Relevance Trees

Relevance Trees are a structured method for breaking down a broad problem, objective or research subject into increasingly specific components arranged in a hierarchical tree. The method helps researchers move systematically from a general question toward the sub-objectives, issues, activities, technologies or alternatives that need to be examined.

Relevance Trees are an established futures research method. Futures Research Methodology 3.0 treats Relevance Trees as a separate method, while noting their close conceptual relationship with Morphological Analysis. Earlier versions had combined the two approaches.

The apparent simplicity of the method can be misleading. Drawing branches beneath a broad topic is easy; constructing a hierarchy that adequately represents the problem is considerably harder. A relevance tree can look systematic while omitting an important dimension, duplicating another, mixing different levels of abstraction or introducing branches that have little connection with the original objective.

A relevance tree is not rigorous because it contains many branches. It is rigorous when each level decomposes its parent meaningfully without unnecessary overlap, omission or loss of connection to the original problem.

On this page

Relevance Trees Explained Simply
What Is a Relevance Tree?
How Relevance Trees Work
Types and Uses of Relevance Trees
How to Construct a Relevance Tree
Defining Levels and Branches
Relevance and Weighting
Dudovskiy Relevance Tree Branch Validation Framework
Relevance Trees vs Morphological Analysis
Relevance Trees vs Decision Trees
Application of Relevance Trees: an Example
Advantages and Limitations of Relevance Trees
Common Mistakes When Using Relevance Trees
Relevance Trees in Business Research
Relevance Trees in the Age of AI and Digital Research
When to Use Relevance Trees
Dissertation Example
Exam Tip

Aspect	Relevance Trees
Main purpose	Decompose a broad problem or objective systematically
Basic analytical question	What components, sub-components or means are relevant to the higher-level issue or objective?
Structure	Hierarchical tree
Starting point	Broad problem, topic, need or objective
Subsequent levels	Dimensions, sub-objectives, activities, means or increasingly specific components
Typical applications	Futures research, technological forecasting, strategic planning, research planning and problem structuring
Main strength	Makes the structure of a complex problem explicit
Main methodological risk	Producing a neat hierarchy that does not adequately represent the underlying problem

Relevance Trees Explained Simply

Imagine a company wants to investigate the broad question:

How can we reduce the environmental impact of our packaging?

That question is too broad to analyse effectively without further structure. A relevance tree could begin by dividing it into three major areas:

Reduce packaging material

Use more sustainable materials

Improve packaging recovery

The first branch could then be decomposed into reducing package size, reducing material thickness and eliminating unnecessary components. The second might divide into recycled materials, renewable materials and alternative materials. The third could contain reuse, recycling and take-back systems.

The tree therefore moves from:

Broad objective → Major dimensions → Sub-dimensions → Specific options

Instead of immediately choosing one solution, the researcher systematically maps what contributes to the higher-level objective.

The important question, however, is whether the branches actually represent that objective adequately. If transport efficiency is an important contributor to packaging’s environmental impact but is completely absent from the tree, the hierarchy may appear comprehensive while an important part of the problem has disappeared.

What Is a Relevance Tree?

A Relevance Tree is a hierarchical analytical structure in which a broad subject or objective is progressively decomposed into smaller and more specific elements. Contemporary foresight descriptions characterize the method as beginning with a general subject and proceeding through increasingly disaggregated components. It can also start with a high-level goal and identify progressively more specific goals and the means through which they might be achieved.

The tree metaphor reflects the structure. The initial problem or objective forms the starting node. Major branches represent its principal components or sub-objectives. These divide into smaller branches, which may themselves be decomposed until the researcher reaches a sufficiently detailed analytical level.

For example:

Objective

→ Increase organizational resilience

Major dimensions

→ Operational resilience
→ Financial resilience
→ Supply-chain resilience
→ Digital resilience

Sub-dimensions of supply-chain resilience

→ Supplier diversity
→ Inventory flexibility
→ Geographic diversification
→ Logistics alternatives

The hierarchical relationship is crucial. Each lower-level element should have a meaningful relationship with the node immediately above it.

This makes Relevance Trees more than a brainstorming device. Their purpose is to impose a defensible structure on a problem so that the researcher can see how specific issues relate to progressively broader objectives.

How Relevance Trees Work

Relevance Trees use hierarchical decomposition. A complex whole is progressively divided into constituent elements while preserving the relationship between each element and its parent.

