In a behavioral story's result, when is a percentage stronger than an absolute number?

## The denominator problem Every percentage in an interview answer is a fraction with the bottom half deleted. That is exactly why weak candidates reach for it: 'a large reduction in false positives' survives scrutiny far longer as a percentage than as 'three alerts became one'. It is also why an experienced behavioral interviewer's reflex on hearing any percentage is to ask what it was taken from. If the candidate cannot answer, the number is discounted entirely and the answer takes damage beyond the number itself. The symmetric trap runs the other way. An absolute deletes the scale. 'About 7 engineer-hours a month' is impressive only against a prior of roughly 30 to 40 hours a month; against a prior of four hundred it is barely visible. Neither form is honest alone — they conceal different halves of the same fact. ## The default: give both The strongest spoken result names the pair and then, if it helps, the ratio: '214 open findings down to 38 — a bit over four-fifths of the queue.' The ratio does the emotional work, the pair does the evidential work, and together they leave nothing for a probing question to expose. Delivered in one sentence it costs about three extra seconds of airtime. ## When the absolute alone is right When the base is small. A percentage taken over a tiny denominator is technically true and rhetorically dishonest, and interviewers reliably spot it. Three of five is not a story about a large reduction; it is a story about five things. Say the small numbers plainly. Candidates routinely underestimate how well a modest, precisely stated result lands compared with an inflated ratio that collapses on the first probe. Also when the absolute is intrinsically meaningful: engineer-hours, days of exposure, count of services. 'The median age of an unresolved high-severity finding went from 19 days to 6' needs no ratio at all — everyone in the room knows what a day is. ## When the ratio alone is right When the raw base is not yours to share. Many employers treat volumes, incident counts and revenue as internal, and a candidate who recites them signals they will do the same with the next employer's figures. The graceful move is to give the shape and say why: 'I can give you the relative move but not the raw volume — that one was treated as internal.' That sentence costs nothing and reads as professional discretion rather than evasion. Ratios are also right when the absolute is so large it becomes noise, or when the comparison across differently sized units is the actual point — for example the same automation covering 9 of 23 services. ## Rates need a per-what A third form appears constantly and is mishandled just as often: the rate. 'Alerts dropped' is not a rate. 'Alerts needing a human look dropped from about 41 a week to 9' is. Any rate must carry both the unit and the interval, and the interval must be the one you actually measured over rather than the one that makes the arithmetic pleasant. ## The tidiness tell Round ratios read as invented, because measured ones almost never come out round. A candidate who says the queue was cut in half is asking to be probed; a candidate who says 214 became 38 has already shown the measurement happened. The failure mode this whole topic guards against is attaching a clean percentage to a result whose measurement was never taken — and the tidiness of the number is the first thing that gives it away. ## A quick decision table | situation | say | |---|---| | large base you can state | both: pair plus ratio | | small base | the absolute pair only | | base treated as internal | the ratio, and say why the base is withheld | | intrinsically meaningful unit (days, hours) | the absolute pair | | never measured at all | a labelled estimate as a range, not a percentage | ## The self-test Before any story goes into your preparation, take every percentage in it and try to write the two numbers it came from. Any percentage that fails that test either gets replaced by the honest absolute or gets dropped. That single pass removes most of the material a probing interviewer would otherwise dismantle.