|
15–18 min read |
Table of contents
- The temptation to start collecting information
- 1. Define the problem before investigating it
- 2. Formulate the questions that need to be answered
- 3. Establish the scope of the investigation
- 4. Develop competing hypotheses
- 5. Design an investigative strategy
- 6. Decide when the investigation is sufficient
- 7. From investigation to decision
- An investigation is only as useful as the questions it is designed to answer
- Practical tool — The investigation design framework
The temptation to start collecting information
A major project is months behind schedule. The client blames the contractor for poor planning. The contractor blames the client for late approvals and shifting requirements. Each side has assembled a folder of emails, minutes and invoices that, in its view, settles the matter. A third party is asked to “look into it”.
The instinctive response is familiar: request the documents, schedule interviews, and try to work out who is responsible. It feels productive. Material accumulates, and with it a sense of progress.
Yet one question usually remains unasked: do we know precisely what we are trying to establish?
An investigation that begins with collection, without first defining its object, tends to become disorganised, expensive and, often unnoticed, biased. Without a defined question, the investigator cannot tell which documents matter, which witnesses are relevant, or when to stop. Worse, the first plausible explanation quietly becomes the lens through which everything else is read.
This article proposes a different starting point, summarised in one principle: the quality of an investigation is largely determined before the first piece of evidence is collected. The design of an investigation conditions what it will be possible to see, to understand and, ultimately, to conclude. There is a real difference between searching for information that confirms an explanation and building an investigation capable of choosing between several explanations. The first risks becoming an exercise in confirmation; the second is designed to test and distinguish between competing explanations. The seven-part framework presented at the end of this article (problem definition, investigative questions, scope and authority, competing hypotheses, evidence strategy, sufficiency assessment and decision relevance) offers a way to a
1. Define the problem before investigating it
1.1. Distinguishing the dispute from the underlying problem
A dispute is what the parties say is wrong; the underlying problem is what actually needs to be understood. The two are rarely identical. “The contractor is responsible for the delay” is a position in a dispute. The underlying problem might be: why did the project fall behind its planned schedule, and what contributed to it?
Treating the dispute as the problem has a practical consequence: the investigation inherits the parties’ framing, and its output is limited to assessing their competing accounts. Reformulating the problem in neutral terms opens the possibility that neither party’s account is complete, or that the causes lie elsewhere, for instance in an ambiguous specification, an unrealistic baseline schedule or a third-party dependency.
1.2. Separating allegations, concerns and established facts
At the outset, information typically arrives in three forms that should be kept distinct:
- Allegations: specific assertions made by a party, usually in support of its position (“the supplier delivered late on three occasions”).
- Concerns: general or unspecific worries, often without identified facts behind them (“something seems wrong with the procurement process”).
- Established facts and undisputed matters: matters supported by sufficiently reliable and verifiable sources, or expressly accepted by the relevant parties (“the contract was signed on 12 March”). Their evidential status should be recorded separately, distinguishing independently verified facts from matters accepted but not independently verified.
Maintaining this separation from the first day prevents a common drift: an allegation repeated often enough, or confirmed by a second party with an interest in the same outcome, begins to be treated as fact. A simple working register, with one column for each category and a note of the source, is usually sufficient. “Undisputed” and “verified” should not become synonyms: the parties may share the same error, or may have no interest in contesting a point.
1.3. Defining the investigative mandate, purpose and expected outcome
An investigation can serve different purposes: understanding what happened, determining causes, assessing compliance with obligations, quantifying loss, supporting a settlement, or preparing for formal proceedings. Each purpose implies a different standard of rigour, a different scope and a different form of output.
The purpose should be written down in a sentence or two and agreed with whoever commissions the work. If it cannot be stated clearly, the investigation is not yet ready to begin.
An investigation does not take place in an institutional vacuum. Beyond the problem to be understood, the investigator needs a clear mandate: the mission entrusted to them and the limits of their authority. The mandate is distinct from the purpose: an investigation may have a perfectly clear purpose and still be conducted under an insufficient or ambiguous mandate. Before starting, the following should be settled:
- Who commissions the investigation, and to whom does the investigator report?