Consider a broad research subject such as:

Future competitiveness of electric vehicle manufacturers

The first level might identify:

Technology | Supply chain | Market | Regulation | Organizational capabilities

The technology branch might then divide into:

Battery technology | Manufacturing automation | Software | Charging integration

Battery technology could subsequently divide into:

Energy density | Cost | Charging speed | Battery life | Material requirements

Each movement downward increases specificity.

Moving upward should produce the reverse logic. Battery life contributes to battery technology; battery technology contributes to technological competitiveness; technological competitiveness contributes to the broader question of future competitiveness.

This upward test is useful because it exposes weak branches. If the researcher cannot explain clearly how a terminal node contributes to its parent, the node may be misplaced or irrelevant.

A relevance tree therefore works in both directions:

Top-down: What components make up this issue?

Bottom-up: How does this component contribute to the higher-level issue?

Types and Uses of Relevance Trees

Relevance Trees can be constructed for different analytical purposes, and the labels attached to the levels should reflect the purpose of the particular tree.

An objective-oriented relevance tree begins with an overall objective and progressively identifies sub-objectives and means of achieving them. For example, the broad objective of achieving carbon-neutral manufacturing could be decomposed into energy, materials, production processes, logistics and waste, with progressively more specific interventions beneath them.

A problem-oriented or subject-oriented tree decomposes a broad area into increasingly specific components. This is useful when researchers need to establish the structure and boundaries of a complex research problem before investigating individual elements.

Relevance Trees have also been used in technological forecasting and planning. Historical business-planning applications used forms such as perspective trees to structure threats and opportunities and objective trees to clarify program options.

The exact labels are therefore less important than the hierarchical logic. Researchers should not force every relevance tree into an identical structure. The question is whether each successive level represents a meaningful decomposition of the level above.

How to Construct a Relevance Tree

The process begins by defining the central problem, topic or objective. This starting node needs to be sufficiently clear that researchers can judge whether subsequent branches are genuinely relevant to it.

The next stage is to identify the major dimensions or first-level branches. Literature review, expert interviews, workshops, Environmental Scanning and other exploratory techniques may contribute to their identification. At this stage, researchers should resist the temptation to begin immediately with detailed factors. The first level establishes the basic architecture of the entire tree.

Each major branch is then decomposed into more specific components. The same question is repeatedly applied:

What are the principal components, sub-objectives or means represented by this node?

Decomposition continues until the terminal branches reach a level appropriate for the analytical purpose. There is no universal number of levels that makes a relevance tree rigorous. A simple strategic problem may require relatively few levels, whereas a detailed technology-planning exercise may require considerably deeper decomposition.

The completed tree should then be reviewed rather than accepted simply because every branch contains several nodes. Researchers should examine whether important areas are missing, whether branches overlap and whether sibling nodes operate at comparable conceptual levels.

Depending on the purpose of the analysis, branches can subsequently be assessed or weighted according to their relevance to higher-level objectives.

Defining Levels and Branches

One of the most important design decisions concerns the level of abstraction.

Suppose the parent node is:

Factors affecting adoption of autonomous delivery vehicles

The following first-level branches might be coherent:

Technology | Economics | Regulation | Consumer acceptance | Infrastructure

Now consider an alternative set:

Technology | Vehicle reliability | Regulation | Cost | Consumer attitudes

This hierarchy is less coherent because “Technology” is a broad dimension while “Vehicle reliability” is a specific technological issue. “Economics” and “Cost” create a similar problem if both appear as equivalent branches at the same level.

The tree mixes levels of abstraction.

A useful test is to compare sibling nodes. Nodes sharing the same parent should normally represent reasonably comparable types or levels of decomposition.

Branches should also be sufficiently distinct. If “digital capability,” “technology capability” and “IT capability” appear as separate branches without clear definitions, the same issue may be represented repeatedly.

However, real-world systems do contain interdependencies. Relevance Trees impose hierarchical structure for analytical purposes, and this should not be mistaken for evidence that the real phenomena are completely independent. Researchers should distinguish between useful decomposition and an unrealistic assumption that branches never interact.

Relevance and Weighting

Some applications of Relevance Trees go beyond structural decomposition and assign relevance values or weights to branches. These values can help indicate the relative contribution of lower-level elements to higher-level objectives.

For example, if a tree examines technologies contributing to sustainable aviation, experts might assess the relative relevance of alternative propulsion systems, aircraft materials, aerodynamic improvements and operational technologies.

Such weighting can support prioritization, but the numbers do not become objective merely because they are numerical.

The basis of the weighting needs to be explained. Values may derive from expert judgement, empirical evidence, structured assessment or another explicit procedure. If the underlying tree is poorly constructed, sophisticated weighting cannot repair it.

