection can overrepresent highly engaged users. Triangulation—combining interviews, observation, behavioral data and operational evidence—improves interpretation.

8. Generative AI in discovery and experimentation

Generative AI can support transcription, coding of qualitative data, synthesis of themes, creation of alternative concepts, simulation of interview prompts, generation of test variants and preparation of research summaries. It can reduce clerical effort, but it does not remove the need for human judgment or direct contact with customers.

8.1 Appropriate uses

Appropriate uses include:

• drafting interview guides for expert review;
• identifying possible themes in a large corpus before human verification;
• generating contrasting hypotheses;
• translating or simplifying research materials with quality checks;
• producing synthetic scenarios for early usability rehearsal; and
• monitoring experiment logs for missing fields or inconsistent terminology.

8.2 Risks and controls

Generative AI may hallucinate facts, reproduce stereotypes, expose confidential information, collapse minority perspectives into majority themes or generate persuasive but invalid explanations. Synthetic users are not substitutes for real users because a model’s output reflects its training and prompt rather than lived experience.

Governance should include:

• data classification before model use;
• prohibition or restriction of confidential and personal data in unapproved systems;
• disclosure when participants interact with AI;
• human verification of summaries and decisions;
• provenance and version records for prompts, models and outputs;
• bias and subgroup checks; and
• an appeal or correction process when AI-supported decisions affect people.

A useful principle is AI for augmentation, not epistemic substitution: AI can accelerate analysis, but responsibility for interpretation, consent, fairness and decisions remains with accountable people.

9. Emerging models of experimentation

The classical MVP is only one form of experiment. Emerging practice uses a broader minimum viable evidence system, selected according to the uncertainty being investigated.

New experimentation models include:

• platform experimentation: reusable feature flags, identity systems, analytics and guardrails that allow many teams to test safely;
• ecosystem experimentation: trials involving partners, suppliers, regulators and communities rather than one firm alone;
• frugal and offline experimentation: low-cost methods designed for constrained environments;
• behavioral and adaptive experimentation: designs that change exposure based on incoming evidence while preserving statistical and ethical integrity; and
• mission-oriented experimentation: portfolios organized around societal challenges rather than isolated products.

The more consequential the decision, the stronger the requirements for causal inference, transparency, safety and accountability. Speed is not an independent virtue when an experiment can cause irreversible harm.

10. An integrated scaling playbook

A manager preparing an innovation for scale can use the following sequence:

1. Name the assumption: State the customer, problem, proposed mechanism and expected outcome.
2. Classify the stage: Identify whether the work is discovery, validation, growth or scaling.
3. Map stakeholders and constraints: Include users, payers, beneficiaries, regulators, data owners, employees and partners.
4. Select the smallest responsible experiment: Match the method to the uncertainty and risk.
5. Define evidence in advance: Specify measures, comparison logic, time window and stop conditions.
6. Run with safeguards: Protect privacy, safety, equity, security and informed participation.
7. Interpret disconfirming evidence: Examine whether the problem, segment, solution, channel or operating model is wrong.
8. Decide explicitly: Persevere, pivot, pause, scale or stop.
9. Codify learning: Record assumptions, evidence, decisions and ownership.
10. Increase organizational capacity: Standardize reliable elements while preserving feedback and local adaptation.

Applied case: Meridian HealthConnect

The following is an illustrative case.

Background

Meridian HealthConnect, a fictional regional healthcare provider, developed a digital follow-up service for patients with hypertension. Nurses initially contacted a small group of patients by telephone and text message. The pilot suggested that reminders and medication questions could reduce missed follow-ups.

Problem

Senior management proposed rapid expansion to 50,000 patients. The innovation team argued that the initial evidence was insufficient. The pilot used highly engaged patients, manual nurse work and a single clinic. It had not established whether patients would consent to digital contact, whether the workflow could operate at scale, or whether outcomes improved rather than merely message response rates.

Analysis

The team separated assumptions into four categories:

• Desirability: patients find the service useful and accessible;
• Clinical effectiveness: the service improves appropriate follow-up behavior;
• Operational feasibility: nurses can manage alerts within safe response times;
• Compliance and equity: consent, data protection, clinical accountability and accessibility are adequate.

The team also identified a risk that automated reminders could overwhelm patients or create false reassurance. Therefore, a high response rate was not accepted as the sole success measure.

Approach and solution

Meridian designed a staged, minimum viable compliant system. It used approved communication channels, explicit consent, human review of escalation alerts, accessibility alternatives and an incident process. A second pilot included clinics with different patient demographics and measured response, follow-up completion, nurse workload, adverse events and subgroup differences. The team maintained a decision register and reviewed results with clinical, legal, data-protection and patient representatives.

