System instability, system risk. (Managing Change).

The cliches abound.

Change is inevitable. The only thing that doesn't change is that everything changes. If it ain't broke, break it. Change lanes before you get run over.

Change is important, but change must be controlled so risk does not escalate to unacceptable levels. In general, more change may mean more opportunity but also more risk.

If you don't know what quantity and intensity of change is reasonable at a given time, and how much will bring the business or department to its knees, then added change can be a very dangerous move.

While it is true that change or evolution in a business or process is important, the other key factors lie in knowing how much to change and when to change, given the current processes, personnel and other factors within the company or department.

Too much change can be just as dangerous as too little. In fact, it may be more dangerous.

Avalanche!

In the late 1990s several factors combined to create the first truly scientific systems to explain complex behaviors. The science, based primarily on work with avalanches and earthquakes, was then able to be adapted to societal events, human endeavors and several other areas of social structures and events such as the stock market, business behaviors, conflicts and war.

Of course, you may ask yourself how can a department or business be in any way like an avalanche?

As it turns Out, avalanches, earthquakes, stock market prices and human behavior follow power laws. Power laws have no inherent scale but work along a path of growing power as frequencies reduce. The big one is rarer than the little one, but anything between can also occur based on multiple factors.

The power law says that weaker events are more common; stronger (and more dangerous) events are rarer along a power curve. The trouble is, you can never know when a single event will occur, nor predict accurately its magnitude. This is true for avalanches and for human activities.

However, there are factors that raise risk levels for the larger, stronger events. When a crisis does hit, chances are that there was some prior buildup of lot of smaller bad risks to a point where a minor added event could tip things over and send the avalanche of problems crashing down.

The most easily explained base model relates the so-called sand pile model. The sand pile model relates to dropping a grain of sand (one grain at a time) from one position above a flat surface. Eventually, a pile is produced. In time, once enough grains are accumulated, the peak of the pile will gain enough steepness where, given one added grain of sand, it may avalanche.

However, the avalanche itself is unpredictable either by which exact sand grain will initiate it or how intense it will be. As it turns out, the sand actually self-organizes to the point where one grain's slippage then cascades a series of other movements.

The magnitude follows a power law. Sometimes a grain may do nothing, sometimes it may produce a tiny avalanche and occasionally a major avalanche may occur from a grain.

The studies found was that any given grain will produce some amount of pressure on the other grains where it lands. Depending on steepness and configuration of the pile, risk is highest when multiple areas are steep (therefore unstable), and the instability is disbursed over many surrounding, interlinked areas.

At this point, a risk of a large avalanche is highest, but again unpredictable by individual grain, the exact location it will start (it will start at the region of weakest stability, but those areas cannot be defined yet by merely looking at the pile) or intensity.

Lesson for business

The lesson for business is that any complex system with multiple weak or vulnerable features is at a high risk of a crisis (avalanche).

The features themselves could initially be business strengths, but the addition of indiscriminant or poorly designed changes (piling sand) could be enough to destabilize a strong area. The exact time of that crisis or the intensity of it remain unpredictable, but the risk is high for some level of crisis at some nearby time.

This means that if multiple changes are inflicted on an otherwise stable system simultaneously, that business system will be more vulnerable to any other added problems.

For instance, if you change equipment, processes, reporting structure, software, staffing ratios, policies and management systems all at once, the probability of an avalanche is much greater than if each business component is changed and stabilized sequentially.

Multiple changes in just one area can lead to the same risks. For example, continuously changing staffing ratios before the staff has settled into a steady workflow can also destabilize a business.

Powers at work

The power function says that small avalanches or quakes are much more likely to occur than big ones.

In research tests, it was found that large catastrophic events are only about 1/4th as likely to occur as moderately catastrophic events, and that moderate crises are only about 1/4th as likely as minor catastrophes, which in turn are only 1/4th as likely as tiny crises. The result is a power relationship, where the more intense the adverse event is, the less likely it is to happen.

The power relationship can be easily seen in the stock market where minor ups and downs are very common and moderate fluctuations less likely. Catastrophic or major adjustments in the market are even rarer, but also occur.

The same applies to business in general. Small crises are not uncommon. Moderate adverse events are less common and severe crises are generally rare. However, there is always uncertainty.

With the Avalanche Uncertainty Principle: Societies, people and sand piles are unpredictable. You can identify the risks they face and get a sense as to when they are at a high level of risk, but you cannot know ahead of time exactly what event will tip the scale, when it will actually tip, nor how big the crash will be.

The lesson for business here is that, when at highest risk, it is much more likely a business will suffer several annoying small to medium problems than a huge one. But the risk of "the big one" is still there-and no one can predict what level of crisis will hit and at when it will occur.