- What is the extent of the mandate, and what powers of access to information does it carry?
- May the persons concerned be interviewed, and on what terms?
- Are there specific confidentiality requirements?
- Who will take the decision informed by the findings?
This does not turn the investigation into a legal exercise. It is a reminder that an investigation must be designed within a framework of authority, accountability and legitimacy; otherwise its findings may be challenged on grounds unrelated to their substance.
1.4. Identifying the decisions that the investigation is expected to inform
The most useful test of an investigation’s design is the decision it will support: whether to settle, to terminate a contract, to discipline an individual, to commence proceedings, to change a process. Identifying the decision in advance determines what level of certainty is needed, which questions are essential and which are merely interesting.
An investigation that answers questions no decision-maker needs answered is, however thorough, wasted effort.
2. Formulate the questions that need to be answered
2.1. From a broad concern to specific investigative questions
A concern such as “the project was delayed” is not an investigative question. It must be broken down into questions that can be answered with identifiable sources: When was each milestone planned and when was it achieved? Which activities were on the critical path? What changes were requested, when, and by whom? What was the effect of each on the schedule?
A well-formed question is specific, answerable in principle, bounded in time and subject, and neutral in its wording. “Why did the contractor fail to plan properly?” fails the last test: it contains its own conclusion.
2.2. Distinguishing factual, causal and responsibility-related questions
Three categories of question call for different types of evidence and should not be conflated:
- Factual questions ask what happened. (“On what date was the revised design delivered?”)
- Causal questions ask why it happened or what effect it had. (“Did the late delivery of the design delay the foundation works?”)
- Responsibility-related questions ask who is accountable under a given standard. (“Was the client contractually obliged to deliver the design by that date?”)
Factual findings inform causal analysis, while responsibility-related assessments require both an understanding of the relevant events and the applicable standards. These dimensions are distinct, but they may inform one another throughout the investigation: the contractual framework, for instance, may determine which facts are relevant, and an obligation may be established independently of any causal analysis. Many flawed investigations skip directly to responsibility, working backwards to justify an attribution that was assumed from the start.
2.3. Identifying what is known, unknown and disputed
For each question, a short status assessment is valuable: what is already known and uncontested, what is unknown, and what is disputed between the parties. This mapping serves several purposes. It prevents effort being spent re-verifying uncontested points, directs attention to the real areas of disagreement, and makes visible the gaps that no party has addressed because neither has an interest in doing so.
2.4. Prioritising questions according to their relevance to the dispute
Not all questions carry equal weight. Some are determinative: their answer could change the outcome of the decision. Others are contextual or marginal. Prioritisation should follow two criteria: the importance of the question to the decision identified in step 1.4, and the feasibility of answering it within the available time and resources. Determinative questions that are also feasible come first.
3. Establish the scope of the investigation
3.1. Defining the relevant period, events and actors
Scope has three dimensions. The temporal dimension fixes the period under review, which should usually extend somewhat before the first alleged event to capture context. The factual dimension identifies the events, transactions or decisions that are in question. The personal dimension identifies the individuals and entities whose actions, knowledge or decisions are relevant.
Each boundary should be justified by reference to the questions formulated in step 2, not by convenience or by the parties’ preferences.
3.2. Identifying the documents, data and sources likely to matter
With questions and boundaries in place, it becomes possible to identify, in advance, the categories of source likely to be informative: contractual documents, contemporaneous correspondence, system logs, financial records, meeting minutes, third-party reports, witness knowledge. A source map linking each question to its probable sources is a practical tool at this stage (see 5.1).
Contemporaneous and independent sources can provide valuable safeguards against retrospective reconstruction and interested accounts. However, neither contemporaneity nor independence guarantees reliability. Each source must ultimately be assessed in relation to its origin, purpose, integrity and consistency with other available information.
3.3. Determining what falls outside the investigation
Exclusions should be as deliberate as inclusions. Stating explicitly what the investigation will not address, whether other projects, other periods, or unrelated allegations, protects against scope creep and manages expectations. It also provides a record of the reasoning if the question of scope is later challenged. Scope must also remain within the mandate defined in section 1.3: what the investigator is authorised to access, examine and report on.
