Milestone map
Milestone map
3 milestones
Baseline Voluntary Turnover Rate and Diagnose Retention Risk Factors
2–4 weeks for data collection and diagnosis
Retention improvement starts with understanding why people are leaving — and why those who stayed chose to stay. Exit interview data, stay interviews, and engagement survey signals are the three evidence sources that drive a credible diagnosis. The baseline voluntary turnover rate (voluntary departures ÷ average team size in the past 12 months) is the anchor metric; the diagnosis must go beyond the number to identify the specific factors driving voluntary exits in this team.
Proof required
Submit: (1) a baseline voluntary turnover calculation for the past 12 months: number of voluntary departures, average team size, calculated voluntary turnover rate, and the data source; (2) exit interview synthesis or equivalent evidence (stay interview summaries, anonymised 1-on-1 feedback, engagement survey open-text responses) covering at least 3 data points — what were the stated or implied reasons people left or considered leaving; (3) a retention risk diagnosis (200 words minimum): the 1–2 dominant risk factors based on the exit data, and a pre-committed intervention with the specific change you will implement to address the dominant risk factor.
What gets checked
- Voluntary turnover calculation uses a clear methodology — include the formula used, the data source for departure counts, and the time period; a single figure with no supporting data is not a baseline
- Exit data is genuine — a diagnosis based on 'I think people left because of X' without exit interview evidence or equivalent is speculation; minimum 3 data points (each a departed employee's stated reason, or a stay interview response, or an engagement survey open-text)
- Intervention is specific to the dominant risk factor — 'improve culture' does not address a specific risk factor; 'implement quarterly career development conversations with a written growth plan for each team member' addresses a risk factor of limited career visibility
Resources
Foundationstart here
Depthgo deeper
Masteryfor the dedicated
What a verifier looks for
- Turnover methodology: ask the submitter to state the formula they used and the data source — '30% turnover' needs the underlying numbers; ask how they counted average team size
- Exit data quality: ask how the exit data was collected — retrospective recall of 'what I think people said' has low reliability; actual exit interview notes, anonymised 1-on-1 records, or engagement survey data has higher reliability
- Intervention specificity: ask what a team member would experience differently after the intervention is implemented — a vague intervention description produces uninterpretable results
- Conflating involuntary and voluntary turnover — performance-based departures, layoffs, and voluntary resignations are different signals; the baseline must isolate voluntary departures only
- Diagnosing compensation as the root cause when it is a symptom — compensation dissatisfaction often masks a deeper driver (no promotion path, lack of recognition, poor management); a diagnosis that goes no deeper than 'people want more money' will produce an intervention that does not address the real cause
- Setting the pre-committed intervention before diagnosing the risk factor — an intervention decided before the exit data is reviewed is a hypothesis, not a diagnosis-driven intervention; the M1 evidence must clearly precede and inform the intervention choice
Implement Retention Intervention and Collect Early Signal Data
3–12 months of post-intervention leading signal collection (depending on intervention type and available data)
Retention is a lagging metric: the 12-month voluntary turnover rate will not show the full effect of an intervention for a year. This milestone collects the early leading signals that the intervention is working — stay interview responses post-implementation, engagement survey score changes, number of at-risk conversations surfaced and addressed, career development plan completion rates, or any other leading indicator specific to the intervention. The 12-month outcome data at M3 is the definitive proof, but M2 leading signals are required to demonstrate the intervention is active.
Proof required
Submit: (1) an intervention implementation log: the start date, the specific changes implemented (one entry per change with the date and description), and any design adjustments made during implementation; (2) 3–6 months of post-intervention leading indicator data: the specific leading indicator relevant to your intervention type — career development conversation completion rates, engagement survey scores, stay interview sentiment summaries, number of at-risk situations surfaced and resolved; (3) the first available 12-month voluntary turnover datapoint post-intervention (if the 12-month window has elapsed); if not yet elapsed, a quarterly voluntary turnover rate showing direction of change with honest caveats about statistical significance.
