FPAC Part II common mistakes
FPAC Part II · Understand the traps and how to avoid them.
Building a full three-statement model with circular interest references before checking that the balance sheet balances on a simple version, then spending exam time hunting a plug that appears only after linking debt schedules.
Candidates trained on real modeling work default to integrated builds, but exam items usually test a single linkage; circularity introduces errors that are slow to trace under time pressure.
What to do instead
Build the simplest balancing version first (cash as the plug), confirm assets equal liabilities plus equity, then add debt/interest schedules and iterative calculation only if the question requires it.
Computing ratios from average balances when the question's data only supports ending balances (or vice versa), then selecting the answer choice built on the other convention.
Textbook formulas often specify averages, so candidates apply them reflexively without checking what the stem provides or what the answer options imply.
What to do instead
Before calculating, scan the stem for opening and closing balances; if only one balance is given, use it, and note that distractors are usually built on the mismatched convention.
Treating a lease, pension, or stock-comp disclosure as off-balance-sheet and leaving it out of adjusted leverage and coverage ratios.
Candidates read the primary statements and stop, missing that the question is testing analytical adjustment of obligations disclosed only in the notes.
What to do instead
When a stem mentions leases, pensions, or share-based compensation, default to adjusting: capitalize the obligation, add back the expense where the question asks for adjusted EBITDA, and state the adjustment explicitly.
Mixing sign conventions in cash flow analysis, such as adding a decrease in a liability or subtracting a source of cash, producing a free cash flow figure that is directionally wrong.
Working capital line items flip sign depending on whether the analyst is building the indirect cash flow statement or analyzing sources and uses, and candidates carry the wrong convention across question types.
What to do instead
Write the rule at the top of scratch work (asset increase = cash use; liability increase = cash source) and sanity-check the sign of the final FCF against the direction of the change described.
Answering a credit-analysis question with the profitability trend when the question asks about debt capacity or covenant headroom.
Candidates anchor on the most salient number in the exhibit, which is often the margin or growth line, rather than the metric the covenant is actually written on.
What to do instead
Restate the question in covenant terms before computing (for example, leverage = debt/EBITDA), identify the governing definition, and compute headroom against that definition only.
Describing a data-quality problem as 'the data is dirty' instead of naming the specific defect (duplicates, stale extract, mismatched granularity, survivorship bias) and its effect on the analysis.
Candidates recognize that something is wrong but lack vocabulary to classify it, so they give a generic answer that earns no credit on a scored rubric.
What to do instead
Memorize a short taxonomy (completeness, accuracy, timeliness, consistency, uniqueness) and practice phrasing each defect as 'X is present, which biases metric Y in direction Z.'
Choosing the most sophisticated tool or technique offered (machine learning, real-time streaming) for a problem that a controlled spreadsheet reconciliation would solve.
Candidates equate advanced methods with correct answers, but exam items often reward fitness for purpose, cost, and auditability over sophistication.
What to do instead
Evaluate each option against the actual requirement, data volume, and control needs; prefer the simplest solution that meets the stated need and can be independently verified.
Ignoring the control environment around an automated report, assuming system-generated output is inherently reliable.
Automation feels authoritative, so candidates skip questions of source-system access, parameter changes, and report logic ownership.
What to do instead
For any automated-output scenario, ask who can change the logic, whether the extract is complete, and how the output is reconciled to the ledger; frame the answer around those controls.
Confusing correlation with causation when interpreting a dashboard or regression output, and recommending an action based on the association alone.
Charts and regression summaries invite causal language, and candidates write conclusions the data cannot support.
What to do instead
State the association, then list at least one plausible confounder or reverse-causality path, and phrase the recommendation as a test rather than a conclusion.
Failing to distinguish population, sample, and reporting period when validating an analysis, so a conclusion drawn on a partial extract is presented as complete.
Candidates focus on the arithmetic and skip scoping questions about what was included and over what window.
What to do instead
Before accepting any result, confirm the row count and date range against the source, and state the scope limitation explicitly in the conclusion.