ISE548 Chap.7 Event Trees and Expected Loss
Event Trees and Expected Loss
Branch forward from a trigger
Event tree analysis follows an initiating event through successive protection successes and failures. This chapter multiplies probabilities along complete paths, adds mutually exclusive outcomes and checks that terminal probabilities sum to one. Expected loss then supports a transfer or control decision while leaving appetite and dependence visible.
New numerical cases preserve the tutorial's method without copying its proprietary stem.
The tree is conditional on a defined trigger and follows barriers in the order they are demanded. Complements belong at each branch, not across unrelated stages. Terminal outcomes are checked for completeness before consequences are attached.
The final comparison keeps probability, loss, premium, retained exposure and time basis in compatible units.
What this chapter covers
- 01
Initiating events and branches
- 02
Conditional protective layers
- 03
Path multiplication
- 04
Mutually exclusive outcome sums
- 05
Expected loss and insurance
- 06
Dependence and sensitivity
Worked example · free
Price a cooling failure
- 3Find complete failure probability.
- 2Find safe-operation probability.
- 3Calculate and interpret expected loss.
Key terms
- Initiating Event
- The trigger from which an event tree follows protective successes and failures.
- Path Probability
- The product of the conditional probabilities along one complete branch sequence.
- Terminal Outcome
- An end state reached by a complete event-tree path.
- Expected Loss
- The probability-weighted average loss across defined outcomes and a stated time basis.
Event Trees and Expected Loss FAQ
Why do I add some probabilities and multiply others?
Multiply events that must occur along the same complete path. Add different mutually exclusive paths that reach the same outcome. Drawing and labelling the branches before calculating makes the distinction visible.
How do I check an event-tree calculation?
List every terminal path and verify that their probabilities sum to one. Then confirm complements, conditional labels and units. The sum check catches missing branches but does not prove the input estimates are credible.
Does lower expected cost always decide the treatment?
No. Expected value does not capture liquidity, legal duties, concentration or intolerance for catastrophic outcomes. Use it as one decision input and state the factor that could override the mean.
When is a branch probability conditional?
A backup-success probability is conditional when the backup is demanded only after the first protection fails. Evidence collected during ordinary standby may not represent performance under the heat, power loss or maintenance condition that defeated the first layer. Label the condition on the branch and use data from comparable demands where possible. Otherwise show a range and test sensitivity.
How do shared causes affect an event tree?
Protections that share power, location, software, maintenance or environment may fail together. The later branch probability should reflect operation after the earlier failure rather than assume independence automatically. If the shared cause creates several simultaneous routes, a fault tree or network model may represent the system more clearly. State the modelling boundary instead of hiding dependence inside a neat product.
How should insurance be compared with expected loss?
Add the premium to the probability-weighted retained loss after deductibles, limits and exclusions. Keep operational delay, customer impact and duties outside cover visible. Compare both options over the same initiating-event or annual horizon. Insurance may improve liquidity or appetite even when its expected cost is higher, but the answer should state that management reason separately from the arithmetic.
What makes a sensitivity result useful?
Vary the uncertain branch or consequence and identify the value where the preferred treatment changes. That breakpoint tells the team which estimate is decision-critical. If all plausible values support the same action, the decision is robust despite uncertainty. If a small change reverses it, collect evidence aimed at that parameter and avoid presenting one point estimate as settled.
What checks should be completed before interpreting an event tree?
Confirm that the initiating event and time basis are defined, every branch has its complement, later probabilities are conditional on reaching them and terminal paths sum to one. Attach consequences only after the paths are complete. Then examine common causes, units and sensitivity. These checks separate a drawing or arithmetic error from a genuine decision about treatment, transfer or appetite.
Exam move
Redraw a two-layer event tree with new probabilities and label every complement. Calculate each terminal path, verify the sum and attach a consequence. Then change one branch probability through sensitivity analysis and identify the value at which your treatment choice changes.
Working through Event Trees and Expected Loss in ISE548? Sia is AskSia’s AI Management tutor — ask any ISE548 Event Trees and Expected Loss question and get a clear, step-by-step explanation grounded in how ISE548 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.