3.4. Balancing completeness, proportionality and available resources
Completeness is an aspiration, not a requirement. The relevant standard is proportionality: the effort invested should be commensurate with the stakes, the complexity of the matter and the use to which the findings will be put. A dispute over a modest sum does not justify a forensic review of every email; conversely, an allegation of serious misconduct may warrant considerable depth.
Where limits on time, budget or access prevent full coverage, they should be acknowledged at the outset and reflected in how conclusions are later qualified.
3.5. Adapting the level of investigation to the circumstances
The framework presented in this article is not intended to impose the same level of procedural formality on every investigation. Its application should be proportionate to the nature of the matter, its complexity and the potential consequences of the decision.
In a straightforward case involving a limited issue, readily verifiable facts and modest consequences, the investigator may address several steps together or handle them informally. A short factual review may be sufficient, provided that the problem is clearly understood, the relevant information is checked and the resulting conclusion is appropriate to the decision at hand.
More complex or contested matters call for a more structured approach. Where several plausible explanations exist, relevant information is dispersed, or the findings may have significant financial, contractual, legal or human consequences, the seven steps should be explicitly considered and documented to an appropriate degree.
The level of formality should also remain responsive to what emerges during the investigation. An initially straightforward matter may reveal inconsistencies, additional issues or consequences that justify expanding its scope and depth. Conversely, an investigation should not become more elaborate than necessary when the questions have been adequately answered.
Proportionality therefore applies not only to the resources devoted to an investigation, but also to the depth and formality of the method used. The objective is not to complete every procedural step for its own sake, but to ensure that the investigation is sufficiently rigorous for the decision it is intended to inform.
4. Develop competing hypotheses
4.1. Why the first explanation should not become the only explanation
Human reasoning gravitates towards the first coherent explanation it encounters, particularly when it is offered confidently or fits a familiar pattern. Once adopted, it shapes what is sought, what is noticed and how ambiguous material is interpreted. An investigation built around a single explanation will almost always find support for it, because supporting evidence is what it is looking for.
The remedy is procedural rather than psychological: require, as a matter of method, that more than one explanation be formulated and tested.
4.2. Identifying alternative explanations
In the schedule-delay example, plausible hypotheses might include: (a) the delay resulted mainly from the contractor’s inadequate resourcing; (b) it resulted mainly from late or changing client instructions; (c) it resulted from a flaw in the original baseline schedule that neither party identified; (d) it resulted from an external event outside either party’s control; or (e) a combination of several of these.
The parties’ own positions usually supply the first two. The investigator’s task is to add explanations that neither party has an interest in raising, including the possibility that neither party’s framing is entirely correct, or that the premise of the dispute is itself partly mistaken.
4.3. Determining what evidence would support or weaken each hypothesis
For each hypothesis, ask two questions before collecting anything: What would we expect to see if this were true? and What would we expect to see if it were false? The most valuable evidence is diagnostic: it is significantly more consistent with one hypothesis than with competing explanations. Evidence that is equally compatible with several hypotheses, however abundant, adds little to the ability to distinguish between them.
This step converts a vague search for “relevant material” into a targeted search for specific, discriminating items.
4.4. Avoiding confirmation bias during the investigative process
Several practices reduce the risk of confirmation bias:
- Record each hypothesis and revisit it at defined points, noting what has strengthened or weakened it.
- Actively seek disconfirming evidence for the hypothesis currently favoured.
- Phrase interview questions in an open, non-leading manner.
- Where resources allow, have a second reviewer challenge the emerging analysis.
- Keep a clear distinction in notes between what was observed and what was inferred.
None of these eliminates bias; together they make it visible and therefore manageable. The investigator should also examine the conditions in which information is produced and communicated. An account can be sincere without being accurate; a person may act in good faith while having an incomplete perception of events; and a contradiction between accounts does not necessarily mean that someone is lying.