So, a secondary lesson for business is that by reducing active risk, the chance of "the big one" is much smaller. In other words, the more stable the multiple sub-components of a social system are, the less likely it is that the social system will experience much more than occasional small fluctuations in daily events. Things will be generally held at the low end of the power curve.

If, on the other hand, instability is everywhere, a move up the power curve toward major catastrophe is more likely.

Necessary change

For a business, the real trick is to determine what is actually necessary to change or add in a business, and to do so with the least disruption on other components possible.

It has always been my sense that we would be wiser to change one big thing at any given time, test it, polish it, establish it and make sure it is functional and stable before changing another big thing.

Certainly there may be exceptions when a company or department is already tipped into huge problems and quick change means survival, But for the average workings of a business or department it is best not make across-the-board changes of multiple established systems simultaneously, or to change again and again before a steady situation is established.

To do so may bump risk up to dangerous levels of pre-collapse.

Of course, when instability is seen in a system, there may be of some value in creating the "controlled avalanche." It is a planned mini-disaster used to avoid the larger one later, as firefighters do with controlled burns and back-burning.

When a business or department is already unstable and risking major problems, a preemptive strike may actually defuse some risk. Companies have done this by selling off losing subsidiaries. Departments have done it by removing or relocating problem bosses or employees. Some load is taken off the sand pile and things can eventually be stabilized further.

When multiple major changes are mandatory, finding ways to stabilize at each step of the way is also advised. Each point where stability is gained tends to strengthen the system and leave it less vulnerable for the major avalanche. Ignoring these history lessons leaves a business condemned to repeat bad history and to experience the one major catastrophic event that could take years to repair--or put the business out of business altogether.

Strategic planning

Good management of change requires some intelligent strategic planning related to intensity, timing and other compounding events occurring in a department or business. It means knowing the vulnerabilities within the system and anticipating issues. It also means knowing when not to change something because it is at a time and under a situation where the vulnerabilities make it even more risky.

Finding ways to lessen the risk of the big avalanche requires experience. There is no easy formula, nor is there an established set of rules. Perhaps the handiest way to address risk I have yet found comes from game theory and probability forecasting, but even this can't tell you when or if the big one will actually hit.

In the world of game theory and probability forecasting, many options exist to help determine relative risk. The easiest may be a relative risk grid. By determining the best, most likely, and worst-case scenarios, then assigning each a percentage value (how likely one thinks they are to happen given the current environment), you can begin to outline what is likely and how likely it is.

When you get into multiple, stacked risks, the risks start to move into a power curve relationship. The benefit may be that a gain of $30K is cumulative under "Most Likely," but the total risks of the "Worst Case" may climb to a Loss of $75K. The problem is that the complex system will never let you know exactly when you might get tipped into catastrophe. But avalanche and earthquake science says that you are more likely to approach disaster the more instabilities (new changes) you add into the system. In this case, it may be better to manage "A" before tackling "B" instead of doing them all at once.

The higher the instability in a system climbs, the higher the risk that some "big" adverse event will occur.

If my worst-case scenario is 60 percent probable, most likely scenario 30 percent probable and my best scenario 10 percent probable, then I know I'm in trouble.

Assigning percentages should take into account as much fact as possible, but often requires gut feelings. Sometimes a person has to get creative and make best guesses based on stepping back, looking at the entire business or department and picking a probability. With practice that "gut feeling" or intuition can really help a manager.

Down to the very bottom line, the leader must know his or her department or business, the staffs' capabilities and the weaknesses to make determinations about how much change is acceptable. Changing just for the sake of change, or because a leader feels it would be exciting or because the competition just did it, is a wrong approach.

Before change is implemented, there should be a defined goal, clear pathways toward the goal, mileposts along the way and a vulnerability analysis.

The vulnerability analysis must take into account the systems and social impacts of yet another change at this time, if risk is to be at an acceptable level. And finally it must all be weighed in the face of the intensity and instability created by other ongoing or recent changes.

Change may be good, but changing wisely and with foresight about what the impact of the changes will be is far better.

Risk Probability Table:

          Best Scenario  Most       Worst-Case
                         Likely     Scenario

Option A  Make $20K      Make $10K  Lose $2K
Option B  Make $30K      Make $10K  Lose $50K

Example: Probability of Bad Press

RISK OF INSTABILITY  LOW  MODERATE  HIGH

0-30%                .67    .33      0
30-50%               .64    .18     .18
50-70%               .07    .44     .48
70-100%               0     .25     .75

Reference

(1.) Buchanan, M. Ubiquity: the science of history... or why the world is simpler than we think, crown Publishers, N.Y. 2001.

RELATED ARTICLE: Game Theory and Probability Tools.

Game theory relates to the study of actions (and non-actions) and how they impact outcomes. Game theory is the maximization of one's own probability of a positive outcome based on the choices of action or inaction.

