AI Business Reports: Keep Percentage Points and Relative Change Clear

An AI-written report says your booking rate improved by 25%. Another line says bookings fell by 25%. Before assuming that one calculation is wrong, inspect the starting values and the quantities being compared. Both statements can describe the same fictional business period accurately.

For an independent creator or small business owner, the useful question is what changed: the number of people booking, the share of eligible visitors who booked, or that share relative to its previous value? A clear report keeps those measures visible before recommending a response.

Two ways to describe a change in a rate

Eurostat’s percentage-point glossary distinguishes a difference between percentages from a percentage change. Percentage points describe the arithmetic gap between two rates. The House of Commons Library’s guide, published on 17 March 2026, explains percentage change as the difference divided by the starting value, multiplied by 100.

Take an increase from 8% to 10%. The gap is 2 percentage points. Relative to the original 8%, that gap represents a 25% increase: (10 − 8) ÷ 8 × 100. Both descriptions are valid when their units are stated.

An increase of 25 percentage points would mean something much larger: a move from 8% to 33%. A writer who substitutes percentage points for percent has changed the claim. Before approving an AI summary, ask it to show the calculation behind the wording.

A fictional booking report explains the apparent contradiction

Imagine a creator running a short paid workshop. For this exercise, the booking rate means the proportion of eligible visitors who make at least one booking within seven days of their first qualifying visit. Each visitor is counted once in the denominator and at most once in the booking count.

Compare two invented acquisition periods, each with its full seven-day follow-up completed. These are illustrative counts, not actual results or an evaluation of an AI product:

• First period: 1,000 eligible visitors, of whom 80 book. The booking rate is 80 ÷ 1,000 × 100 = 8%.
• Second period: 600 eligible visitors, of whom 60 book. The booking rate is 60 ÷ 600 × 100 = 10%.

The rate rises by 2 percentage points, equivalent to a 25% relative increase. Yet the booking count falls from 80 to 60, a decrease of 20 bookings, or 25% of the original 80. The second period has a higher booking share within a smaller visitor group.

A useful summary therefore gives the owner all three quantities: eligible visitors, bookings and rate. Calling the period simply better or worse would hide which outcome the owner is trying to assess.

State what the comparison can support

In this fictional dataset, there is no evidence about why the visitor count or booking rate changed. The numbers do not identify a campaign effect, explain visitor intentions or establish that a redesigned page caused the increase. The report should describe the observed arithmetic before proposing an investigation.

Suppose the creator asks whether the new rate would produce more bookings from 1,000 visitors. Applying 10% to 1,000 gives 100 bookings, compared with 80 at 8%: an arithmetic difference of 20. That is a conditional calculation using the selected rates, not a forecast that the next 1,000 visitors will behave that way.

This distinction matters when an assistant moves from a report into planning language. Ask it to identify the assumptions behind each scenario. A calculation can be correct while the conditions needed to use it remain untested.

The starting point changes a relative comparison

The direction of the comparison also matters. A rise from 8% to 10% is a 25% relative increase. A fall from 10% to 8% is a 20% relative decrease, because the starting value is now 10. The percentage-point difference has the same magnitude, two points, in either direction.

Keep the baseline next to a relative-change headline. In the fictional report, the phrase 25% higher booking rate needs the original 8% to be interpretable. Giving the reader both 8% and 10% makes the comparison easier to inspect.

If the earlier rate is zero, the usual relative-change formula divides by zero. Report the starting and ending rates and their percentage-point difference instead of inventing a finite percentage increase. Likewise, a rate based on no eligible visitors is unavailable under this example’s definition; it should not be entered as zero.

Give the assistant a brief with named quantities

For a reporting task like this one, I would ask for a short calculation record before the narrative. This is a proposed review routine, rather than a tested intervention or a claim about a particular reporting integration:

• Identify the eligible population, qualifying booking action and observation window.
• Show the numerator and denominator for each rate, with the period labels.
• Calculate the percentage-point difference and label it in full.
• Calculate the relative rate change using the earlier rate as the baseline.
• Report the booking-count change separately and flag unavailable comparisons.

Then ask for a plain-language sentence that preserves the distinction. For our fictional workshop, that sentence can say the booking rate rose from 8% to 10%, while bookings fell from 80 to 60 as the eligible visitor count declined.

Review the sentence against the calculation record. If the assistant changes from visitors to sessions, or from people booking to booking events, stop and establish the intended definition. A label change can alter the question even when the arithmetic still works.

Connect the report to a decision

The creator may decide to investigate the smaller visitor group while retaining the higher rate as an observation worth following. Another question is whether the periods are comparable in audience and measurement. Neither decision requires pretending that one headline captures the whole business.

For follow-up measures, MaryChuks.com’s guide to comparing retention at the same cohort stage adds a useful timing check. A correctly labelled rate still needs comparable observation periods.

Before accepting the next AI-generated growth claim, name the quantity, the baseline and the unit of change. Keep the counts beside the rates. That gives a business owner enough information to distinguish a numerical improvement in one measure from the next question the business needs to answer.

Sources

Official statistical guidance checked on 29 October 2026. All workshop figures are fictional; the calculations were checked. Eurostat’s glossary does not display a publication date.

• Eurostat: Glossary — Percentage point.
• House of Commons Library: How to understand and calculate percentages — 17 March 2026.