Applied Systems Analysis in 12 Stages

Key ideas#

The stages at a glance#

Tarasenko’s technology has 12 stages:

1. Record the problem.
2. Diagnose it.
3. List the stakeholders.
4. Uncover the “problem mess”.
5. Define the configurator.
6. Identify goals — see Identifying Goals and Defining Criteria.
7. Define criteria (same page).
8. Study the system by experiment — see Measurement Scales.
9. Build and improve models — see Modelling and Management.
10. Generate alternatives — see Methods for Generating Alternatives.
11. Choose — see Choice Theory and Voting Paradoxes.
12. Implement the improving intervention — see Implementing Interventions.

The path is not linear: a gap found late sends you back, and Tarasenko treats the number of such loops as a rough measure of the problem’s difficulty.

Ground rules before you start#

Tarasenko lists three conditions to agree before taking on the work: access to the information the work needs (with the analyst’s commitment to confidentiality), personal involvement of the organizations’ leaders, and no demand for a fixed specification up front, since the result takes shape during the analysis. Without them, he advises declining the engagement.

Stages 1–2: record, then diagnose#

Record the problem in the client’s own words, without judging it, and do not start solving it yet. The client’s picture of the situation is probably inadequate (otherwise they would have solved it), a symptom is often mistaken for the problem, and sometimes the problem to solve belongs to someone else. In a classic case Tarasenko cites, complaints about slow lifts were settled with mirrors in the lobby rather than new lifts.

Diagnosis decides the type of problem: change how people perceive the situation, or change the situation itself. Tarasenko calls it more art than science and borrows J. S. Mill’s advice: look for what every failure shares and what never appears in a success.

Stage 3: stakeholders#

For a complete list, check four groups: customers and users of what the situation produces; performers who do the work; host organizations (the one containing the problem and the one solving it); and owners of the material resources involved. Following Ackoff, add three often-forgotten “silent” stakeholders: future generations, past generations (represented by the culture they left), and the natural environment.

Leaders of the problem-containing and problem-solving organizations are mandatory participants; others are desirable. If someone cannot be reached, find the closest accessible representative through a chain of contacts; voicing their view yourself is the last resort.

Stage 4: the problem mess#

Ackoff’s “mess” is the whole set of stakeholders’ subjective views of the situation. The client’s problem is only its core and cannot be fixed in isolation. Tools Tarasenko describes for surfacing it:

• Nominal Group Technique: each person writes problems down alone, the statements are posted, then the group discusses them.
• Metaphors of organization, after M. Jackson: view it as a machine, an organism, a brain, a culture, a political system. Each metaphor reveals different problems.
• Ackoff’s analysis: describe how the system works now, what obstructs change, and what happens if nothing changes. The last scenario often shows the organization undermining itself.

Then structure the mess: a problem tree (causes below, consequences above), or Warfield’s Interpretive Structural Modeling, a directed graph built from the group’s pairwise judgments of whether one problem aggravates another. Ackoff’s alternative is to go straight to idealized design.

Stage 5: the configurator#

The configurator is the smallest set of professional “languages” (legal, financial, technical and so on) enough to describe the situation fully. Real problems ignore discipline boundaries: a road accident needs legal, medical, technical and economic descriptions at once. Infer each stakeholder’s languages from how they talked about the mess; their union is the configurator. Languages your team speaks poorly show which experts to bring in.