Entrepreneurship and innovation · Module 1

Startup strategy, business model validation and product-market fit

How a new venture chooses a strategy when almost everything is uncertain, tests the assumptions that could sink it, and recognizes product-market fit when it arrives.

Beyond the IB course UndergraduateMBA About 25 minutes · 2,801 words · checked 16 November 2026

From the IB course to this module

The IB course covers entrepreneurship and starting up (1.1), market orientation (4.1), marketing planning and the USP (4.2) and market research (4.4). The Lab briefings Finding and testing an opportunity and Business models, the canvas and the lean start-up introduce customer discovery, the Business Model Canvas, the lean start-up and the minimum viable product. This module adds startup strategy as a set of hard-to-reverse choices, validation as a statistical experiment with decision rules, the debate between effectual and planned approaches, product-market fit measured by behaviour, and the evidence on whether these methods work.

Section 01: Conceptual foundation

A startup, in Steve Blank's phrase, is a temporary organization searching for a repeatable and scalable business model. Its central problem is uncertainty, not execution: it does not yet know who its customers are, what they will pay or what it will cost to serve them.

Founders make two kinds of decision. Reversible ones, such as a price test or a landing page, are cheap to try and undo, and should be made quickly. Irreversible ones, such as key hires, partners, the technology platform and the brand, commit the venture to a path and close others. Validation turns beliefs about customers, value and economics into evidence, testing the riskiest and cheapest-to-test assumptions first. Product-market fit is the point at which the evidence shows the market pulling the product: customers keep using it and paying for it without being pushed. Strategy decides which market to test first and how to compete once a model works.

Section 02: Advanced explanation

Strategy when everything is uncertain

Gans, Scott and Stern (2018) identified the founder's paradox: the value of a strategy can be learned only by committing to it, and committing closes other paths. They reduced the choice to two questions. Should the venture collaborate with incumbents or compete with them? And should it control its idea, investing in protection such as patents or an architecture it owns, or execute, moving fast to build customers and brand? The answers give four strategies:

  • an intellectual property strategy (collaborate and control), licensing to or partnering with incumbents;
  • a value chain strategy (collaborate and execute), becoming a specialized supplier inside an existing industry;
  • a disruption strategy (compete and execute), entering where incumbents will not follow quickly;
  • an architectural strategy (compete and control), creating and owning a new value chain or platform.

Each implies different hires, funding and partners, so the choice should be made deliberately rather than by drift. Kerr, Nanda and Rhodes-Kropf (2014) add that entrepreneurship is experimentation whose cost shapes which ideas get tried: software ideas can be tested for little, while a new drug or battery chemistry needs years and millions before the first real signal. Cheap experiments favour learning by doing; expensive ones favour thinking hard about strategy first.

Validation as experiment: hypotheses, evidence and decision rules

Customer development (Blank, 2013) tests the problem before the solution: founders interview potential customers to learn whether the problem is real and urgent, then test whether a proposed solution solves it, before building at scale. The lean start-up (Ries, 2011) frames this as a build-measure-learn loop around a minimum viable product, the least effort that produces real evidence, with each loop ending in a decision to pivot or persevere.

Good validation is designed like an experiment. Rank assumptions by the damage if they are wrong and by how uncertain they are; write each as a falsifiable hypothesis; and set the threshold for success before the test. Then read the result with its uncertainty. If 12 of 200 landing-page visitors pre-order, the rate is 6%, but the 95% interval runs from about 3.5% to 10.2%, so a 5% threshold is neither met nor failed with confidence; a margin of two percentage points would need about 540 visitors. Evidence also varies in quality: what people do (deposits, pre-orders, repeat use) outranks what they say in interviews, and early adopters and friends are poor guides to the mainstream.

Does any of this work? In a randomised trial with Italian start-ups, Camuffo, Cordova, Gambardella and Spina (2020) found that founders trained to frame theories and test hypotheses abandoned weak ideas sooner, pivoted more often and earned more revenue than a control group.

Effectuation, planning and plans as hypotheses

Sarasvathy (2001) studied how expert entrepreneurs decide and found a logic she called effectuation. Rather than setting a goal and selecting the means to reach it (causation), they start from their means (who they are, what they know and whom they know), commit only what they can afford to lose, build partnerships with stakeholders who self-select by committing resources, treat surprises as opportunities and try to control the future rather than predict it. Causation suits markets that can be measured; effectuation suits those that do not yet exist.