Within this theory is a need to optimize choices in order to direct probabilities toward a desired end. A game, then, would be any event or episode where you must employ some sort of strategy to reach a desired outcome.

In effect, every interpersonal communication where one or more people have a certain desire, need or aim constitutes a game. If I want a raise in pay, you are playing the "raise game." If you want a contract with me, you are playing the "contract game."

Of course, we have all heard people make negative comments about playing games in business or with relationships, but from a game theory point of view, every interaction where there may be give-and-take is a game. We all want to maximize our own outcomes, so we choose strategies that we believe are most likely to meet our needs.

Compounding the games are such factors as cooperation or competition. In general, games (interactions) may be cooperative, non-cooperative (competitive), both cooperative and competitive, or independent (unaffected by what others do).

At any time where interests or needs may be in conflict, game theory becomes most critical because it means that each player in the game is trying to maximize his or her outcome. This means a potential of competing strategies and a risk of win or lose.

Game variables

The variables in game theory include information, beliefs and the chosen tactics (within a strategy). In general, the more information related to the game a person has, the more correct their beliefs are and the better their choice of tactics are, the better their outcomes.

The variables are also part of the tool set in a game. The tools are knowledge (information), accurate beliefs, optimal tactics and knowing the probabilities related to the other tools.

Knowing the probabilities requires a strong sense of how certain tactics used in the strategy will impact others, and thereby result in your own preferred outcomes (versus what others may desire).

Knowledge of probabilities can aid a player in devising better tactics and strategies in a game.

For instance, in an extremely simple game like a coin flip, there are certain probabilities linked to outcomes. If A and B flip coins, they have a chance of a match with both heads or both tails, and a chance of mismatch with heads-tails or tails-heads. Therefore, there is an even chance (assuming the coins are 'fair coins and not weighted or double-headed) of a match or a non-match.

Human games get much more complex. The classic example of a human game is The Prisoners' Dilemma. In this game, two crime suspects are put in separate cells. If they both confess, each will get three years in jail. If only one confesses, he will be freed and used to testify against the other who will, in turn, get 10 years in jail. If neither confess, they will both spend one year in jail. This game is easily placed into a matrix as follows:

Probability tables have been around for years but it is also clear that human behavior ant it is not always easy to place probabilities on a simple grid or table.

Still, there are many human events which using probabilities since they can at least give what would otherwise be only subjective chance.

Often determining a best move in a game requires making a best guess as to what the impact of an action might be. With the tools noted above, there is some level of objectivity afforded, even if percentage chance may be partially a guess.

One of the more useful approaches lies in probability tables when making a determination between two choices. This is the case above with The Prisoners' Dilemma. To cooperate risks zero or three years (mean = 1.5 Years). To not cooperate risks one or 10 years (mean = 5.5 years).

If one has no idea how the other person will react, cooperation would seem the best statistical choice for a one-time event. Knowing the other person, and planning on a longer relationship, however, might lead a person to not confess, even though there is higher relative risk if the other person does confess. If the other person holds off as well, both people get one year and are relatively satisfied compared to the three- or 10-year options.

So, while knowing the other person's beliefs and having information is important, knowing the chances/probabilities/odds can raise the likelihood of a better choice when deciding a course of action or inaction.

A common tool in probability table determinations is the scenario versus probability table. In this system, one may be faced with two or more choices, each of which may have a good, neutral or bad outcome. By plotting out the knowledge-based "best guess" for each scenario, it may make it easier to determine what to do.

Knowing that people do not always make the safest choice, at least with the probability approach a manager can better decide what the benefits and risks might be when it is plotted out in table form. This is yet another tool in the manager's risk-reduction toolbox.

             A Confesses   A Does NOT Confess
B Confesses  3 Years Each  A: 10 Years
                           B:  0 Years
B Does NOT   A:  0 Years   1 Year Each
Confess      B: 10 Years

               Best Case  Most Likely  Worst Case

Choice A

Profit         15,200     12,000       10,000
Prob. Outcome  25%        50%          25%

Choice B

Profit         20,000     10%          15,000
Prob. Outcome  50%        5,000        40%

From just a profit point of view, one would be tempted to choose B
because there appears to be a 60 percent chance of more money. But when
one considers the probable outcome times the scenarios, things could
change. Calculating profit times chance, above, one gets:

   Best   Likely  Worst  SUM
                         Probability

A  3,750  6,000   2,500  12,250
B  2,000  7,500   2,000  11,500

In this case, A is actually the statistically more desirable (safer)
choice.

Richard M. Burton, MD, MBA, PhD, FACPE is the senior medical executive of the Cigna Western region, centered in Phoenix, Ariz. He can be reached by e-mail at [e-mail].