Industrial Engineer Office Automation common mistakes
Industrial Engineer Office Automation · Understand the traps and how to avoid them.
When asked to calculate standard time, the candidate times the task once, records the observed time, and reports it as the standard time without applying the performance rating (leveling) factor or the allowance factor. For example, an observed time of 4.0 minutes is reported as the standard time even though the candidate is told the worker was rated at 110% and the allowance is 15%.
Candidates memorize the formula 'standard time = normal time × (1 + allowance)' but treat normal time as equal to observed time. They skip the rating step because it feels subjective, or they confuse normal time (observed × rating) with standard time (normal × allowance factor). Under exam time pressure, the intermediate step is dropped, and the answer looks plausible because it is close to the observed value.
What to do instead
Write the two-stage formula explicitly before substituting numbers: normal time = observed time × performance rating; standard time = normal time × (1 + allowance) for an allowance on normal time, or normal time / (1 − allowance) when the allowance is stated as a percentage of total time. Then label each intermediate value. Practice with mixed rating values (below and above 100%) so the direction of the adjustment becomes intuitive. Check whether the problem states allowance as a fraction of normal time or of total time, since that changes the divisor.
When asked to identify the bottleneck in a process, the candidate picks the station with the longest single task time or the most workers, rather than the station with the lowest capacity (highest cycle time) across the whole line. In a service process, they pick the step with the longest queue rather than the step whose service rate limits throughput.
Bottleneck reasoning requires converting each station's work content into a rate (units per hour) and comparing rates, which is an extra computation step. Candidates anchor on the visually longest bar in a process description or on the step described as 'busy', and they conflate utilization with capacity. They also forget that a station with two parallel workers can have more work content but higher capacity than a single-worker station.
What to do instead
Standardize on a rate-based method: for every station compute capacity = available time / work content per unit (or units per hour), then select the minimum. Express the answer as 'Station X limits output to N units/hour' and verify by computing the line's throughput and the utilization of every other station. Practice problems that include parallel servers and rework loops, where the bottleneck is not the step with the largest raw work content.
When normalizing a relational schema, the candidate stops at 3NF and declares the design complete, missing that a table contains a determinant that is not a candidate key (a BCNF violation), or conversely over-decomposes a table that was already in BCNF. They also misidentify partial dependencies by treating any multi-attribute primary key as automatically creating a 2NF problem.
Candidates learn 1NF/2NF/3NF as a checklist and treat BCNF as an advanced topic to skip. Partial dependency is confused with 'any dependency involving part of the key', when it only matters if a non-prime attribute depends on part of a candidate key. Without listing all candidate keys and all functional dependencies first, the normal-form test has no basis.
What to do instead
Before answering, write out the attribute set, all candidate keys, and every stated functional dependency. Test each FD X → Y: for 2NF, check whether X is a proper subset of a candidate key and Y is non-prime; for 3NF, check whether X is a superkey or Y is prime; for BCNF, check whether X is a superkey. State the violated normal form and the decomposition explicitly. Practice on small schemas where the candidate keys are not obvious (composite keys, overlapping keys).
On workflow and process-modeling questions, the candidate describes the 'to-be' process as an idealized flow with no exception paths, no rework loops, and no handoff delays, then concludes that cycle time improves by the sum of removed task times. They ignore queueing and wait time between steps, which usually dominates the total lead time.
Textbook process maps show value-adding steps, so candidates equate process time with lead time. Little's Law and queueing effects are covered in a different domain, so candidates do not connect them to workflow redesign. The result is an optimistic answer that ignores the largest component of lead time in office processes: waiting between handoffs.
What to do instead
For every process question, separate processing time from wait/queue time and compute lead time as the sum of both. Apply Little's Law (WIP = throughput × lead time) to check consistency. When proposing a redesign, quantify the effect on both processing time and wait time, and state assumptions about arrival variability and utilization. Practice with a process containing an approval loop to see how rework inflates effective work content.
When a question asks about LAN/WAN design, subnetting, or bandwidth planning for an office, the candidate answers with the nominal link speed (for example, '1 Gbps') and ignores effective throughput after protocol overhead, contention, and the number of users sharing the link. They also misapply subnet masks, treating a /24 as providing 256 usable host addresses rather than 254.
Candidates memorize link speeds and classful defaults but do not practice host-count arithmetic. The '256 minus two' rule is known but applied inconsistently when the prefix is not /24. Shared-medium contention and overhead are abstract, so they are dropped from capacity estimates.
