The Stay Interview That Won't Fix the Real Problem
Meridian Software, a 240-person B2B SaaS company, lost 22% of its 70 engineers last year against a 12% industry benchmark. Of 15 exit interviews, 11 cited 'no growth path' and 'my manager doesn't advocate for me.' Engagement survey response was 61%, with the lowest scores on career development (2.9/5) and recognition (3.1/5). Base pay sits at the 50th percentile; the board approved a 3% merit pool. Engineering managers average 13 direct reports, and six of nine were promoted from senior engineer within the past 18 months with no management training. The CHRO has one HR business partner supporting all of engineering and a $180K discretionary budget. The CEO has asked HR to launch a stay-interview program next quarter and report back on retention improvement in six months.
1. Based on the evidence in the scenario, what is the most likely primary driver of engineering turnover? Explain why the CEO's proposed stay-interview program is unlikely to move retention on its own.
Read the worked answer
Diagnosis. The convergent signals point to manager quality and career progression, not pay. Pay is at the 50th percentile, and no exit interview named compensation. The two lowest engagement items (career development, recognition) are the same themes named in 11 of 15 exits. The structural facts explain the survey scores: managers with 13 direct reports have no time for individual advocacy, and managers promoted 12–18 months ago with no training are unlikely to be skilled at coaching or sponsorship. This is a capability and span-of-control problem expressing itself as an engagement problem.
Why the CEO's proposal underdelivers. Stay interviews are a data-collection tool, not an intervention. Run by untrained managers with 13 reports, they will either (a) surface problems the company has no capacity to fix, increasing cynicism, or (b) be skipped or performed superficially, generating false reassurance. A six-month measurement window is also too short: the median tenure at exit and the time-to-hire pipeline mean improvement in the annualized turnover rate will not be detectable that quickly. The program also places the burden of retention on the employee ('tell us what would keep you') rather than on the system that is losing them.
Framing for the CEO. Position stay interviews as a diagnostic input inside a broader intervention, with the measurement window reset to 12–18 months using leading indicators in the interim. Refusing the idea outright is unnecessary; resequencing it is the judgment call.
2. You have $180K and one HR business partner. Sequence three interventions over the next 12 months and justify what you are deliberately not doing.
Read the worked answer
1. Fix span of control and add a management layer (months 1–4). Highest leverage, lowest cost. Reorganize 70 engineers into teams of 6–8 under team leads, which converts 9 managers into roughly 10–12 leads with manageable spans. Cost is largely re-grade and title, not new headcount: budget roughly $40–60K for re-grades and for promoting two senior engineers into lead roles. Rationale: without this, no amount of training or recognition has anywhere to land, because managers physically cannot do individual advocacy at 13:1.
2. Manager capability program targeted at the six newly promoted managers (months 2–9). Not a generic leadership curriculum. Build six 90-minute modules on one-on-ones, career conversations, giving recognition, and escalation - delivered by the HRBP plus an external facilitator. Budget $25–35K. Tie completion to a visible expectation from the CTO so it is not optional.
3. A defined individual-contributor career ladder plus a lightweight recognition mechanism (months 4–10). Publish explicit criteria for Engineer / Senior / Staff with a calibration process, so 'no growth path' has a concrete answer. Pair with a low-cost peer recognition mechanism tied to the ladder's competencies. Budget $20–30K in facilitation time; the 3% merit pool stays reserved for market and promotion adjustments rather than being spread thinly.
What I am not doing, and why. No market-wide pay move - the data does not support it, and $180K would be consumed without touching the stated driver. No company-wide engagement campaign - response rate of 61% signals survey fatigue already. No stay-interview program in the near term, except a pilot with two reorganized teams in month 6 to test the diagnostic. Reserve roughly $40K for off-cycle retention of flight-risk engineers identified during career conversations; hold it centrally so managers cannot spend it as an undisciplined counteroffer fund.
3. What leading and lagging metrics would you track, and what evidence would cause you to revise the diagnosis?
Read the worked answer
Lagging (12–24 months): annualized voluntary turnover among engineers, regretted turnover specifically, and 12-month retention of engineers who received a promotion or re-grade.
Leading (quarterly): average span of control; percentage of engineers with a documented career conversation in the last 90 days; internal mobility rate (internal fills as a share of senior openings); engagement survey scores on career development and recognition with a target of raising response rate above 80%; ratio of counteroffers accepted to regretted exits.
Course correction triggers. (1) If career conversations are happening and spans are fixed but exit themes still cluster on growth, the constraint is likely too few senior roles - the answer becomes creating a technical track or accepting managed attrition. (2) If engagement rises but regretted turnover does not, look upstream at hiring: are we recruiting engineers whose growth expectations this company structurally cannot meet? (3) If exits begin citing compensation after the ladder is published, the ladder has created visible internal inequity and pay becomes the live issue. (4) If survey response falls further, treat it as a trust signal and switch to a smaller, qualitative listening method rather than re-running the survey louder.
Discipline point. Pre-commit to these triggers now, in writing, so that the six-month check-in is a genuine test of the hypothesis rather than a search for confirming anecdotes.