What does ‘conversion rose 20%’ mean in an AI report?
An AI-written report says your enquiry-to-booking rate rose 20%. Before putting more money into advertising, ask it to show the two rates. In a hypothetical example, 10 bookings from 100 enquiries becomes 12 from 100. That's a 20% relative increase, but a rise of 2 percentage points: two more bookings per hundred enquiries. Both descriptions are correct. They don't leave the same impression when the starting rate is missing. The Office for National Statistics' existing style guide defines a percentage point as the difference between percentages. This isn't a new AI finding; it's a useful distinction to keep when an assistant turns a spreadsheet into prose. I'd ask for ‘10% to 12%, up 2 percentage points’ alongside the actual bookings and enquiries for each period. The same rates could come from very different numbers of enquiries, so the counts belong in the report too. Neither phrasing proves the advert caused the change. That's a reporting rule I'd use, not an error I've observed in a particular tool. It lets the owner judge what the change is worth without reopening the sheet to work out what ‘up 20%’ meant.
Comments
I'd check whether ‘enquiry’ meant the same thing in both periods. If the receptionist used to count every incoming message but now removes spam and repeat contacts before the AI gets the sheet, the booking rate can look better without anyone winning more work. Ask the person maintaining that list what changed, and keep the counting rule beside the report. A cleanup may be worthwhile in its own right; it shouldn't become evidence that the sales team improved. That's a reporting check I'd add, not a mistake I've observed in a particular tool.
Please don't turn the booking rate straight into a receptionist score. If a repair shop can't handle a particular fault, telling someone before they book saves them a wasted visit. I'd want the report to keep ‘we couldn't provide what they needed’ separate from ‘they chose another shop.’ Otherwise helpful refusals can look like poor sales, and staff get pushed to fill appointments that shouldn't happen. That's a distinction I'd ask for before using an AI-written report to judge someone's work.