The main post discussed ‘cases’ for risks. There is always a reasonable worst case, and usually some other cases, each milder or more extreme.
The main post included an example of launching a product in a new market, with uncertainty around the average sales volume in the new market. This page gives full details of the product launch example, that is, the set of risk cases and the scale for consequences. It also discusses some aspects of the example in depth.
This product launch example shows a range of cases for the average volume of product sales in a new market. There is a single planned range for the sales volume. Each risk case is a deviation of reality from that planned volume, or if you prefer, an extent of error in the assumed volume. Deviations of each size are assumed to have a known, fixed effect on profitability.
That assumption is unrealistic. It keeps the example simple. The separate page on separating size cases and consequence cases shows a slightly more realistic version of this example, in which there is some uncertainty about the consequences that might follow from each range of deviation in the average sales volume.
|Risk: Average sales volume for the main product is different from expectations.
|Consequence level (for profitability)
|Actual average sales below 50% of break-even volume.
|Very low (indicatively, 5%)
|Reasonable worst case
|Actual average sales are 50%-110% of break-even volume.
|Low (indicatively, 15%)
|Actual average sales 110% to 130% of break-even volume.
|Possible (indicatively, 35%)
|Planned case (not a ‘risk’)
|Actual average sales 130% to 150% of break-even volume.
|Possible (indicatively, 25%)
|Actual average sales 150%-250% of break-even volume. Further sales might be possible if supplies were increased.
|Low (indicatively 13%)
|Actual average sales over 250% of break-even volume. Further sales would be assured if supplies were increased. If supply increases are not possible, there could be price increases to balance demand with supply.
|Very low (indicatively, 2%)
Case likelihoods are shown with an indicative single percentage value. No likelihood percentage value is a numerical fact. Likelihood percentages should never be interpreted as either precise or factual. In the risk table above, the indicative likelihood percentages represent the assessment’s best guess of the case likelihood, taking into account any available data, plausibility, and allowances for uncertainty arising from the lack of data and knowledge. There are other ways of representing likelihoods, and of representing the uncertainty around likelihoods. In practice I prefer to keep a distance between realistic risk assessment and illusory numbers, but likelihoods are unavoidably tied to numbers. My solution is to keep saying this over and over: No likelihood percentage value is a numerical fact. Likelihood percentages should never be interpreted as either precise or factual. Risk decisions always rely on assumptions as well as facts. Precisely calculated probabilities based on extensive factual data have a legitimate specialised place in risk management.
|Consequence type: Profitability
|End-point effect on objectives
|The enterprise becomes insolvent and has debt that cannot be settled in any finite time. Attempting to continue operations will only make things worse.
|The enterprise is making losses that cannot be sustained.
|The enterprise is making a loss. There may be strategic reasons to persist with it for a limited time.
|The enterprise is just breaking even, stakeholders looking for better investment opportunities, or continuing for reasons other than profitability.
|The enterprise makes a reasonable return on investment, justifying continuation of the activity.
|The enterprise makes a high return on investment, generating the potential for expansion.
|At least 20%
|The enterprise makes a very high return on investment, allowing the enterprise to move up into a different league.
|At least 1%
In-depth comments on the product launch example
Sales volume can deviate from expectations for many reasons. The reasons might be understood, or the volume can just be ‘uncertain’. In this example, the enterprise does not have a strong insight into specific drivers of the volume. All sales volume risk is rolled into a single risk definition, without identified causes for the deviations.
Sales volume is a strong influence on profitability, the consequence type used in the example. Other influences on profitability include cost of product, salaries and expenses, interest rates, finance availability, etc. Each of these factors has its own risks. There are also enterprise consequence types other than profitability; it is assumed that sales volume surprises will not lead to those other consequences.
The profitability consequence level associated with each case of sales volume variation has been made up from nothing. Each consequence is assumed to be fixed. In a real enterprise, the links between cases and consequence levels would depend on the cost structure, for instance the gross margin and the mixture of fixed and variable costs. Consequence levels for risk cases would not necessarily match the scale consequence levels in one-for-one steps. During development of the cases, the assessors might start with different volume ranges that cross over consequence levels. They may ultimately refine the case consequence ranges to match the scale consequence levels, as shown in the example.
This example aims to eliminate ‘natural variability’ of any relevant number. There are numbers representing different ranges for sales volume. Those numbers are not meant to represent variability, in the statistical sense. The differing number ranges aim to represent uncertainty about the true average sales volume over future time. That uncertainty comes solely from ignorance about the underlying product demand, and about how that demand will be realised in sales. In statistical speak, the future sales volume is not a random variable (sales in each month), but an unknown parameter (long-term average sales). The unknown parameter has been estimated subjectively from evidence and plausibility available before any trial, like a Bayesian ‘prior’.