The organization used a controlled rollout rather than immediate universal deployment. Expansion occurred only when predefined safety and capacity conditions were met. The operating model changed as learning accumulated: routine reminders became more automated, while clinically significant alerts remained under professional review.

Outcome

The pilot did not produce a simple “scale” or “stop” answer. It showed improved follow-up completion among consenting digital users, but lower access among patients with limited connectivity and higher-than-expected nurse workload for a subgroup with complex needs. Meridian therefore expanded the service for suitable patients, funded a telephone alternative, revised triage rules and postponed full automation.

Learning points

1. A regulated MVP must include compliance, safety and accountability.
2. Positive user engagement is not equivalent to clinical effectiveness.
3. Scaling exposes operational and equity assumptions that a small pilot can conceal.
4. A portfolio decision can be selective: scale one segment, redesign another and stop an unsafe feature.
5. Human oversight should be reduced only when evidence supports that transition.

Challenges and limitations

Lean methods have important limitations. Customer statements may not predict behavior, especially when payment or risk is absent. Small samples can reveal problems but rarely establish population-level effects. A/B tests may optimize a local metric while damaging trust, long-term retention or social welfare. Rapid experimentation can privilege users who are digitally visible and exclude marginalized groups.

Organizational incentives may also distort learning. Teams may avoid experiments that could disconfirm executive commitments, redefine metrics after unfavorable results or report activity as progress. Portfolio governance can become a political ranking exercise rather than a disciplined response to uncertainty.

Finally, not all innovation problems are reducible to customer preference. Climate transition, public health, infrastructure and institutional reform involve systems, power relations and collective action. Lean experimentation is useful within these domains, but it must be combined with policy analysis, stakeholder governance, systems thinking and impact evaluation.

Best-practice principles

• Start with the riskiest assumption, not the easiest feature.
• Match evidence strength to potential harm and decision consequence.
• Separate learning metrics from outcome metrics.
• Use staged investment and explicit stop, pivot and scale criteria.
• Treat compliance and ethics as design requirements, not late-stage approvals.
• Include non-users, excluded groups and affected communities in discovery.
• Preserve dissenting evidence and minority experiences.
• Use AI transparently, with human verification and data safeguards.
• Codify learning without allowing documentation to replace learning.
• Scale the operating system only after understanding what must remain adaptive.

Chapter Summary

Scaling Lean innovation requires a transition from testing a product hypothesis to developing a reliable, ethical and economically sustainable system. Corporate entrepreneurship must balance exploration with exploitation and connect venture teams to organizational assets without imposing premature operating targets. Social ventures must test desirability, effectiveness, viability, equity and legitimacy across beneficiaries, users and payers. Regulated sectors require minimum viable compliant systems, staged exposure, auditability and proportionate safeguards.

Innovation portfolios should allocate staged resources under uncertainty, using innovation accounting and decision rules rather than superficial activity measures. Organizational learning depends on converting individual insight into shared routines, records and capabilities while preserving double-loop questioning. Continuous and remote discovery can improve responsiveness but requires attention to representation, accessibility and triangulation. Generative AI can augment research and experimentation, yet it introduces risks involving privacy, bias, hallucination and accountability.

The future of Lean innovation lies in evidence systems that combine experimentation with governance, impact evaluation, safety and systems thinking. The defining managerial capability is not merely launching experiments quickly, but knowing what to learn, how much evidence is sufficient, when to scale, and when responsible innovation requires a pause or refusal.

Key Terms

Ambidextrous organization: An organization capable of pursuing exploration and exploitation simultaneously.

Continuous discovery: Ongoing customer research and testing integrated into product work.

Double-loop learning: Learning that questions underlying goals, assumptions or problem definitions.

Emergent impact: Longer-term social or environmental change produced by an intervention.

Experiment register: A structured record of hypotheses, methods, evidence and decisions.

Exploration: Search for new opportunities under uncertainty.

Exploitation: Improvement and efficient use of established capabilities.

Innovation accounting: Measurement that links learning and behavior to investment decisions.

Innovation portfolio: A collection of initiatives distributed across stages, risks and time horizons.

Minimum viable compliant system: The smallest system capable of testing a hypothesis while meeting relevant legal, safety and ethical requirements.

Minimum viable product: The smallest product or service release that tests a relevant assumption with users.

Organizational learning: Changes in shared knowledge, routines, assumptions or capabilities resulting from experience.

Real option: The value of preserving a future investment choice through an initial learning investment.

Regulatory sandbox: A controlled environment for testing innovation under defined regulatory conditions.

