Skip to content
Back to search results

MICMAC Structural Analysis

Introduced by
Michel Godet with Jean-Claude Duperrin (LIPSOR)
Year
2006
Type
Matrix-based pipeline for classifying system variables by influence and dependence
Family
Futures studies / French prospective school
Font size

What does it do?

MICMAC — Matrice d'Impacts Croises Multiplication Appliquee a un Classement, or Cross-Impact Matrix Multiplication Applied to a Classification — is the structural-analysis tool at the front of Michel Godet's la prospective method.

Developed by Godet with Jean-Claude Duperrin, it starts from a square matrix in which experts record the direct influence of each system variable on every other, then raises that matrix to successive powers to uncover the indirect influences that propagate along chains of variables. Summing the rows and columns of the iterated matrix yields each variable's overall influence and dependence, and plotting variables on the influence-dependence plane sorts them into driving (key) variables, dependent (result) variables, relay variables, and autonomous variables. The purpose is not prediction but diagnosis: to reveal which hidden variables truly drive the system, so that later scenario work focuses on the factors that matter.

The intuition behind the method

When experts list the variables shaping a system, they can usually say which factor directly pushes on which other factor, and they can write those direct effects into a grid. But the variable that matters most is often not the one with the loudest direct effects; it is the one whose influence ripples through the network, pushing on something that pushes on something else, until it quietly steers the whole system. MICMAC's trick is to let those ripples play out mathematically. Multiplying the influence matrix by itself counts the two-step paths of influence; multiplying again counts the three-step paths, and so on, until the ordering of variables by total reach stabilises. A variable that looked minor in the direct matrix can rise to the top once indirect chains are counted, while one that seemed central may turn out to be merely a relay. Plotting every variable by how much it influences others and how much it depends on them produces a map that tells the analyst, at a glance, where to push to move the system and which variables are simply along for the ride.

History

MICMAC was developed in France in the 1970s by Michel Godet together with Jean-Claude Duperrin as part of the structural-analysis toolkit of la prospective, the French strategic-foresight tradition centred at the Conservatoire national des arts et metiers and later the LIPSOR laboratory. It drew on the broader cross-impact tradition in futures studies but added the distinctive idea of multiplying the influence matrix to extract indirect effects and reclassify variables. Godet integrated MICMAC with companion tools — MACTOR for actor analysis, morphological analysis for scenario construction, and SMIC Prob-Expert for probabilised expert judgement — into a coherent prospective method documented in works including Creating Futures: Scenario Planning as a Strategic Management Tool (2006). LIPSOR's release of free MICMAC software helped spread the method across European foresight, territorial planning, and corporate strategy practice.

When to use it

Use MICMAC at the diagnostic, structuring stage of a foresight or strategy study, when you have a long list of interacting variables and need to discover which of them really drive the system before investing in scenario construction. It is well suited to complex problems with many interdependent factors where intuition about importance is unreliable and where a panel of experts can credibly judge pairwise direct influences. As the first analytical module of la prospective, it pairs naturally with downstream actor analysis (MACTOR) and morphological scenario building. It is less appropriate when the system has only a handful of obviously central variables, when no qualified group can assess the influence matrix, or when the goal is quantitative forecasting rather than qualitative diagnosis — MICMAC reveals structural roles and relationships, not magnitudes or probabilities of future outcomes.

Limitations and considerations

  • The entire analysis rests on subjective expert judgements of pairwise influence, so garbage in yields garbage out.
  • The ordinal influence scale and equal weighting of links impose strong simplifications on genuinely graded relationships.
  • It is static, capturing the system's structure at one moment without modelling how relationships themselves evolve.
  • Filling an n-by-n matrix grows quadratically with the number of variables, becoming tedious and error-prone for large lists.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Exposes indirect and hidden driving variables that a reading of direct influences alone would miss, via matrix iteration.
  • Reduces a sprawling list of interacting factors to a clear influence-dependence map that focuses subsequent analysis.
  • The matrix-filling process itself is a valuable structured discussion that surfaces unspoken assumptions about the system.
  • Provides a transparent, reproducible classification of variables into driving, relay, dependent, and autonomous roles.

Applications

  • Identifying the key driving variables of a system at the start of a scenario-planning or foresight study.
  • Structuring complex strategic problems by reducing many interacting factors to a clear influence-dependence map.
  • Supporting territorial and regional foresight by diagnosing which variables govern a region's development.
  • Informing corporate strategy by distinguishing the levers management can pull from the outcomes it merely registers.
  • Feeding morphological scenario construction and actor analysis with a focused set of genuinely influential variables.

How it works

How it works

1

The analysis begins by assembling the full list of variables that describe the system under study, drawn from internal factors the organisation controls and external factors in its environment. Godet emphasises doing this through expert and stakeholder workshops, brainstorming and then consolidating a comprehensive but bounded set — typically several dozen variables, each clearly defined so that participants share a common understanding. The completeness of this list is important, because MICMAC can only reclassify variables that have been included; a driving force omitted here cannot be discovered later. Defining the variables precisely also makes the subsequent pairwise influence judgements consistent and defensible.