What to do instead
For every capacity question, compute usable hosts as 2^(32 − prefix) − 2 and state the network and broadcast addresses. For throughput, apply an explicit overhead assumption (for example, 70–80% of line rate for effective TCP throughput) and divide by the number of concurrent users when the question describes sharing. Practice prefixes from /20 to /30 until the arithmetic is automatic.
On project-management questions, the candidate computes the critical path but then reports the project duration as the sum of all activity durations, or fails to recompute the critical path after a change in an activity duration. They also confuse total float with free float and treat any activity with zero float as automatically on the critical path without checking the network logic.
Candidates perform the forward pass but skip the backward pass, so they never derive float. Total float and free float are defined similarly in memory and get swapped. When a question says 'activity X is delayed by 3 days', the candidate adds 3 days to the project duration without checking whether X had float to absorb it.
What to do instead
Always do both passes: forward pass for early start/finish, backward pass for late start/finish. Compute total float = LS − ES (or LF − EF) and free float = ES of successor − EF of the activity. When a duration changes, recompute the path lengths rather than adjusting the old answer. For delay questions, compare the delay to the activity's total float before concluding the project end date moves.
On information-security questions, the candidate selects the control that is technically strongest (for example, full-disk encryption or a firewall rule) when the scenario calls for an administrative or procedural control (for example, segregation of duties or a periodic access review). They also answer 'encrypt it' for any data-protection question regardless of whether the risk is unauthorized access, loss of availability, or integrity.
Candidates study controls as a list and do not map control type (preventive, detective, corrective; administrative, technical, physical) to the specific risk described. Encryption is memorable and feels like a universal answer, so it crowds out controls that address the actual threat. The distinction between confidentiality, integrity, and availability is known in the abstract but not applied to the scenario.
What to do instead
Before choosing a control, name the asset, the threat, and the CIA property at risk, then select a control type that addresses that property. Check whether the question asks for prevention, detection, or correction, and whether the control must be administrative, technical, or physical. Reject answers that address a different CIA property than the one described, even if they are technically sound.
When interpreting a quantitative result, the candidate reports a point estimate (for example, a forecast of 1,200 units or a cost estimate of $50,000) without any measure of uncertainty, or treats a correlation coefficient as evidence of causation when asked to recommend an action.
Forecasting and regression questions usually supply enough data to compute a point estimate, and candidates stop there because the confidence interval or standard error requires an extra step. Correlation-as-causation is a familiar trap but is missed when the scenario is phrased as a business recommendation rather than a statistics question.
What to do instead
For any estimate, state the point value, the interval (using the standard error or a stated confidence level), and the decision implication of the interval's width. For regression, report R² and the coefficient sign, and phrase conclusions as 'associated with' unless the question describes a controlled experiment. Practice writing one-sentence recommendations that include both the estimate and its uncertainty.
On spreadsheet and office-productivity questions, the candidate describes what a function does in general terms ('VLOOKUP looks up a value') instead of the specific behavior the question tests, such as whether the lookup is exact or approximate, what happens when the lookup column is not the first column, or whether a relative reference will shift when copied.
Candidates learn function names and rough purposes but not argument order, match-type behavior, or reference anchoring. Absolute versus relative references are understood conceptually but not applied when reading a formula. The result is an answer that sounds correct but misses the tested detail.
What to do instead
Memorize the argument order and default behavior for the functions the exam covers: VLOOKUP(lookup_value, table_array, col_index, range_lookup) with range_lookup defaulting to approximate; INDEX/MATCH as the left-lookup alternative; and the $ anchoring rules for absolute and mixed references. When reading a formula, trace which references change on copy. Practice by predicting the output of a formula before checking it.
On ergonomics and methods-study questions, the candidate recommends a change based on general comfort advice ('adjust the chair', 'take breaks') without tying it to a specific principle such as anthropometric percentile selection, NIOSH lifting-equation components, or RULA/REBA posture scoring. They also confuse the purpose of a time-study observation with that of a methods-improvement study.
Ergonomics content is often studied as a list of guidelines rather than as quantitative tools. Candidates do not practice applying the lifting equation or posture-scoring thresholds, so they fall back on generic advice. Methods study and work measurement are taught together, and the candidate does not separate 'should this task be done this way' from 'how long does this task take'.
What to do instead
For ergonomics questions, identify the specific tool implied by the scenario: NIOSH for manual lifting (compute the recommended weight limit and lifting index), RULA/REBA for posture, anthropometric percentiles for workstation dimensions. State the principle and the numeric threshold. For methods questions, first ask whether the task should be eliminated, simplified, or combined before measuring time; for time-study questions, confirm the method is standardized before timing.