This concern appears in the methodological literature. Freeman’s work on empirical relevance trees argued that doubts about the adequacy of the tree structure or the validity of relevance numbers can seriously weaken the conclusions obtained from the analysis. (Research Explorer)

Researchers should therefore treat weighting as a second analytical layer:

First: Is the tree itself defensible?

Then: Is the assessment of relevance defensible?

Dudovskiy Relevance Tree Branch Validation Framework

The Dudovskiy Relevance Tree Branch Validation Framework is a practical decision aid for evaluating the structure of a relevance tree after an initial hierarchy has been developed.

It does not replace established Relevance Tree methodology or claim a new form of hierarchical decomposition. Instead, it addresses a practical methodological problem: how can a researcher determine whether the proposed branches provide a defensible representation of their parent nodes?

The framework evaluates branches through six checks.

1. Parent relevance

Ask whether the branch genuinely contributes to explaining, decomposing or achieving its immediate parent node. A branch may be interesting in its own right but still belong elsewhere in the tree.

2. Distinctiveness

Compare the branch with its siblings. Substantial conceptual overlap suggests that branches may need to be combined, redefined or reorganized.

3. Coverage

Consider the sibling branches collectively. Do they provide adequate coverage of the parent issue, or is an important dimension missing? Coverage does not necessarily require an exhaustive list of everything imaginable; it requires adequate representation for the purpose of the analysis.

4. Level consistency

Check whether sibling branches operate at approximately comparable levels of abstraction. A broad category should not normally sit beside one highly specific example of that category without methodological justification.

5. Traceability

Follow the branch upward through the tree. Can the researcher explain the logical connection from the terminal node through each parent level to the central problem or objective? If the chain becomes difficult to justify, the branch structure should be reconsidered.

6. Analysability

At the terminal level, determine whether the branch has become sufficiently specific to investigate, compare, evaluate or translate into action. Continuing decomposition indefinitely adds complexity without necessarily adding analytical value.

The sequence can be summarized as:

Proposed Branch → Parent Relevance → Distinctiveness → Coverage → Level Consistency → Traceability → Analysability → Defensible Branch Structure

The central principle is:

The quality of a Relevance Tree depends not on how extensively a problem is divided, but on whether the resulting hierarchy preserves meaningful relationships between the research problem, its dimensions and its specific components.

Relevance Trees vs Morphological Analysis

Relevance Trees and Morphological Analysis are closely related but perform different analytical tasks. Their relationship is sufficiently close that earlier editions of the Millennium Project’s Futures Research Methodology treated them together; Version 3.0 presents them separately. (The Millennium Project)

Relevance Trees	Morphological Analysis
Decomposes a problem hierarchically	Structures a multidimensional configuration space
Moves from broad to increasingly specific	Identifies parameters and their alternative states
Focuses on parent-child relationships	Focuses on combinations of states
Produces a hierarchical tree	Produces a morphological field or box
Asks “What makes up this problem?”	Asks “What configurations are possible?”
Useful for structuring objectives and components	Useful for systematically exploring alternative configurations

Suppose researchers investigate the future of commercial agriculture.

A Relevance Tree might decompose the subject into technology, labour, climate, regulation, supply chains and consumer demand, and then divide each dimension into more specific issues.

Morphological Analysis might subsequently define important parameters and alternative states—for example, low/moderate/high automation or centralized/regional/local supply structures—and examine the configurations created by combining them.

The methods can therefore complement one another. A Relevance Tree can help structure the problem; Morphological Analysis can then explore alternative configurations within that problem space.

Relevance Trees vs Decision Trees

The tree-like appearance can make Relevance Trees and Decision Trees seem similar, but their analytical purposes differ.

A Relevance Tree primarily decomposes a problem, objective or subject into increasingly specific components. Its branches express relationships of relevance, contribution, decomposition or means-to-objective structure.

A Decision Tree represents alternative decisions and their possible consequences. In formal decision analysis, branches may also incorporate probabilities, outcomes and expected values.

The distinction is therefore conceptual:

Relevance Tree: What components or means are relevant to the broader issue?

Decision Tree: What choices are available and what outcomes may follow from them?

A relevance tree should not be interpreted automatically as a sequence of decisions merely because it is presented visually as branches.

Application of Relevance Trees: an Example

Consider a research project examining the long-term resilience of international pharmaceutical supply chains. Beginning directly with dozens of risks would make it difficult to determine whether the study covers the problem systematically. The researcher therefore constructs a Relevance Tree with:

International pharmaceutical supply-chain resilience

as the central node.

