Dependency matrix

[STUB]

List of methods / Knowledge management methods

The Dependency matrix is a method for generating improvement initiatives, through identifying meaningful relationships between real-world problems and the available organizational knowledge, at an appropriate level of granularity. It is part of the Knowledge mapping assemblage.

Name and history

The Dependency matrix method was formerly known as QQE matrix, where "QQE" stood for Qual, Quant, and Emergence.

Method

Purpose

The main goal is to find useful connections between knowledge objects, created upstream through the process of Decision mapping and ASHEN, to things that decision makers consider important and pay attention to. Critically, these are real-world problems, as opposed to idealistic future goals.

The Dependency matrix is both a diagnosis and an intervention design component within the Cynefin® framework.

Approach

The process, run by a complex facilitator, walks a group through the production of a matrix based on input data.

Output

The output is the completed matrix of data and an agreed prioritised list of actions or micro-projects. It identifies opportunities and therefore a portfolio of strategic knowledge projects.

The method outputs normally constitute inputs for the subsequent application of a contextualization method, such as 3-points, where the Cynefin® framework is overlapped to them and serve to address each of them coherently with the determined domain.

Prior knowledge

This Method requires skills in complex facilitation, in order to avoid any outside influence and in order to avoid premature convergence. Facilitators should be familiar with the principles for managing knowledge and with those of complex facilitation.

No prior knowledge is needed of participants

Preparation

Input gathering

To populate the rows in your grid, generate a short list of "things that matter" to the organization - either things decision makers value or things that matter in a more abstract sense. These will be typically "core processes" or "essential functions". There are two ways to do this:

You can gather material separately from different decision makers, by asking them to answer the question: "What keeps you awake at night?". That can be either captured by hand or with SenseMaker®. Then bring the results together and present them back to the decision makers as a group. The facilitator needs to encourage dissent and initiate discussion around the list to refine it. Then ask the decision makers to score each agreed-upon item based on its importance to the organization. Use the same scale you used to for the knowledge objects.
Or, you can derive character archetypes from anecdotes related to the issue at hand. Gather a group of people and go through the same process as described for decision makers above, except ask the people in the group to represent the viewpoints of each archetypal character as if it were one of the decision makers. This can also be done with themes or values as well as character archetypes - in that case each "decision maker" is someone who cares primarily about one theme or value.

One possibility is to do both of these processes - with both real and archetypal decision makers - and then either merge or juxtapose the results, for a richer result. At the tops of each column of your grid you should now have the name of a core process or essential function and an "importance" score for it (possibly as a color code).

The columns of the matrix grid will host the knowledge objects which arose from the Knowledge disclosure points and perspective-question answers, namely from Knowledge mapping. These will also have a "score" based on the evaluation criterion for the object, such as its vulnerability to loss (knowledge) or controllability (dynamics).

Before you start calculating your grid, stop and have people agree on a set of criteria that will determine what patterns are strong enough to take action on. Would three medium-to-high values represent a pattern, or would four be required? Make sure this decision is made before you start, because otherwise people may be tempted to change the rules to fit their prior expectations or agendas.

Physical environment

The ideal setting for the Dependency matrix is a standup environment.

Layout

One room with enough wall space one main work areas
One or more support tables

Stationery and tools

Enough butcher paper to cover the width of the necessary wall space
Tape or bluetack to hold the paper on walls
The prepared hexies with knowledge objects (or printed on sticky labels and then applied to them)
The prepared hexies with identified problems (or printed on sticky labels and then applied to them), or pre-printed strips of paper containing the problem statements
Blank hexies of different colors
Marker pens for all participants
A picture-capturing device, such as a camera or mobile phone

Environment setup

Prepare a rectangular grid on butcher paper to allow enough column-space to host all the knowledge objects, and enough row-space to host all problem statements
Place butcher paper on the wall space
Place the hexies with knowledge objects on the grid as column-headers
Place the problem statements hexies or paper strips on the grid as row-headers
Place the marker pens, and the blank hexies on one or more tables

Virtual environment

Whiteboard app that allows you to replicate the respective setup as in a physical environment
The prepared hexies
Blank hexies

Also see Virtual facilitation.

