Total profit can hold steady while the business underneath gets worse. Ratio, trend and common-size analysis look at the shape of the numbers, not just the totals, across several periods.
Where it comes from
Auditors call this analytical review. ACCA's audit material describes analytical procedures as comparing figures with expectations, prior periods and related figures to spot what needs explaining. Financial analysts use the same three moves:
- Common-size: every expense as a percent of sales, so periods of different sizes compare.
- Trend: each figure over several periods, against the same period last year.
- Ratios: gross margin, operating margin, current ratio, days sales outstanding, inventory days, and the few operating ratios that matter to this business.
How a bot helps
A bot can lay out 12 months of statements as common-size tables and ratios in minutes, flag every figure that moved more than a set amount, and line up each flag with what happened that month. That gives a person a short list of "why did this move?" questions.
It should keep definitions fixed. If one period's "cost of sales" includes freight and another's doesn't, the trend is an artifact. The bot should list every definition it used and every place a definition changed.
On the bots' own work
Common-size a bot team's costs: what share goes to each bot, each model, each routine. A routine whose share keeps creeping up is worth a look even if the total is fine.
When to use it
Quarterly, or when something feels off and nobody can say what. Medium: the tables are fast, the questions take people.
Where bots differ
Ratios invite stories. A bot will explain a margin drop fluently with nothing behind it. Each flag should come out as a question with the evidence to pull, not as an answer.
Sources: ACCA: analytical procedures.
Back to Engineering tools for bots.
Short outline
Ratio and trend analysis: short outline
What it's for
Ratio, trend and common-size analysis compare the shape of the financial statements across consistent periods to find what is slipping, and turn each unexpected move into a question with evidence to check.
Columns
| Column | What goes in it |
|---|---|
| Measure | The ratio or common-size line |
| Definition | How it is calculated, and any change in definition across periods |
| Periods | One column per period |
| Expected | What it should be: last year, plan or a stated benchmark |
| Move | Change against expected |
| Flag | Over the set threshold or not |
| Question | What needs explaining |
| Evidence to pull | The records that would answer it |
| Answer and basis | The explanation, labeled measured, inferred or guess |
Steps, each with its "done when"
- Pick the measures and the threshold. Done when the measures, periods and the size of move worth flagging are set before the analysis.
- Fix the definitions. Done when each measure has one definition, and any period where the underlying figures were defined differently is noted.
- Build the tables. Done when every measure has a value for every period, and totals tie back to the statements.
- Flag and ask. Done when every move over the threshold has a question and the evidence that would answer it.
- Answer or hand on. Done when each question is answered with a labeled basis, or has an owner and a date.
Common mistakes
- Trends built on figures whose definition changed partway.
- Comparing a seasonal month with the month before instead of the same month last year.
- Writing a bot's explanation as fact.
- Twenty ratios where five that matter to this business would do.