Exit and Stay Conversation Assessment for Managers and HR TeamsThe same decision, before and after the resignation. What the timing costs you.

Methodology: Thirty-eight original exercises across five formats: sixteen single-choice situations written as eight matched pairs, six select-every-that-applies exercises, six binary claims about the turnover evidence, five matching exercises and five ordering exercises. One response instruction is declared for the whole instrument and it is a BEHAVIOURAL TENDENCY instruction: every situation asks what the respondent is most likely to do, never what should be done, because the thing being measured is the judgement a manager actually exercises when somebody is thinking of leaving or has already said so. Construct statement: it measures a manager's judgement in stay and exit conversations - reading an early signal, opening the conversation, knowing what is actually theirs to promise, closing with an owned commitment, and turning what was said into something the organisation can act on - and specifically what the timing of the conversation costs that judgement. It does not measure whether any particular person will leave, whether the respondent's own team is at risk, the respondent's warmth, empathy, or general management ability, and it is not a prediction about any named individual. Scoring is a matched-trial within-person contrast. Eight underlying decisions are each put twice: once before anybody has resigned, when the signal is cheap to act on, and once after the resignation is in, when the same structural decision arrives at its most expensive. The two sets are matched on the decision, on the four moves offered, on the score gradient across those moves, on the share of items where the strongest move is to hold rather than act (three of eight in each set), on reading load and on option length; they are interleaved so that no reader sees the pairing. The headline figure is the difference between graded quality on the stay set and graded quality on the exit set, taken on the raw scale and never as a difference between two chance-corrected figures, because the two sets have different declared nulls by construction and that difference is itself a finding. It is printed beside the cost that a respondent answering at the declared answer priors would show across the same two sets, and the reader's own figure is read against that line. Overall quality across all thirty-eight exercises is reported separately, chance-corrected against a declared per-option answer prior authored on every keyed exercise, and is never folded into the cost. Every figure carries its standard error band at 68 and 95 per cent; the five competencies each carry a three-way placement rather than a number, because their assumed omega from six to nine items sits below .70, and that refusal is printed. An unanswered exercise leaves the denominator rather than scoring zero, an empty sitting scores exactly zero, and a matched pair with one half unanswered leaves the cost entirely. Counter-offers are never keyed as the strong move by default; the exercises say what the decision turns on. Keying sources: Lee and Mitchell (1994), the unfolding model of voluntary turnover; Holtom, Mitchell, Lee and Inderrieden (2005) on shocks as causes of turnover; Mitchell, Holtom, Lee, Sablynski and Erez (2001) on job embeddedness; Griffeth, Hom and Gaertner (2000) and Rubenstein, Eberly, Lee and Mitchell (2018), the meta-analyses of turnover antecedents; Hom, Lee, Mitchell and Griffeth (2017), one hundred years of turnover research; Allen, Bryant and Vardaman (2010) on retention misconceptions; Rousseau (1995) on psychological contracts and Robinson and Rousseau (1994) and Zhao, Wayne, Glibkowski and Bravo (2007) on breach; Klotz and Bolino (2016) on resignation styles; Spain and Groysberg (2016) on making exit interviews count; Finnegan (2012) on stay interviews; Hanisch and Hulin (1991) on withdrawal behaviours; Gollwitzer and Sheeran (2006) on implementation intentions; and McDaniel, Hartman, Whetzel and Grubb (2007) on behavioural-tendency response instructions. Every exercise is an original work written for this instrument. It reproduces no item from any published or commercial instrument, and it is not affiliated with, endorsed by or derived from any assessment vendor, certification body or named retention programme.