RAMSynapse
Log inSign up

Reliability Allocation · Chapter 6

Common Pitfalls

The mistakes seen in practice and the guards against them.

Allocation fails socially more often than it fails mathematically: the arithmetic is too simple to break, but the judgements around it bend under programme pressure. Every entry below is a pattern seen on real programmes; the guards are correspondingly procedural.

PitfallWhat it looks likeThe guard
Equal apportionment that never diesThe day-one placeholder split still governing budgets at CDRFlag equal splits as provisional with an expiry milestone
Budgets where there is no design freedomA COTS unit allocated twice what its vendor has ever demonstratedCap COTS budgets at evidence; escalate the shortfall, don't bury it
Series arithmetic across redundancyRedundant channels summed as if independent series elementsBudget the redundant group as one element; its internals belong to the RBD
Ratings without ground rulesFOO scores drifting toward whoever argued loudestWritten criteria per factor, independent scoring, archived rationale
ARINC over a stale predictionLast year's estimates scaled into this year's contractsRe-run the prediction first; date-stamp the weights
Unit and clock mismatchBudgets in FIT against predictions in f/10⁶ h, or different mission timesOne declaration of unit, clock and mission time, shared with prediction
Structure edits outside the budgetElements added after the split, carrying no allocationConservation check (Σ children = parent) after every structure change
Allocation as decreeBudgets issued without feasibility checks, absorbed as fiction by the teamsEvery budget line held against similar-equipment or vendor evidence
No reserveThe first surprise reopens every sibling's contractual targetAllocate tighter than the requirement; hold the difference as reserve
Target mistaken for achievementAllocated MTBF columns quoted as if they were predictionsKeep the allocated-vs-predicted view current; never report one without the other

Two of these deserve a closing sentence each. The decree failure is the deepest: an allocation that was never negotiated against feasibility does not distribute a requirement; it distributes disbelief, and every team quietly reverts to designing to what it thinks is possible while the budget table becomes ceremonial. And the static-budget disease is the slowest: an allocation is a snapshot of a negotiation, and a snapshot left unrevised while the design, the predictions and the structure all move ends up describing a programme that no longer exists. The fix for both is the same habit: treat the allocation as a living contract, re-opened deliberately, on evidence, at every milestone that changes what anyone knows.


Want to see this on a live system model? Request a walkthrough.