The first level contains:

The Supply branch is decomposed into active pharmaceutical ingredients, excipients, packaging materials and critical suppliers. Critical suppliers are then examined through geographic concentration, supplier substitutability and dependency.

The Manufacturing branch contains production capacity, geographic distribution, manufacturing flexibility and quality-control capability. The Logistics branch includes transport capacity, cold-chain infrastructure, border disruption and inventory positioning.

Rather than simply accepting this structure, the researcher evaluates it. For example, cybersecurity initially appears under both Information and Manufacturing. The overlap prompts the researcher to clarify whether cybersecurity should remain a separate first-level dimension or whether distinct operational and information-security components should be explicitly defined.

Expert interviews reveal another omission: workforce availability. The researcher evaluates where this belongs and revises the tree accordingly.

The relevance tree has therefore performed two functions. It has decomposed a complex research problem into analysable components, but it has also exposed overlap, omission and ambiguity in the researcher’s initial conceptualization of the problem.

The final tree can subsequently guide data collection, expert assessment or further futures analysis.

Advantages and Limitations of Relevance Trees

A major advantage of Relevance Trees is their ability to impose structure on broad and ambiguous problems. Researchers can see how specific issues connect with broader dimensions rather than working with an unorganized list of variables, technologies or objectives. The hierarchical representation can also reveal gaps and make the boundaries of an analysis easier to communicate.

The method encourages systematic decomposition. A researcher considering only familiar aspects of a problem may discover that an entire branch has been overlooked once the issue is represented hierarchically. Relevance Trees can also provide a useful foundation for subsequent methods, including expert assessment, technological forecasting, Morphological Analysis and scenario development.

Their apparent orderliness is also a limitation. The tree is a representation created by the researcher, not proof that the underlying system naturally possesses the same hierarchy. Complex social and business systems contain feedback, interaction and overlap that a simple parent-child structure may conceal.

The method also depends on judgement about what belongs in the tree, where it belongs and how far decomposition should continue. A missing first-level branch can affect everything beneath it. Likewise, inappropriate weighting can create an appearance of quantitative precision unsupported by the underlying evidence.

Relevance Trees are therefore strongest when the structure itself is treated as something requiring justification rather than merely as a convenient diagram.

Common Mistakes When Using Relevance Trees

One common mistake is starting with branches before defining the central problem precisely. If the root node is ambiguous, researchers have no stable criterion for determining whether subsequent branches are relevant.

Another is confusing detail with completeness. A tree may contain dozens of terminal nodes while still omitting an entire first-level dimension. Deep decomposition of one branch cannot compensate for missing coverage elsewhere.

Researchers may also create overlapping branches. If the same concept can reasonably be placed under several sibling nodes, the categories may need clearer definitions or restructuring. Some cross-branch interaction is inevitable in complex systems, but persistent conceptual duplication indicates a structural problem.

Mixing levels of abstraction is another frequent weakness. “Economic conditions,” “interest rates” and “competitor pricing” would not normally represent equivalent levels of decomposition without a specific rationale.

Finally, researchers sometimes assign weights before validating the structure. This creates precision around a hierarchy whose conceptual foundations may still be unstable.

Relevance Trees in Business Research

Relevance Trees have a long-standing connection with business planning and technological forecasting. Swager’s 1973 Business Horizons article describes their use for structuring and partitioning forecasting issues, including perspective trees for identifying threats and opportunities and objective trees for clarifying program options.

Contemporary business researchers can use the method for similarly complex problems. A study of a company’s future competitiveness, for example, might begin with market position, technology, operations, workforce, finance and regulatory environment before progressively decomposing each branch.

The method can also support innovation studies. A broad objective such as developing commercially viable low-carbon aviation could be decomposed into propulsion, aircraft design, fuels, infrastructure, regulation, economics and customer acceptance. Researchers could then identify specific technological or strategic questions beneath each branch.

The main business-research value lies in making explicit how specific factors relate to a broader strategic question. This can be especially useful before prioritization because it reduces the risk that immediately visible issues dominate merely because they are familiar.

Relevance Trees in the Age of AI and Digital Research

Generative AI can accelerate the construction of Relevance Trees dramatically. Given a broad research problem, an AI system can propose major dimensions, generate sub-branches, reorganize categories and expand a tree to several levels within seconds.

That capability changes the bottleneck. Generating branches becomes easy; validating the architecture becomes more important.

AI-generated trees may reproduce conceptual overlap without recognizing it. For example, an AI system might independently create “digital transformation,” “technology adoption” and “digital capability” as sibling branches because each phrase appears frequently in relevant literature. The resulting tree looks comprehensive while potentially counting related concepts several times.

