Lakhlani International
Research

How to Write a Research Question, Hypothesis Register and Validation Plan

Most research projects that disappoint were never really designed. They started with a topic ("explore the market for this"), collected material and ended with a long document that did not settle anything. The cure is to design the research around a decision before collecting any evidence.

Lakhlani International's Research & Innovation page lists four deliverables that do exactly this: a research question and hypothesis register, a proposed validation plan and evidence criteria, a synthesis of findings, uncertainties and limitations, and a decision brief. This article shows how to write each one yourself, with templates. The examples use placeholders to show format, not real findings.

Step 1: Define the decision

The first step in the page's method is to define the decision and research question. Begin with the decision, not the topic.

Ask: what choice will this research inform, who will make it and by when? Examples of decisions:

  • →Whether to build a first version of a product.
  • →Whether to enter a new customer segment.
  • →Whether a new technology can replace an existing process.
  • →Whether to enter a research partnership.

If no decision depends on the answer, the research is probably curiosity, which is fine but should not be funded as if it were decision support.

Step 2: Write a research question that can be answered

The page lists research questions that are too broad to guide a decision as a core problem. A good research question is specific, answerable with available methods and tied to the decision.

Weak questions and better versions

  • →"What is the market for this technology?" is weak. Better: "Which two customer segments have the most urgent version of this problem, and will either commit to a paid pilot at the proposed price?"
  • →"Can AI help our operations?" is weak. Better: "Can a defined method reduce the time to process a specific type of request to a stated level, on our real data, without a stated error rate being exceeded?"
  • →"Is this a good idea?" is weak. Better: "Do the assumptions that the business depends on hold up in realistic conditions?"

Test your question with three checks. Could two people read it and agree on what counts as an answer? Could the answer change the decision? Can it be answered with the time and money available?

Step 3: Break it into hypotheses

A hypothesis is a specific, testable statement that you believe may be true and could be shown to be false. Break the research question into hypotheses, and label each as a feasibility, desirability or viability hypothesis, as discussed in the article on feasibility versus demand.

Good hypotheses are:

  • →Specific: they name who, what, how much.
  • →Falsifiable: it is clear what result would show them to be wrong.
  • →Linked to the decision: if the hypothesis fails, the decision changes.

Step 4: Build the hypothesis register

A hypothesis register is a table that records every hypothesis, why you hold it, how you will test it and what happened. Keep it as a living document.

Register fields

  • →ID. A short reference (H1, H2 and so on).
  • →Hypothesis. The statement being tested.
  • →Type. Feasibility, desirability or viability.
  • →Why we believe it. Existing evidence or reasoning, with sources.
  • →Risk. How likely to be wrong and how costly if so.
  • →Test. The method that will be used.
  • →Supports it. Evidence that would support it, set in advance.
  • →Weakens it. Evidence that would weaken it, set in advance.
  • →Owner and date. Who runs the test and when.
  • →Result. What happened, once tested.
  • →Alternative explanations. Other reasons the result may have occurred.
  • →Status. Untested, supported, weakened, inconclusive.

A worked placeholder row shows the format.

Example rows with placeholder thresholds

  • →H1, desirability. Hypothesis: quality managers at mid-sized food processors lose significant time to a specific manual check. Supports it: most interviewees describe recent, costly incidents unprompted. Weakens it: few mention it unprompted or describe a workaround.
  • →H2, feasibility. Hypothesis: the method detects the defect reliably on real samples. Supports it: detection meets the agreed threshold across a realistic sample. Weakens it: detection falls below the threshold in realistic conditions.
  • →H3, viability. Hypothesis: delivery cost per customer is below the price customers would pay. Supports it: costing shows a margin at tested price points. Weakens it: cost exceeds the price customers will accept.

The thresholds are placeholders. Set yours before testing.

Step 5: Review evidence and find the gaps

Before testing, review what is already known: published studies, market data, internal records, competitor products and customer conversations. The page describes this step as reviewing available evidence and identifying material gaps. Note where existing evidence is strong, weak or missing. Do not repeat tests that others have already done well.