5. Design an investigative strategy
5.1. Matching investigative questions with appropriate sources
A strategy matrix, with questions in rows and sources in columns, makes the plan explicit. Each question is linked to the sources most likely to answer it, with an indication of reliability and accessibility. Questions that depend on a single source are flagged as vulnerable; questions for which no source can be identified are flagged as potentially unanswerable and the implications communicated early.
5.2. Planning document reviews, interviews and data analysis
The three principal methods serve different functions:
- Document review establishes what was recorded at the time and provides a framework against which accounts can be tested.
- Interviews supply context, explain gaps and identify further sources, but are vulnerable to memory, interest and suggestion. They are usually best conducted after the principal documents have been reviewed, so that accounts can be tested against the record.
- Data analysis can reveal patterns, anomalies and timelines that are not apparent from individual documents.
The sequencing of these methods is a design decision, not an administrative one.
Preserving information. A strategy that identifies the right documents but does not secure them can compromise everything that follows. From the outset, keep four related operations distinct: identifying the relevant information, preserving the sources and records that exist, ensuring their traceability, and examining their integrity and authenticity. Where records are at risk (automatic deletion, system migrations, departing staff), appropriate preservation measures should be considered as soon as the risk is identified, and implemented before relevant information can be lost or altered.
5.3. Identifying dependencies between investigative steps
Some steps cannot begin until others are complete: an interview may depend on a document that has not yet been obtained; a data analysis may depend on the extraction of records from a system. Mapping dependencies identifies the critical path of the investigation itself and avoids wasteful repetition, such as interviewing the same person twice because a key document arrived late.
5.4. Determining where specialist expertise may be necessary
Some questions exceed the investigator’s competence: scheduling analysis, forensic accounting, digital forensics, engineering or technical standards. Identifying these needs at the design stage allows experts to be engaged in time, with clear instructions, rather than at the end to validate conclusions already reached. Expert input should answer defined questions; it should not be left to define them.
6. Decide when the investigation is sufficient
6.1. The difference between having collected information and having answered the question
Volume is not a measure of progress. The relevant test is whether each investigative question can now be answered, with what degree of confidence, and on the basis of which evidence. An investigation can hold thousands of documents and still be unable to answer its central question; it can also reach a reliable conclusion from a modest, well-chosen body of material.
Sufficient for what? The level of confidence required depends on the purpose of the investigation and the consequences of the decision it is intended to inform. An internal process review, a contractual dispute and an allegation of serious misconduct may require different levels of evidential support. There is no universal threshold of sufficiency.
6.2. Identifying information gaps that could change the conclusion
The critical review is to identify what is missing and to ask whether its absence matters. A gap is significant if the missing information, depending on what it showed, could lead to a different conclusion. Gaps that could not change the outcome, however regrettable, do not justify further effort. Those that could should be pursued if feasible, or recorded as limitations if not.
6.3. Assessing the proportionality of further investigation
Further work has diminishing returns. At each stage, weigh the expected value of additional investigation (the likelihood that it would alter the conclusion and the importance of the decision) against its cost in time, money and disruption. When the marginal contribution is low, stopping is a legitimate and defensible decision, provided the reasoning is documented.
6.4. Recognising when uncertainty must remain explicit
Some questions cannot be answered conclusively: records may be missing, witnesses unavailable, or the evidence genuinely equivocal. In such cases, the sound course is to state the uncertainty and its extent, not to resolve it by assumption. Expressing findings with calibrated confidence, for example “established”, “probable”, “possible” or “not determinable”, is more useful to a decision-maker than an overstated certainty that will not withstand scrutiny. These terms should be defined for the purposes of the investigation rather than treated as universal evidential categories. A level of confidence should also attach to a specific proposition, not to a person or a file as a whole: one may be highly confident about the date of a delivery and much less so about its effect on the project schedule.
7. From investigation to decision
7.1. What the investigation can establish
A well-structured investigation can establish facts, reconstruct sequences of events, test competing explanations against the evidence, identify causal links to the extent the evidence permits, and clarify which points remain contested and why. These are the building blocks of any subsequent assessment.