The planning school replies that thinking ahead pays. A meta-analysis by Brinckmann, Grichnik and Kapsa (2010) found that business planning is positively related to small-firm performance, but less strongly for new ventures than for established firms, with the relationship varying by context. Writing a plan is not the same as following one.

Discovery-driven planning (McGrath and MacMillan, 1995) reconciles the two. It starts from the profit the venture must earn to be worthwhile, works backwards to the revenues and costs that would require, lists every assumption behind them and sets milestones at which each assumption is tested before more money is committed. The plan becomes a schedule of experiments rather than a forecast to be defended.

Product-market fit: measuring pull rather than push

Product-market fit is easier to recognize than to define, so it is best measured by behaviour. The strongest signal is a retention curve by cohort that flattens above zero: a stable core of customers who keep using and paying. Suppose cohort A retains 60% of customers after one month, 42% after three, 35% after six and 33% after twelve, while cohort B retains 55%, 30%, 15% and 5%. Only A shows fit, a core of about a third who stay; B is leaking towards zero however fast it is acquiring. Other signals are a rising share of customers who arrive by word of mouth, frequent use, willingness to pay at full price and shorter sales cycles. Surveys asking how disappointed users would be without the product are popular with practitioners, but their thresholds are rules of thumb, not established findings.

Three cautions apply. Fit is specific to a segment and a use: a product can fit one niche perfectly and the mainstream not at all. Fit can be imitated with money: discounts and paid acquisition can produce growth without retention, which disappears when the subsidy stops. And fit can be lost when competitors or platforms change. When the evidence says no, founders can pivot, changing segment, problem, solution, channel, revenue model or technology while keeping what they have learned; stopping is also a legitimate result of validation, because it preserves capital that escalating commitment would waste.

Section 03: Models and frameworks

Where each comes from, how to use it, and where it stops being useful.

Blank (2013)

Customer development

Search for a business model before executing one: discover customers and their problems, validate that they will buy, then create demand and build the company, returning to discovery when evidence fails.

How to use it

List the hypotheses on each block of the business model, interview and test with real customers, and move to the next stage only when the evidence meets thresholds set in advance.

Where it fails

Interviews reveal stated rather than revealed preferences, and the method fits markets where customers can be found and asked. For radically new products customers may not recognize the problem yet.

Steve Blank (2013). Why the lean start-up changes everything. Harvard Business Review. Alexander Osterwalder and Yves Pigneur (2010). Business Model Generation. John Wiley and Sons.

Ries (2011)

Build-measure-learn and the minimum viable product

Build the smallest product that produces real evidence, measure customer behaviour against hypotheses, learn, then pivot or persevere. Progress is measured by validated learning, not output.

How to use it

Choose the riskiest assumption, design the cheapest test that would change a decision, and record the result against a threshold set beforehand.

Where it fails

Small, fast tests favour incremental features over bold visions, and some products cannot be tested minimally (medical devices, aircraft, regulated finance). A poor MVP can damage the brand or invite imitation.

Eric Ries (2011). The Lean Startup. Crown Business.

Sarasvathy (2001)

Effectuation

Expert entrepreneurs start from available means, invest only affordable losses, form commitments with partners, exploit contingencies and seek to control rather than predict the future.

How to use it

Use it when the market cannot yet be measured: map your means, define what you can afford to lose, and look for partners who will commit resources early and help shape the opportunity.

Where it fails

It rests largely on think-aloud studies of experts and is hard to test for outcomes. In markets that can be researched, ignoring prediction wastes information.

Saras D. Sarasvathy (2001). Causation and effectuation: toward a theoretical shift from economic inevitability to entrepreneurial contingency. Academy of Management Review.

Gans, Scott and Stern (2018)

The entrepreneurial strategy compass

Two choices, collaborate or compete with incumbents and control or execute, define four strategies: intellectual property, value chain, disruption and architectural.

How to use it

Before committing hires, funding and partners, test which strategies are feasible for the idea, choose one deliberately and design early experiments that would distinguish between them.

Where it fails

Real ventures mix strategies and change them as they grow, and the framework says little about how to execute the one chosen.

Joshua Gans, Erin L. Scott and Scott Stern (2018). Strategy for start-ups. Harvard Business Review.

McGrath and MacMillan (1995)

Discovery-driven planning

Start from the profit a venture must earn, work back to the revenues and costs that implies, list the assumptions and test them at milestones before committing more money.

How to use it

Build a reverse income statement, an assumptions checklist and a milestone plan, then update the plan as each assumption is confirmed or rejected.

