Task analysis fails by being plausible. Every row has a time in it, the times are not obviously wrong, and nothing in the document says which of them anybody has ever seen performed.
| Pitfall | What it looks like | The guard |
|---|---|---|
| Times estimated by the designer | Every task rounded to the half hour | Estimate by step, from an elemental basis, then validate on the article |
| Never walked on the hardware | An analysis completed before the first article exists | Plan the validation event; the findings pay for it |
| Elapsed time and man-hours conflated | One column called "time" | Two columns, plus the crew profile that connects them |
| Crew size averaged | 1.67 technicians for three hours | Crew per step; the average belongs to no step and no plan |
| Frequency missing | Task times with no rate against them | Every task carries a frequency, from the failure rate or the interval |
| Access and preparation excluded | The task starts at the item and ends at the item | Start at the aircraft as found and end with it serviceable |
| Fault isolation assumed instant | Diagnosis time absent, or the same for every task | Take it from the testability model, ambiguity groups included |
| Support equipment invented | A task that needs a rig nobody has costed | Every tool named, quantified per site, and priced |
| Special tools missed | A task validated with a substitute that damages the item | The validation is where these surface, and the reason to hold it |
| Consumables ignored | Seals, lockwire and lubricant absent from provisioning | They are demands per task, and they are what stops a job at 2 a.m. |
| Skill level left blank | A task any technician can do, on paper | Trade and skill per step; a task nobody at that base can do is not a task |
| Hazards recorded as a warning | "Caution: hot surfaces" and no cooling time | If it changes the task time or the crew, it belongs in the steps |
| Prerequisites unrecorded | Three tasks planned in parallel on one bleed system | Concurrency constraints are part of the analysis, not of scheduling |
| Preventive tasks left out | Only corrective work analysed | On this example preventive work is 69 per cent of the man-hours |
| Ranked by duration | The longest task treated as the biggest problem | Rank by frequency × duration; a short frequent task usually wins |
| One level analysed | Only the flight-line task, no workshop or depot task | Level of repair cannot be decided without the other levels |
| The analysis never updated | Times from the prototype, in service five years later | Field data corrects task times exactly as it corrects failure rates |
| Written as prose | A paragraph per task, no data | Structured fields, or none of the downstream products can be built |
Four are worth expanding.
Not walking the task is the pitfall that produces all the others. Times estimated in an office are optimistic in a predictable direction, and the errors are not random: access is worse than the drawing suggests, connectors are unreachable with the fixture fitted, and the special tool that makes the job feasible has not been designed. A validation event on the first article finds these in an afternoon, and every one of them is a change to a document somebody else is already building from.
Elapsed time and man-hours answer different questions. The aircraft cares about the first, the manpower plan about the second, and the support cost model about man-hours per operating hour. A single column called "time" silently answers one of the three and misleads the other two, and there is no way to recover the missing information afterwards without redoing the analysis.
The frequency column is what makes it engineering. A task time is a fact about a procedure; a workload is a fact about a fleet. On the worked example the biggest single consumer of manpower is a 1.2 man-hour filter servicing, because it happens 86 times a year, and it outranks every failure in the system. No amount of attention to the long tasks would have found that.
Preventive work is usually the majority and is usually analysed last. It is scheduled, so it feels like planning rather than engineering, and it lands in the analysis after the corrective tasks are done and the budget is spent. On this example it is 69 per cent of the man-hours, and the interval that drives it is a number somebody chose, which means it is a number somebody can revisit.