FPAC common mistakes
FPAC · Understand the traps and how to avoid them.
Treating the cash flow statement as a plug that automatically reconciles, then submitting an answer where net change in cash does not tie to the opening and closing balance sheet cash figures.
Candidates learn the three statements separately and never practise the articulation between them; under time pressure they compute operating cash flow and stop, forgetting the investing/financing legs and the FX effect on cash.
What to do instead
Before finalising any statement question, run a three-line reconciliation: opening cash + CFO + CFI + CFF + FX effect = closing cash, and separately confirm closing cash matches the balance sheet. Build this as a fixed checklist step you perform on every practice case, not just the ones that ask for it.
Building a forecast model where the balance sheet does not balance, then forcing the difference into a single unexplained line rather than locating the driver of the imbalance.
Candidates link the income statement to retained earnings but forget that working capital, capex, and debt movements each need a matching balance sheet entry; the plug hides the error and the model appears to work.
What to do instead
Insert a live balance check cell that flags any nonzero difference, and debug by isolating one schedule at a time (working capital, PP&E, debt) with the check visible. Practise rebuilding a small three-statement model from a blank sheet until the check stays at zero through a full forecast period.
Computing ratios and metrics using inconsistent definitions across periods, for example switching between average and closing balance sheet figures, or between calendar and fiscal periods, without noting the change.
Candidates pull figures from different source tables and do not record which convention each came from; the resulting trend looks plausible so the inconsistency goes undetected.
What to do instead
Write the definition of every ratio you use at the top of your workings, including the balance sheet convention and period basis, and apply it mechanically. When a source forces a different basis, restate the prior period on the same basis before comparing, and state the restatement explicitly.
Projecting revenue or margins by extending the historical growth rate without testing the underlying drivers against capacity, pricing, or unit economics.
Straight-line extrapolation is fast and produces tidy numbers, so candidates under time pressure skip the driver build and never sanity-check whether the implied volume or price is achievable.
What to do instead
Always decompose the forecast into at least volume and price (or units and revenue per unit) and test each against a stated assumption. Add a reverse check: compute the implied market share or unit growth your forecast requires and ask whether it is credible given the historical range.
Valuing a business with a discount rate or terminal growth assumption that is internally inconsistent, such as terminal growth above the long-run nominal GDP growth rate used to justify the discount rate.
Candidates treat WACC and terminal growth as independent inputs to memorise rather than as linked assumptions; the arithmetic still produces a number, so nothing signals the inconsistency.
What to do instead
State the inflation and real growth embedded in your terminal growth rate and confirm it sits below the long-run nominal growth of the economy the company operates in. Cross-check the implied terminal exit multiple against comparable transactions and flag any result that exceeds a plausible trading range.
Presenting a data analysis conclusion drawn from a dataset without checking completeness, duplicates, or period alignment first, so the headline number is computed on a biased sample.
Candidates move straight to the analytical technique because that is the part they studied; data preparation feels like administration and gets compressed under exam time pressure.
What to do instead
Run a fixed five-minute profiling pass before any analysis: row count versus source, null rate per key field, duplicate check on the primary key, date range coverage, and outlier scan. Report the profile alongside your conclusion so the reader can see the basis of the sample.
Confusing correlation with causation when interpreting a regression or a dashboard trend, and recommending an action based on the association alone.
Statistical output presents coefficients and significance levels that look authoritative, and candidates equate a low p-value with proof of a causal mechanism.
What to do instead
For every relationship you report, name at least one plausible confounder and state whether the design controls for it. Use language that matches the evidence (associated with, predicts) and reserve causal claims for designs that support them, such as controlled experiments or natural experiments.
Selecting or tuning a model on the same data used to evaluate it, producing performance metrics that do not survive out-of-sample.
Candidates learn the modelling step in isolation and treat validation as a formality; without a held-out set the reported accuracy always looks good.
What to do instead
Split the data before any feature selection or tuning, keep the test set untouched until the final evaluation, and use cross-validation only within the training portion. Report both training and out-of-sample metrics side by side so any gap is visible.
Automating a reconciliation or reporting workflow without building in exception handling, so the process silently produces blanks or zeros when an input format changes.
Candidates test automation on the happy-path dataset they built it from and never simulate missing files, renamed columns, or changed date formats.
What to do instead
Before signing off, run the workflow against deliberately broken inputs: a missing file, a renamed column, a duplicated row, and a null in a key field. Confirm each produces a visible error or exception report rather than a silent default, and document the expected input schema.
Describing a data model or analysis in a memo without stating the assumptions, source lineage, and limitations, leaving the reader unable to judge reliability.
Candidates focus on the result and treat documentation as padding; in a timed written answer, narrative space feels better spent on conclusions.
What to do instead
Adopt a fixed closing block for every analytical memo: data sources and extraction date, key assumptions, known limitations, and what would change the conclusion. Draft this block first so the analysis is built to answer it, then write the findings around it.