Where it fails

It depends on honest assumptions and on milestones that really test them; plans written to win funding often lack both.

Rita Gunther McGrath and Ian C. MacMillan (1995). Discovery-driven planning. Harvard Business Review.

Section 04: Real business applications

Real organizations appear only with dated, cited facts; illustrative examples are labelled.

Real organization, 2017-10

Testing in a regulated market: the FCA's regulatory sandbox

The UK's Financial Conduct Authority opened its regulatory sandbox in June 2016 to let firms test new financial products with real customers, using tools such as restricted authorization and safeguards such as extra capital requirements and an exit plan for every test. Its October 2017 review reported 146 applications across the first two cohorts, 50 accepted and 41 tested or testing. Of the first cohort, 75% completed their tests, around 90% of those were continuing towards a wider launch, at least 40% received investment during or after testing, and around a third used what they learned to pivot their business model significantly before launch. Many tests were designed to assess customer uptake and commercial viability; smaller firms without an existing customer base found it harder to recruit customers to test with. The FCA cautioned that it was too early to draw robust conclusions about the sandbox's overall effect on competition.

Interpretation. Where the law restricts experiments, a sandbox lowers the cost of a live-market test. The pivot rate shows how often real customers overturn founders' assumptions. The success figures are the regulator's own, for firms selected as ready to test, with no comparison group.

Real organization, 2025-09

Mistral AI: an entry strategy built on open models

In a statement of its mission in September 2023, Mistral AI said it would train its own models and release them openly to build "a credible alternative to the emerging AI oligopoly", and released its Mistral 7B model under the Apache 2.0 licence, usable without restriction. It also committed to a commercial offering alongside the open models: optimized proprietary models for deployment on customers' own infrastructure, distributed with weights and code, and hosted services for enterprises. In September 2025 it announced a €1.7 billion Series C round at an €11.7 billion post-money valuation, led by the semiconductor equipment maker ASML as part of a strategic partnership, with existing investors including Andreessen Horowitz, Bpifrance, General Catalyst, Index Ventures, Lightspeed and NVIDIA, and said the round reaffirmed its independence.

Interpretation. In the compass, open release is an execution move: it builds adoption, reputation and talent quickly. Charging for deployment, customization and service is how the company seeks control points to earn from. The 2025 round adds collaboration with an incumbent in an adjacent value chain. The risk is that open weights give away much of the core asset.

Illustrative

A false positive from friends and family

A founder (illustrative) launches a meal-planning app to her own network: 30% of the first 200 sign-ups become paying subscribers. Paid advertising to strangers then converts at 3%, and the plan built on the first figure collapses. The early sample was neither random nor independent of the founder; the threshold should have been set and tested on the market she intends to serve.

Section 05: Case studies

Fictional case, simulated figures

Kinfolk Care: which assumptions survived?

Kinfolk Care (fictional) is building a platform that matches families with vetted carers for elderly relatives at short notice, charging families a £25 booking fee and carers a 15% commission. Before raising money, the two founders ran four tests, each with a threshold agreed in advance. They believe the results justify a seed round.

Kinfolk Care, validation tests and results
Assumption (test) Sample Result Threshold set in advance 95% interval
Families struggle to arrange short-notice care (40 problem interviews)4031 (78%)60%62% to 88%
Families will pay a £25 booking fee (landing page with deposit, 600 visitors)60027 (4.5%)5%3.1% to 6.5%
Families rebook within 30 days (concierge pilot, 45 families)4521 (47%)40%33% to 61%
Carers will accept a 15% commission (survey of 60 carers)6022 (37%)50%26% to 49%

Intervals are Wilson 95% intervals for a proportion. Simulated data.

Questions

  1. For each test, judge whether the assumption is supported, refuted or unresolved, using the interval rather than the point estimate.
  2. Why is the carer result the most important finding for a two-sided platform?
  3. The interviews scored far higher than the deposits. Explain the gap and what it implies about the quality of different kinds of evidence.
  4. Propose the next two experiments, with decision rules, and say whether the founders should pivot their revenue model before raising money.

Section 06: Critical evaluation

What the evidence supports, where the debate is, and what the models leave out.

The case for treating a new venture as a series of experiments has firmer evidence than it once did. The randomised trial by Camuffo and colleagues is unusually strong for management research, and the logic of testing the riskiest assumption first follows from simple decision analysis. The statistics of small samples are not in dispute either: many validation decisions are made on evidence too thin to distinguish a good idea from a mediocre one.

