The Analytic Network Process (ANP) is a way to structure a decision, capture judgments about importance and influence, and turn those judgments into priorities for the options you care about (the alternatives).

This page is a short orientation—not a full textbook. For deeper theory, see the literature on ANP (e.g. work by Thomas Saaty) and related tools such as SuperDecisions and pyanp.

The idea in one minute

  1. You have a decision: which alternative is best, or how to allocate priority among alternatives.
  2. You group the factors that matter into clusters of nodes (criteria, stakeholders, options, …).
  3. You draw connections that say “this node influences that cluster.”
  4. You enter judgments (pairwise comparisons or ratings).
  5. The software builds matrices, computes a limit of influences, and reports priorities.

A pure hierarchy (AHP-style) is the special case with no feedback loops. ANP allows cycles and feedback between clusters.

Pieces of a model

Piece Plain language In Studio
Alternative An option you are ranking or choosing among Often an “Alternatives” cluster (or scores from subnetworks)
Cluster A group of related nodes Colored boxes on the Structure canvas
Node One element inside a cluster Circles/labels inside a cluster
Connection “A influences B’s group” Arrows / links between nodes and clusters
Judgment How much more important A is than B, or a rating Judgments stage (pairwise or ratings)

See the glossary for more terms.

From judgments to scores

Rough pipeline (details on Calculations):

  1. Unscaled supermatrix — local priorities from pairwise/ratings in each column
  2. Cluster matrix — how strongly clusters influence each other
  3. Scaled (weighted) supermatrix — cluster weights applied; columns normalized
  4. Limit matrix — long-run influence (method chosen per network in the Inspector)
  5. Global priorities — priorities of all nodes from the limit matrix
  6. Alternative scores — priorities for the alternatives (synthesized when there are subnetworks)

Stages in ANP Studio

Studio walks you through the work in order:

flowchart LR
  Structure["Structure<br/>build model"] --> Judgments["Judgments<br/>enter data"]
  Judgments --> Analysis["Analysis<br/>see results"]
  Analysis --> Researcher["Researcher<br/>explore commands"]
Stage What you do
Structure Create clusters and nodes, draw connections, set network options (synthesis formula, limit-matrix method)
Judgments Enter pairwise comparisons or ratings for each “with respect to” (wrt) context
Analysis Read matrices, globals, alternative scores; try sensitivity and influence
Researcher Optional notebook-style commands for inspection and experiments

Suggested first session

  1. File → Open Sample… → 02_ahp_best_car.anpstudio
  2. Look at clusters and links on Structure
  3. Open Judgments and click a node in the left list to see pairwise (or ratings)
  4. Open Analysis → Synthesis and scroll from Scaled Supermatrix through Limit Matrix to Alternative Scores

When you are ready for feedback networks, open 01_hamburger_marketshare.anpstudio.

Next: Structure · Glossary