Build a traceable analysis workbook

An analysis workbook is a reasoning tool, not a pile of copied numbers. Separate reported inputs, unit conversions and calculations, and conclusions into distinct sheets. For each input store entity, period, currency, display unit, page, source type and version date. Preserve the value as published beside the formula that converts it to a common scale: 82,000 thousand becomes 82 million after division by 1,000. Without this separation a silent edit to a source number can change many ratios unseen. Locking or visually marking source cells helps, but it does not replace an independent calculation check.

Define variables and units before formulas. current_assets_2025_m and current_liabilities_2025_m explain more than B17 and C31. Never divide a thousands cell by a millions cell. Show a ratio formula explicitly, such as current_ratio = current_assets_m / current_liabilities_m, and check that the denominator is not zero and both terms describe the same period and perimeter. Once numerator and denominator share currency and scale, the ratio is dimensionless, but 2.4 times and 240 percent are different display conventions; label the one you use. Retain full precision in calculations and round only the published display.

Place tie checks beside the data, not only in a separate report. balance_check equals assets less liabilities less equity. A cash tie compares closing cash in the position and cash flow statements after checking the definitions of cash and cash equivalents and any foreign-currency effect. Totals can balance despite offsetting classification errors, so add semantic checks: is the performance interval twelve months, are all columns consolidated, and does a margin pair income and expense from the same perimeter? A zero check proves an arithmetic relationship only. The IFRS Conceptual Framework describes indirect verification through checking inputs and recalculating outputs with the same method; it does not certify the original estimates.

Traceability has two levels. At cell level save the table heading or note number and document link. At claim level connect a statement such as margin improved to its rows, formula and comparison period; connect a proposed cause to a disclosure or label it a hypothesis. A 30 percent margin can be calculated correctly yet change meaning after distribution costs are reclassified, so inspect policy notes. Do not turn missing data into zero. Zero is an economic value; missing is a data state. The formula should fail visibly or return unavailable when an input is absent rather than print a reassuring percentage.

Keep an adjustment log with row number, reported value, adjustment, rationale, evidence location, editor and version date. The aim is to let a colleague reproduce or challenge the result, not to manufacture a claim of external audit. When changing an assumption, preserve both the base result and sensitivity result instead of overwriting one. For example, year-end inventory and a quarterly average can produce different liquidity or turnover interpretations when both are available. Their difference exposes seasonality; each scenario must retain its source or stated assumption.

Before issuing a summary, run a small countertest. In a test copy, alter one sample input unit from thousands to millions and check that the unit control blocks or flags the result; then restore the source. Try a zero denominator and an empty cell to ensure neither yields a misleading ratio. Recalculate one sample formula outside the workbook. If a saved file alone cannot reproduce the result, the analysis is not ready for internal handoff. These are specialist work controls, not a claim that an external accountant reviewed the model.