Workflow

STAGE INSTRUCTION	COMMENTARY & TIPS
Ensure that the groups have the required inputs to fill in the axis.
Produce a large worksheet and let the groups build their matrix. You should ensure that both the axis are populated with the input material, so that a complete picture can be produced.
The group should now assess each intersection in the matrix so that a quantitative value is given.	The use of a “RAG”, “HML” or 1-3 score works best for this assessment. This step can produce very good debates on the subjects and you should provide a means of capturing any key issues or themes that emerge from them. Theses will help with any intervention planning.
Now combine the values in the grid.	If you used a numerical scale, you can multiply the numbers at each intersection to achieve a relative combined "score". For example, say you are looking at a "cell" of the grid in which the column (essential function) "ship on time" has an importance value of 5 and the row (dynamics object) "relationships with suppliers" has a controllability value of 3. The combined value would then be 15. If you used a qualitative scale, you can create simple "rules" for combination, such as "the combination is the highest of the two scores", thus L+M=M and L+H=H, or "the combination is the rounded average of the two scores converted to numbers (123)", thus L+M=(1+2)/2=1.5=M and L+H=(1+3)/2=2=M. Basically whatever everyone accepts as meaningful is meaningful.
In any case, color-code the cells in the grid based on the range of numbers or qualitative levels they cover. As you look at the grid, patterns will emerge naturally. And they must be allowed to emerge naturally; any pattern that meets the criteria you set up earlier is allowed, which means no one can artificially create or deny one.	Vertical bar patterns are things that are easy to sell because they obviously impact an important function, but are hard to implement because they require intervention in several object areas. For example, there may be many ways to improve on-time shipping, but it may be difficult to tackle all of those ways at once. Horizontal bar patterns are hard to sell because they impact several functions but in minor ways, but are easy to implement because they only require intervention in one object area. For example, adding a supplier extranet might be a low-hanging fruit, but it will have only minor impacts on several essential functions.
The patterns you choose from the grid as most deserving of action will form the basis for your portfolio of projects.
Once you have identified a portfolio of projects which might prove fruitful to approach based on your matrix, you need to evaluate them to determine which you will actually carry out. To do this you can use two methods: a Cynefin® framework and the perspective question used earlier in the mapping.	Ideally you will use both methods and combine the results, but if you can only use one it is better to use the Cynefin® framework method since it has not been used yet in the mapping process and thus provides another perspective.

Do's and Don'ts

The facilitator should not assist with the assessment of the intersections; this is for them to decide on. Avoid judgement throughout the whole process.

This exercise is best done with small groups from the same functional department so that different departmental perspectives can be seen by comparing the output at stage 3. Capture this components output well as this will drive the formal action plan and engagement report.

Virtual running

The method can be run virtually, using virtual collaboration tools.

Method card material

This material will be extracted for the method cards

Possible symbols or illustrations

Front page description

A Dependency Matrix maps dependencies in two ranges (eg. needs and assets).

Back of card summary

A dependency matrix is used for various purposes, For example (i) applied as part of a process including Decision Mapping and ASHEN. Organisation needs are mapped against knowledge assets. The resulting grid is studied for patterns which then generate a portfolio of projects (ii) mapping archetypes against demographics to create personas.

How can it be used?

for diagnosis

for analysis/understanding

for intervention

References

Articles and books

...

Blog posts

Dave Snowden, Of sandbanks and granite cliffs, Cognitive Edge Blog (November 7, 2014), on dependency matrix (2nd of 2 parts)
Dave Snowden, Dependency matrix, Cognitive Edge Blog (November 6, 2014), on dependency matrix (1st of 2 parts)

Cases

Banking sector
Government
...

Other references