Much else rests on weaker ground. The famous lean start-up stories are selected successes; ventures that tested diligently and failed are rarely written up, so the anecdotes cannot show that the method causes success. Critics argue that rapid testing favours incremental ideas, because bold visions cannot be validated in weeks, and that experimentation is costly or impossible in regulated and capital-intensive sectors, the very problem the FCA's sandbox was set up to ease. Effectuation is grounded mainly in studies of how experts reason, which shows what they do, not whether it works better.

Planning versus experimentation is a false choice. The meta-analytic evidence suggests that planning helps modestly, and more for established small firms than for new ones; discovery-driven planning turns the plan into a schedule of tests.

Product-market fit remains the least rigorous idea in this module. It is widely used and loosely measured, and the survey thresholds are practitioners' rules of thumb. Retention by cohort is the most defensible measure, but even it can be flattered by heavy discounts or a narrow segment. The honest conclusion is that fit is a matter of degree, judged from several behaviours together.

Section 07: Decision exercise

A fictional situation. Decide before you read the guidance.

Raise now or validate further at HiveSense

HiveSense (fictional) sells a sensor that monitors beehive health and alerts beekeepers to swarming and disease. After nine months, 30 hobby beekeepers pay £8 a month; retention after three months is 80%, but no cohort is older than five months. The hardware costs £95 a unit to make in small batches. An investor offers £1.2 million for 25% of the company, on condition that HiveSense ships 1,000 units within a year. The company has £150,000 of cash.

  1. A. Take the money and scale

    Accept the offer, order 1,000 units and spend on marketing to hobby beekeepers. Fastest growth, but committed before longer-term retention and unit costs at volume are known.

  2. B. Run a paid pilot first

    Decline for now and sell 200 units over four months at full price, measuring six-month retention and the cost of acquiring customers, then raise on stronger evidence. Uses most of the cash and risks losing the investor.

  3. C. Test a pivot to commercial beekeepers

    Approach commercial pollination businesses with hundreds of hives each, offering a service contract. Fewer, larger customers and a different sales process; the investor's condition would no longer fit.

Judge the options on

  • Strength of the evidence for product-market fit
  • Cash runway and survival risk
  • Reversibility of the commitments involved
  • Dilution and the fit of the investor's conditions
  • Size of the market each path can reach
Read the guidance

Thirty customers with 80% three-month retention is encouraging but thin: the interval around it is wide, no cohort has passed a full season, and hobby beekeepers are early adopters. Option A converts that thin evidence into irreversible commitments, a large inventory order and a target that forces spending on acquisition before fit is shown. If retention sags after the summer, the company would hold stock it cannot sell and an investor whose condition it has missed.

Option B buys the evidence that matters most, longer retention and acquisition cost at full price, for a cost the company can almost afford, and keeps the decision reversible. Its risk is cash: it needs a small bridge or careful staging, and an honest conversation with the investor, who may accept a later round on better evidence.

Option C may be the larger opportunity, because commercial customers lose more money per failed hive and buy many units, but it is a different business with a different sales cycle. A strong answer pursues B and runs a small test of C in parallel, a few conversations and one paid trial, so that the next round is raised on evidence about which market really pulls.

Section 08: Further reading

  • Steve Blank (2013). Why the lean start-up changes everything. Harvard Business Review.

    A short statement of customer development and why it changes how new ventures are run.

  • Eric Ries (2011). The Lean Startup. Crown Business.

    The book that popularised the build-measure-learn loop, the minimum viable product and the pivot.

  • Saras D. Sarasvathy (2001). Causation and effectuation: toward a theoretical shift from economic inevitability to entrepreneurial contingency. Academy of Management Review.

    The theory of effectuation and its contrast with causal reasoning.

  • Joshua Gans, Erin L. Scott and Scott Stern (2018). Strategy for start-ups. Harvard Business Review.

    The entrepreneurial strategy compass and the founder's paradox of committing in order to learn.

  • Arnaldo Camuffo, Alessandro Cordova, Alfonso Gambardella and Chiara Spina (2020). A scientific approach to entrepreneurial decision making: evidence from a randomized control trial. Management Science.

    A randomised trial showing the effect of a scientific approach on start-up decisions and results.

  • Rita Gunther McGrath and Ian C. MacMillan (1995). Discovery-driven planning. Harvard Business Review.

    Discovery-driven planning, the bridge between business plans and experiments.

Every work cited here is listed in the platform's sources register. Real organizations appear only with facts from the sources shown; everything else is labelled illustrative or fictional.