Be wary of one-sided evidence. Published research tends to report positive results more often than negative ones. Daniele Fanelli's 2010 study found the odds of reporting a positive result were around five times higher in some social-science disciplines than in space science, which he attributes partly to bias. When reviewing published or vendor evidence, ask what negative results might not have been reported.

Step 6: Write the validation plan

A validation plan says how each hypothesis will be tested, with enough detail for someone else to run it.

Validation plan elements

  • →Method. Interviews, observation, prototype test, pilot, survey, cost model, desk research.
  • →Why this method. Why it is proportionate to the stakes and the stage.
  • →Sample. Who, how many, how chosen, and what bias that selection could introduce.
  • →Procedure. What will be done, in what order.
  • →Evidence criteria. The thresholds for support, weakening and each decision.
  • →Resources. People, time, money, facilities and specialists needed.
  • →Risks and limits. What the test cannot tell you.
  • →Timeline and owner. When it will be done and by whom.

Proportionality is the key idea. A cheap test is enough for a cheap decision. An expensive, hard-to-reverse decision justifies more rigorous evidence. If a test needs facilities or specialists you do not have, say so. The page notes that technical testing depends on suitable facilities, qualified specialists, methods and budgets being agreed.

Step 7: Set evidence criteria

For each hypothesis, decide in advance what would count as support or as a challenge. Avoid vague terms such as "strong interest". Prefer observable results such as a number of customers committing at a stated price, a measured detection rate in realistic conditions or a cost per unit.

Also decide how you will treat mixed results. Writing "if results fall between the thresholds, we will run a second, smaller test on X" prevents a debate afterwards.

Step 8: Run, document and synthesise

When the tests are done, write a synthesis with four parts.

  1. Findings. What was found, with the evidence.
  2. Uncertainties. What remains unknown and why.
  3. Limitations. Sample, method and time constraints that affect confidence.
  4. Alternative explanations. Other reasons the results might have occurred, and why they were accepted or rejected.

The page lists a synthesis of findings, uncertainties and limitations among its deliverables and asks for findings, limitations and alternative explanations to be documented. Include negative and inconclusive results. They are findings too.

Step 9: Write the decision brief

The brief should be short and usable by the decision-maker. A suitable structure:

Decision brief sections

  • →The decision. What choice this informs.
  • →What we tested. The hypotheses and methods, briefly.
  • →What we found. Results against the criteria.
  • →Confidence and limits. How sure we are and why.
  • →Options. Proceed, revise or stop, with what each would involve.
  • →Recommendation. One option with the rationale.
  • →Next experiments. What to test next, if proceeding or revising.

The companion article on proceed, revise or stop decisions explains how to set the decision rules.

Common mistakes

  • →Starting with a topic, not a decision.
  • →Writing hypotheses that cannot be proven wrong.
  • →Choosing tests that cannot disagree with you, such as a friendly sample.
  • →Setting thresholds after seeing the results.
  • →Dropping negative results from the report.
  • →Over-engineering the method for a low-stakes decision.
  • →Ignoring costs and limits of testing.

An illustrative scenario (hypothetical)

Consider a hypothetical innovation lead at a mid-sized manufacturer who is asked whether to invest in a new inspection approach. The lead writes the decision (approve or decline a pilot budget by the end of the quarter), a research question (does the approach meet the agreed detection threshold on real production samples at a cost per inspection below the current manual cost?) and four hypotheses with thresholds. A register records each hypothesis, test and result.

The approach meets the detection threshold on most samples but fails on one common material type, and the cost is only marginally lower. The synthesis records the limitation and an alternative explanation (the samples came from one line). The decision brief recommends revising: test on the failing material and a second line before approving the pilot. This scenario is invented for illustration and does not describe any client.

Next step

Write the decision, the research question and three hypotheses with their thresholds on one page. That alone often reveals which assumption is most dangerous.