Scaling transition: Movement from validating a model to delivering it reliably and responsibly at greater volume.

Staged funding: Release of resources in tranches linked to evidence or capability milestones.

Theory of change: A model linking inputs and activities to outputs, outcomes and impact.

Triangulation: Use of multiple evidence sources or methods to improve interpretation.

Review Questions

Short Answer Questions

1. Why does the object of learning change when an innovation moves from validation to scaling?
2. Distinguish exploration from exploitation in an ambidextrous organization.
3. Why may the beneficiary, user and payer be different in a social venture?
4. What is meant by a minimum viable compliant system?
5. State two risks and two controls associated with generative AI in customer discovery.

Long Answer Questions

1. Explain how a large corporation can protect exploratory innovation while enabling access to corporate assets and eventual scaling.
2. Discuss how a theory of change improves experimentation in social ventures, including its limitations.
3. Evaluate the role of staged exposure, auditability and human oversight in regulated-sector innovation.
4. Explain how innovation accounting can prevent teams from confusing activity metrics with meaningful progress.
5. Compare the methodological strengths and limitations of continuous remote discovery with in-person research.

Analytical / Application Questions

1. Design a staged experiment for a digital financial service aimed at low-income customers. Identify assumptions, evidence, safeguards, inclusion risks and scale criteria.
2. An AI-supported product shows high engagement but declining retention and unequal outcomes across user groups. Diagnose possible explanations and recommend a portfolio decision.
3. Construct a portfolio governance approach for six initiatives distributed across core, adjacent and transformational categories. Specify funding logic, review criteria and conditions for pivoting or stopping.

Multiple Choice Questions with Answers

Question 1

Which question is most characteristic of the scaling transition?

A. Do customers recognize the problem?

B. Can the system deliver reliably and responsibly at volume?

C. Can a prototype be understood in a usability test?

D. Will a first interview reveal a customer need?

Correct Answer: B. Can the system deliver reliably and responsibly at volume?

Question 2

In a social venture, why must the payer be distinguished from the beneficiary?

A. They are always legally separate entities

B. The person receiving value may not be the person financing the service

C. Beneficiaries cannot participate in research

D. Payers are never affected by outcomes

Correct Answer: B. The person receiving value may not be the person financing the service

Question 3

What is the main purpose of staged funding in an innovation portfolio?

A. To guarantee that every project reaches launch

B. To release resources as evidence and capabilities justify further investment

C. To eliminate the need for experimentation

D. To rank projects solely by initial revenue forecasts

Correct Answer: B. To release resources as evidence and capabilities justify further investment

Question 4

Which statement best describes double-loop learning?

A. Improving performance without questioning assumptions

B. Repeating an experiment with a larger sample

C. Questioning the goals or assumptions underlying current action

D. Documenting a standard operating procedure

Correct Answer: C. Questioning the goals or assumptions underlying current action

Question 5

A regulated-sector MVP should primarily be understood as:

A. The cheapest possible product, regardless of risk

B. A product released without professional review

C. The smallest system that can test a hypothesis responsibly and compliantly

D. A fully automated system with no human oversight

Correct Answer: C. The smallest system that can test a hypothesis responsibly and compliantly

Question 6

Which is a limitation of remote customer research?

A. It cannot produce any behavioral evidence

B. It may exclude people with limited connectivity or digital access

C. It is always slower than in-person research

D. It eliminates participant self-selection

Correct Answer: B. It may exclude people with limited connectivity or digital access

Question 7

In the formula EMV = ∑ p_i V_i − C, what does C represent?

A. The probability of outcome i

B. The value of outcome i

C. The cost of the initiative

D. The number of customers in the sample

Correct Answer: C. The cost of the initiative

Question 8

Which practice is most appropriate when using generative AI to summarize interviews?

A. Accepting the generated themes without review

B. Uploading all confidential transcripts to any public tool

C. Human verification of themes and checks for missing minority perspectives

D. Replacing participant interviews with synthetic users

Correct Answer: C. Human verification of themes and checks for missing minority perspectives

Question 9

Why is high engagement alone an insufficient success measure for a healthcare experiment?

A. Engagement can never be measured

B. Engagement may not demonstrate clinical effectiveness or safety

C. Healthcare users never change behavior

D. Engagement is always a financial metric

Correct Answer: B. Engagement may not demonstrate clinical effectiveness or safety

Question 10

What is the main purpose of triangulation in discovery research?

A. To use multiple evidence sources to improve interpretation

B. To ensure that every participant gives the same answer

C. To replace qualitative research with analytics

D. To avoid defining the research question

Correct Answer: A. To use multiple evidence sources to improve interpretation