AI can also create false completeness. A polished hierarchy with symmetrical branches may give researchers the impression that the problem has been exhaustively represented even when an important dimension is missing. Visual neatness is not evidence of conceptual coverage.

A more rigorous use of AI is iterative and adversarial. Researchers can ask an AI system to propose an initial tree and then separately challenge it: Which branches overlap? What dimensions may be missing? Are sibling nodes at equivalent abstraction levels? Which terminal nodes cannot be traced convincingly to the central objective? What alternative decomposition would produce a materially different representation?

Digital tools can also make large relevance trees easier to revise, compare and collaboratively evaluate. Yet the final methodological responsibility remains with the researcher. The dissertation should explain how the tree was developed and validated rather than presenting an AI-generated hierarchy as self-justifying evidence.

When to Use Relevance Trees

Relevance Trees may be appropriate when:

a research problem or objective is broad and needs systematic decomposition;
researchers need to identify increasingly specific components of a complex subject;
objectives need to be connected with sub-objectives, activities or possible means;
a futures or forecasting study requires a structured representation of an issue before further analysis;
researchers need to identify gaps, overlaps or poorly defined boundaries in a problem;
expert judgement can help validate the hierarchy or assess relevance;
research priorities or technology options need to be connected to broader objectives;
a complex problem needs to be structured before Morphological Analysis, scenario development or another futures method; or
the researcher needs a transparent visual representation of how detailed issues relate to the overall research problem.

Relevance Trees are less appropriate when the central research problem cannot meaningfully be represented hierarchically or when interactions and feedback relationships are the primary analytical concern. In those situations, network, systems or cross-impact approaches may represent the problem more faithfully.

Dissertation Example

Consider a dissertation titled “Technology Priorities for Decarbonizing European Maritime Freight by 2040: A Relevance Tree Analysis.”

The methodology chapter could justify Relevance Tree analysis because maritime decarbonization involves a broad objective supported by multiple technological, infrastructural and operational pathways. The researcher might develop the initial hierarchy from academic literature, industry reports and Environmental Scanning and subsequently refine it through interviews with maritime-industry experts.

The root node would represent the overall objective of reducing maritime-freight emissions. First-level branches might include propulsion technologies, alternative fuels, vessel efficiency, port infrastructure and operational measures. Each branch would then be decomposed into progressively more specific technologies or interventions.

The methodology chapter should explain how branches were defined and revised rather than simply presenting the finished diagram. For example, the researcher could assess parent relevance, overlap between sibling branches, coverage of the higher-level objective and consistency in levels of abstraction. Expert feedback might identify missing branches or lead to the restructuring of existing ones.

If relevance weights were subsequently assigned to terminal nodes, the dissertation would explain separately how those values were obtained and interpreted. The resulting priorities would therefore be grounded in both a transparent hierarchical structure and an explicit assessment procedure rather than emerging from an unexplained ranking exercise.

Exam Tip

When explaining Relevance Trees, do not define the method merely as “a diagram that breaks a topic into smaller parts.” That describes its appearance but misses its methodological purpose. A stronger answer explains that a Relevance Tree uses hierarchical decomposition to connect a broad problem or objective with progressively more specific components, sub-objectives, activities or means.

Remember the central methodological issue:

A detailed tree is not necessarily a good tree.

Its quality depends on whether the hierarchy provides adequate coverage, minimizes unnecessary overlap, maintains sensible levels of abstraction and preserves a logical connection between lower-level branches and the original research problem.

Trying to structure a complex dissertation problem into defensible research components?
Dudovskiy Research Assistant can help identify possible dimensions, develop a Relevance Tree, challenge overlapping or missing branches and explain how the resulting structure can be justified in your methodology.

References

Glenn, J.C. & Gordon, T.J. (Eds.) (2009). Futures Research Methodology—Version 3.0. The Millennium Project. The handbook includes a dedicated Relevance Trees chapter by The Futures Group International and Theodore J. Gordon. (The Millennium Project)

Swager, W.L. (1973). Technological forecasting in planning: A method of using relevance trees. Business Horizons, 16(1), 37–44. (ScienceDirect)

Freeman, J.M. (1983). Relevance Trees by Empirical Method. Department of Management Sciences Series No. 8303, UMIST, Manchester. (Research Explorer)

UNDP Global Centre for Public Service Excellence (2014). Foresight: The Manual. The manual describes Relevance Trees as an analytical technique for subdividing a broad topic into progressively smaller subtopics in a hierarchical structure. (undp.org)