As simple as possible;
but no simpler

Force Directed Graphs

<<Mouse-over >> nodes for Concepts, and edges for Relationships. <<Drag>> to explore



As data visualizations go, force-directed graphs can be fun to play with; and intuitively insightful and informative – especially in an executive management dashboard . A force directed graph is a physics-based simulation of visual elements linked as a graph. Elements are acted on by defined forces setup in the visualization layout.

Forces include:

  • Repulsion or attraction between elements, subject to maintaining an approximate specified distance between elements;
  • Attraction to centers of gravity; and
  • Collision prevention.

The interplay of these defined forces make a force-directed graph a richly tactile tool for visualizing complex and dynamic linkages between concepts.

E.g. As used here to visualize the complex operational linkages between key high level trading & regulatory concepts.

  • Each node represents a concept i.e. a fundamental business/operational concept, or a class of similar regulations (e.g. CCAR and other capital adequacy stress-testing related regulations, as a class).
  • Each edge represents the operational impact/linkage between two concepts – it may not suggest causality or directionality of impact. The strength of this impact is reflected in the attraction force between nodes.
  • The size of a node is proportional to the impact-weighted sum-product of all operational linkages the node has with other nodes.
  • Each node exerts a global repulsing force that is proportional to its size.
  • Finally, the canvas itself exerts pinning forces – a central gravity-like force that prevents nodes drifting away from the center of the canvas, and a clustering non-central but local force that aims to keep the regulatory concepts roughly grouped together (similarly but more weakly clustered groups are: decision driver concepts, revenue & risk origination concepts, operational cost driver concepts, and don’t-go-bust-now concepts).

These forces are tied to weights derived from organizational data, and collectively provide the physics guiding how the nodes move when dragged around; reflecting not only the expected linkages between concepts, but also the client-specific organizational and operational (in)efficiencies around these concepts.

i.e. Some concepts pull more (or less) weight than they should

e.g. In this example Insight (as a function of Research and Predictive Modeling) would appear to affect Risk Appetite (as it should); unfortunately Risk Appetite (an expected fundamental driver) does not seem to pull much weight with the things it should.


Nodes (first 12 – example)

id concept group connectedness
1 Liquidity CLAR/NSFR… Regulation 2.1
2 CCAR/DFAST/BofE/EBA Regulation 3.2
3 BCBS 239 Regulation 2.5
4 Resolution Planning Regulation 4.5
5 Volcker Regulation 3.8
6 DFA EPS Regulation 2.4
7 MIFID II Regulation 3
8 Basel3/FRTB/CRD-IV Regulation 2.7
9 Risk Appetite DecisionDriver 1.4
10 Insight DecisionDriver 1.3
11 Data Source DecisionDriver 1.9
12 P&L Unit RevenueAndRiskDriver 6.9

Connectedness (weighted sum product of Concept’s related Edges) drives Node size.


Edges (first 12 – example)

source target summary impact
Liquidity CLAR/NSFR… Regulation Significant Impact:(Controls,Operations,Staff) 0.6
Liquidity CLAR/NSFR… Funding Significant Impact:(Asset Quality,Inventory,Funding model) 0.8
Liquidity CLAR/NSFR… Scenario Validate,Calibrate,Review:(Controls,Inputs,Model) 0.6
Liquidity CLAR/NSFR… Valuation Validate,Calibrate,Review,Narrative:(Controls,Model,Outputs) 0.55
Liquidity CLAR/NSFR… Risk Factors Minimal Net Impact 0.5
Liquidity CLAR/NSFR… IDs & Taxonomies Significant Impact:(Inputs,Inventory) 0.5
Liquidity CLAR/NSFR… Capital GSIB buffer capital: CET1 up to 4.5% 0.65
CCAR/DFAST/BofE/EBA Regulation Significant Impact:{Controls,Operations) 0.85
CCAR/DFAST/BofE/EBA Scenario Validate,Calibrate,Review:(Controls,Inputs,Model) 0.5
CCAR/DFAST/BofE/EBA IDs & Taxonomies Integration,Aggregation:(Finance,Risk) 0.8
CCAR/DFAST/BofE/EBA Risk Factors Validate,Calibrate,Review:(Controls,Inputs,Model) 0.75
CCAR/DFAST/BofE/EBA Price Model Validate,Calibrate,Review:(Controls,Inputs,Model) 0.65

Impact of one concept on an another, a [0,1] range. May be estimated as a regression of y-predictors like cost changes, resource changes, skill upgrades. Determines the Strength of the Edge.