7.2. What it cannot establish
An investigation cannot establish what the evidence does not support. It cannot, by itself, determine legal liability, which depends on applicable standards and, often, on a decision-maker with proper authority. Nor can it bridge evidentiary gaps through unsupported inference. Distinguishing between findings of fact, analytical conclusions and normative judgements keeps the investigation within its proper role.
7.3. How to communicate limitations and unresolved questions
Limitations belong in the main body of the findings, not in a footnote. A clear account should state the questions addressed, the sources consulted, the sources unavailable, the methods used, the confidence attached to each conclusion and the questions left open. This transparency strengthens the credibility of the work and protects both the investigator and the reader from over-reliance on it.
7.4. Preparing the transition from investigation to assessment
The final step is the hand-over: organising findings so that the decision-maker can move from “what happened” to “what follows”. This means presenting facts, causal analysis and open questions in a form that maps onto the decision identified at the start (see 1.4). Some evaluation of the evidence necessarily takes place during the investigation itself, shaping further questions and the revision of hypotheses. The distinction is therefore not between two unrelated activities: the investigation produces and organises findings and analysis, whereas legal, contractual, disciplinary or managerial assessment belongs to the appropriate stage and to the person competent to carry it out. The investigator should therefore make clear where factual and analytical findings end and where the assessment of their implications begins.
An investigation is only as useful as the questions it is designed to answer
The temptation to begin collecting information immediately is understandable and, in many cases, counterproductive. Evidence has no meaning in the abstract; it acquires meaning only in relation to a question. The quality of an investigation is largely determined before the first piece of evidence is collected. It depends less on the quantity of material gathered than on the discipline of its design: defining the problem clearly, formulating precise questions, establishing a clear mandate and deliberate boundaries, testing competing hypotheses, matching the evidence strategy to the questions, and assessing honestly what has and has not been established.
Structure does not make an investigation slower. It makes it faster where it matters, by avoiding irrelevant searches, and more reliable where it counts, by exposing assumptions before they harden into conclusions. This is the difference between an investigation that looks for support for one explanation and one built to choose between several, with a level of structure and formality proportionate to the matter at hand.
Practical tool
The investigation design framework
The seven steps below provide a framework for designing and guiding an investigation. They can be used as a checklist at the outset and as a reference point throughout, with a level of detail and formality adapted to the circumstances.
| Step | Core question | Output |
|---|---|---|
| 1. Problem definition | What exactly needs to be understood? | A neutral problem statement |
| 2. Investigative questions | What needs to be established? | Prioritised questions |
| 3. Scope and authority | What is included, excluded and authorised? | A defined investigative scope within the authorised mandate |
| 4. Competing hypotheses | What plausible explanations must be tested? | Testable competing hypotheses |
| 5. Evidence strategy | Which sources and methods can answer each question? | An investigation strategy matrix |
| 6. Sufficiency assessment | Is the available information sufficient to answer the questions? | Findings, confidence levels and limitations |
| 7. Decision relevance | What conclusions can inform the decision, and what remains for the decision-maker? | Structured findings and a decision handover |
The seven stages
01 Problem definition
→
02 Investigative questions
→
03 Scope and authority
→
04 Competing hypotheses
→
05 Evidence strategy
→
06 Sufficiency assessment
→
07 Decision relevance
The sequence is not strictly linear. New information may require revisiting the questions or the scope. What matters is that each revision is made deliberately and recorded, so that the investigation remains a controlled process rather than a drift.
The framework should also be applied proportionately. In straightforward cases, several steps may be addressed together or handled informally. In complex, contested or high-stakes matters, each step should be explicitly considered and documented to an appropriate degree. The outputs listed in the table are possible work products, not documents to be produced in every case.
Continue exploring
Investigation is only the beginning of a broader process of understanding and judgment. Explore the related perspectives below.
Next in Responsa
Facts, assumptions and inferences: why the distinction matters
Learn to distinguish what is established, what is assumed and what is inferred when analysing a disputed situation.
Related in Sapientia
Explore the limits of procedural and analytical reasoning, and the role of practical judgment when competing rationalities collide.
Related in Corpus
How to Measure Anything — Douglas W. Hubbard
Explore how measurement, uncertainty and the value of information can improve decision-making.