If you want a second opinion on a research question or validation plan, the Research & Innovation capability sets out what an engagement covers. Findings and decisions are worth recording where a team can find them later; the Knowledge Maturity Management capability covers that.

Research & Innovation

A research question and hypothesis register, a validation plan with evidence criteria, a synthesis and a decision brief.

Explore the capability

Proceed, revise or stop

How to set decision rules at each gate, and why stopping early can be a successful outcome.

Read the decision guide

Feasibility is not demand

How to separate the three tests and start with the assumption that is riskiest and cheapest to check.

Read the earlier article

Knowledge Maturity Management

So findings, decisions and the reasoning behind them stay findable after the project ends.

Explore the capability

Contact Lakhlani International

For a second opinion on a research question or validation plan, start with a conversation.

Go to contact

Frequently asked questions

What is a hypothesis register?

A table that records each testable hypothesis, why you hold it, how it will be tested, what evidence would support or weaken it, and what happened. It keeps assumptions visible and results honest.

How do I write a good research question?

Tie it to a decision, make it specific, ensure it can be answered with available methods and check that two people would agree on what counts as an answer.

What are evidence criteria?

Pre-agreed thresholds that define what results would support or weaken a hypothesis and which decision follows. They should be observable and set before testing.

How much evidence is enough?

It depends on the stakes and the stage. Cheap, reversible decisions need less. Costly, hard-to-reverse ones justify more rigorous testing. Match the method to the decision.

Should negative results go in the report?

Yes. Negative and inconclusive results are findings. Recording them prevents repeating failed tests and gives decision-makers an honest picture.

What is the difference between a synthesis and a decision brief?

A synthesis summarises findings, uncertainties, limitations and alternative explanations. A decision brief uses the synthesis to set out options and a recommendation for a specific decision.

References

  • →Fanelli, D. "Positive" Results Increase Down the Hierarchy of the Sciences. PLOS ONE (2010). https://pmc.ncbi.nlm.nih.gov/articles/PMC2850928
  • →Brown, T. Change by Design: How Design Thinking Transforms Organizations and Inspires Innovation (2009).
  • →Lakhlani International. Research & Innovation capability page. https://lakhlaniinternational.com/capabilities/research-innovation

Related articles

Research

Preparing for an Applied Research Partnership in India: A Brief Template and What the 2025 RDI Scheme Signals

Mayank N Lakhlani

Indian businesses and founders considering research and innovation work have had a new public policy to watch since mid-2025. This article explains what the Research Development and Innovation Scheme says on its face, what is not yet clear from the public sources reviewed here, and how to prepare a clear brief for any research partnership, whether or not public financing is involved.

Read article→

Research

Proceed, Revise or Stop: How to Make Staged Decisions on an Early-Stage Idea

Mayank N Lakhlani

The hardest decision in innovation is rarely whether to start. It is whether to continue. Money has been spent, people are invested and the evidence is mixed. Without an agreed way to decide, teams tend to drift: extending a project because stopping feels like failure, or abandoning it because of one discouraging result.

Read article→

Research

Technical Feasibility Is Not Market Demand: Why Promising Ideas Fail When They Are Tested Late

Mayank N Lakhlani

A team spends a year proving that something can be built. The prototype works, the benchmarks are strong and the demo impresses visitors. Then the product reaches customers and the response is polite indifference. Nothing was wrong with the engineering. The question the engineering answered was not the question that decided the outcome.

Read article→

Research

What Is Applied Research and Innovation Consulting? How It Differs from Basic Research and Product Development

Mayank N Lakhlani

Many new ventures start with a strong belief: a technology that should work, a market that should want it, a way of operating that should be cheaper. Belief is a fine place to begin and a poor place to commit large budgets. The gap between "should" and "does" is where applied research earns its keep.

Read article→

Enjoyed this article?

Explore more ideas, research, and perspectives from the Knowledge Hub — or start a conversation with us.