Formula
V = \{v_1, \ldots, v_n\}\ \text{system variables identified by experts}
2

Experts then fill in the structural-analysis matrix, a square table whose entry in row i and column j records the direct influence that variable i exerts on variable j, usually on a simple ordinal scale such as nil, weak, moderate, or strong, with a separate mark for potential influences. The judgement is made pairwise and directly: does i, on its own, influence j, and how strongly. Godet notes that the discipline of considering every ordered pair often surfaces relationships the group had not consciously articulated, making the matrix-filling workshop valuable in itself. This direct-influence matrix is the sole quantitative input to the method.

Formula
M = [m_{ij}],\quad m_{ij} \in \{0,1,2,3\}\ \text{direct influence of } v_i \text{ on } v_j
3

The mathematical core of MICMAC is the iterated multiplication of the direct-influence matrix. Multiplying the matrix by itself yields a matrix whose entries reflect chains of influence of length two — the indirect effect of i on j through some intermediate variable — and raising it to higher powers captures ever-longer chains. Godet observes that as the power increases the relative ordering of the variables by influence and by dependence converges to a stable hierarchy. This iteration is what allows MICMAC to expose indirect and hidden influences that a reading of the direct matrix alone would miss, often substantially reordering the apparent importance of the variables.

Formula
M^{k} = M \cdot M^{k-1},\quad M^{k}_{ij} = \text{paths of length } k\ \Rightarrow\ \text{indirect influence as } k \to \infty
4

From the stabilised power matrix the method derives two scores for each variable: its influence (or motricity), obtained by summing across its row, measuring how much it acts on the rest of the system; and its dependence, obtained by summing down its column, measuring how much the rest of the system acts on it. Computing both the direct and the indirect versions of these scores lets the analyst compare the picture given by immediate effects with the picture that emerges once propagation through the network is accounted for. These two coordinates summarise each variable's structural role in a single influence-dependence pair.

Formula
\text{Infl}(v_i) = \sum_j M^{k}_{ij},\quad \text{Dep}(v_i) = \sum_j M^{k}_{ji}
5

Finally the variables are plotted on the influence-dependence plane and read off by quadrant. Variables high in influence but low in dependence are the driving or key variables that steer the system and deserve priority in scenario building; variables high in both are relay or stake variables, unstable because acting on them feeds back through the system; variables high in dependence but low in influence are the dependent result variables that register the system's outcomes; and variables low in both are relatively autonomous, largely disconnected from the dynamics. Godet uses this classification to focus the rest of the prospective study, ensuring that scenario construction and actor analysis concentrate on the variables that genuinely govern the system's evolution.

Formula
(\text{Infl},\text{Dep}) \to \{\text{driving},\ \text{relay},\ \text{dependent},\ \text{autonomous}\}\ \text{quadrants}

Common pitfalls

  • Omitting a genuine driving variable from the initial list, since MICMAC can only reclassify variables that were included.
  • Treating the influence-dependence quadrants as exact rankings rather than as a heuristic map open to interpretation.
  • Confusing relay (stake) variables with driving variables and acting on them without anticipating the destabilising feedback.
  • Letting a few dominant experts impose the influence scores, baking their biases into the supposedly objective classification.

Frequently asked questions

Why multiply the influence matrix by itself?

Because direct influence tells only part of the story. A variable may exert little direct influence yet shape the system powerfully through chains — influencing a factor that in turn influences many others. Multiplying the direct-influence matrix counts the two-step paths of influence, multiplying again counts the three-step paths, and continuing reveals the cumulative indirect reach of each variable. As the power rises the ranking of variables by influence and dependence stabilises, often promoting hidden drivers that the direct matrix understated. This propagation of indirect effects is exactly what distinguishes MICMAC from simply reading the raw matrix.

What do the four variable categories mean?

Plotting variables by influence and dependence yields four quadrants. Driving (key) variables are high-influence, low-dependence: they steer the system and are the priority for action. Relay or stake variables are high in both: acting on them is risky because effects loop back through the system. Dependent (result) variables are low-influence, high-dependence: they reflect the system's outcomes rather than cause them. Autonomous variables are low in both: largely disconnected and safe to set aside. The classification tells the analyst where leverage lies and which variables merely register or sit outside the dynamics.

Does MICMAC predict the future?

No. MICMAC is a structural-analysis and diagnostic tool, not a forecasting method. It does not produce probabilities, magnitudes, or timed predictions; it produces a map of the structural roles variables play in a system. Its job within la prospective is to focus the study by revealing which variables genuinely drive the system, so that the later scenario-building, actor-analysis, and probabilistic stages concentrate their effort where it matters. Anyone expecting numerical forecasts from MICMAC has mistaken its purpose; it answers 'what governs this system?' not 'what will happen?'.

Related methods

Sources

  1. Godet, M. (2006). Creating Futures: Scenario Planning as a Strategic Management Tool (2nd ed.). Economica.

    ISBN: 9782717852448
    Copied.