Modelling and Management
In short Whatever you understand or change, you understand or change through a model. This page covers two ways to build models, what makes a model adequate, and seven types of control that tell you how to steer a system depending on what stands between you and the goal.
Based on: F. P. Tarasenko, Applied Systems Analysis (in Russian), ch. 3–4, drawing on R. L. Ackoff.
On this page
Key ideas#
Two ways to build a model#
- Analysis splits the whole into parts, explains each, then combines the explanations. It answers “how is it built?” and yields composition, structure and black-box models. Its trap: cutting the links between parts and losing the properties of the whole.
- Synthesis goes outward: find the larger system yours belongs to, study it, and explain your system’s role through its links with the other parts. It answers “what is it for?”, which analysis cannot.
You usually need both.
Kinds of models#
By material, models are abstract (built from language, from everyday speech to mathematics) or real (a photo, an analogue device, or conventional signs such as money and maps that work because users agree on them). By purpose, they are:
- Cognitive — aimed at true knowledge, tolerant of competing views, always being refined.
- Pragmatic — directive: plans, standards, rules, specifications. Reality is made to fit the model, not the reverse.
A model is adequate if it lets you reach the goal, which is not the same as being true. It also has to fit its users: a perfect diagram in a notation nobody reads is useless.
Tarasenko retells Ackoff’s ladder of knowing: data (what?), information (what is it made of?), knowledge (how is it connected?), understanding (why?), wisdom (what for?). Each rung is worth far more than the one below.
Five components of control#
To control anything you need an object (with outputs, inputs you control, and inputs you can only observe); a goal expressed as a desired path over time, not just an end point; a control action; a model of the object on which you search for that action; and someone or something to apply it. Searching for the action on the real system instead of on a model is, in Tarasenko’s words, just poking at it.
Seven types of control#
The right type depends on what blocks the goal. The illustrations below are general business cases.
| Type | Use it when | Business illustration |
|---|---|---|
| Programmed control | The model is adequate; just execute | A routine, well-understood procedure |
| Trial and error | The model is inadequate: try on the model, apply, observe, correct the model, repeat | Piloting a new process in one branch |
| Regulation | Small, slow drift from the goal; adjust parameters, keep structure | Shifting staff hours to match demand |
| Structural control | Drift too large for parameter tweaks; change the structure | Reorganizing a department |
| Control by goals | The goal cannot be reached with available means; replace it with a reachable one | Rescoping a release |
| Control of “large” systems | No time to find the optimum; speed up modelling or take the first satisfactory option | A deadline-driven pricing call |
| Control without a known end goal | The end state is unknown; adopt a guiding view or take small steps toward improvement | Incremental product evolution |
Tarasenko insists on one distinction. A complex system is one you lack information about. A large system is one where you have the information but not the time to use it. Confusing them leads to adding computing power to what is really a knowledge gap.
How to apply it#
- State the goal as a path: where you want to be, and the milestones on the way.
- List what you control, what you only observe, and what you measure.
- Test your model: has it predicted past behavior? If not, plan for trial and error.
- Size the gap: small and slow, regulate; large, change structure; unreachable, renegotiate the goal.
- If time is the bottleneck, say so and agree what “good enough” means.
- Shape the model for its users, not for yourself.
Example#
The Workshop sets a goal: standard repairs back within three working days, improving month by month. Mechanic hours are controllable; incoming bookings are observable but not controllable.
- In normal weeks managers regulate, moving mechanics between shops.
- When the spring rush swamps the shops, regulation is not enough, so they make a structural change: a quick-service bench for small jobs.
- For full rebuilds three days proves unreachable, so the goal changes to ten days for that category (control by goals).
- When online booking launches, nobody’s model of demand is reliable, so they use trial and error in two shops before rolling out.
Pitfalls#
- Experimenting on the live organization with no model → test on a model or a pilot first.
- Calling a system “complex” when it is just “large”, or the reverse → ask whether you lack information or time.
- Judging pragmatic models (specs, standards) by truth → judge them by whether they get you to the goal.
- Analysis without synthesis → also ask what the system is for in the larger whole.
Sources and further reading#
- F. P. Tarasenko. Applied Systems Analysis (in Russian). KnoRus, 2010, ch. 3 (models and modelling) and ch. 4 (control).
- R. L. Ackoff — the hierarchy from data to wisdom, as retold by Tarasenko.