What gets checked
- Leading indicator is relevant to the intervention — a career development conversation initiative must show career conversation completion rates; an engagement-survey-based intervention must show engagement score changes; the leading indicator must be logically connected to the specific risk factor diagnosed in M1
- Post-intervention data covers at least 3 months — a single post-implementation measurement is noise; 3 months of data shows direction of change; less than 3 months does not constitute leading signal evidence
- 12-month data or honest quarterly proxy — if the 12-month window has not yet elapsed, a quarterly turnover rate comparison with an explicit caveat that it may not reflect the final annual rate satisfies the standard; claiming 12-month improvement on 3 months of data does not
Resources
Foundationstart here
Depthgo deeper
What a verifier looks for
- Leading indicator relevance: ask why the chosen leading indicator is connected to the M1 dominant risk factor — a leading indicator selected for convenience rather than logical connection to the diagnosed risk factor will not show a real retention effect
- Data period: confirm that the leading indicator data covers at least 3 months — ask for the dates of the first and last data points
- Absence of departures: if the submitter cites zero voluntary departures as evidence, ask what the base rate of voluntary departures is in a typical 3-month period for a team this size — zero departures may not be below the expected baseline
- Selecting a leading indicator that is too easy to move — completion rates for activities the manager schedules are easy to hit; the leading indicator must measure a real signal of retention risk reduction, not task completion compliance
- Treating no departures during the measurement period as evidence of retention improvement — zero voluntary departures in a 3-month period may simply reflect a healthy labour market, a difficult economy, or coincidence; zero departures is not evidence of intervention effectiveness without a comparison baseline
- Conflating retention-adjacent metrics with retention — lower absenteeism or higher performance scores may be correlated with retention but are not leading indicators of it; the leading indicator must be directly connected to voluntary turnover risk
Report Retention Outcome with Expert Review
1 week for outcome report, lessons-learned, and expert review session (at 12-month mark or with quarterly proxy)
The final milestone compares the M1 baseline voluntary turnover rate with the post-intervention rate (12-month preferred, quarterly with explicit caveats if 12 months have not elapsed), and requires a real-time review with a named, qualified reviewer. The ADVERSARIAL VERIFICATION Level 2 standard applies: the reviewer must challenge the attribution — a lower turnover rate could reflect macroeconomic conditions, team composition changes, or the labour market, not the intervention.
Proof required
Submit: (1) a retention outcome report: the M1 baseline voluntary turnover rate, the post-intervention rate (12-month preferred), the delta, and a 200-word attribution analysis — which factors besides the intervention might explain the change; (2) a lessons-learned document (300 words minimum): what the intervention addressed, what it did not address, and what the next retention improvement would be if continuing; (3) documentation of a real-time Q&A session with a named reviewer who has ≥3 years of people management experience and a verifiable professional profile — the documentation must include specific challenge questions and your responses; the reviewer must not be your current manager or a direct report.
What gets checked
- Retention outcome uses the same voluntary turnover methodology as M1 — the same formula, the same data source, the same time-period calculation method
- Attribution analysis names macroeconomic and contextual factors — a retention improvement during a tech layoff period cannot be solely attributed to a management intervention; honest attribution acknowledges the labour market context
- Q&A shows the reviewer challenged attribution and intervention design — a reviewer who only confirmed the turnover figures without asking 'how do you know this was your intervention and not the economic environment?' did not meet the adversarial standard
Resources
Foundationstart here
Depthgo deeper
What a verifier looks for
- Data period: confirm whether the outcome report uses 12-month data or a quarterly proxy — if quarterly, ask the submitter how they are accounting for the fact that 3–6 months of low attrition may not predict the 12-month rate
- Attribution: ask the submitter to describe the labour market conditions during the post-intervention period — if it was a period of high unemployment or tech layoffs, low voluntary turnover may reflect macroeconomics rather than the intervention
- Reviewer challenge: look for causal logic questions in the Q&A documentation ('how do you know it was the career conversations and not the compensation changes that happened at the same time?') rather than only process questions
- Claiming 12-month retention improvement on less than 12 months of data without a clear proxy methodology — an honest quarterly proxy with explicit caveats is acceptable; an uncaveated improvement claim on 6 months of data is not
- Attribution analysis that credits only the intervention — retention is influenced by macroeconomics, team composition changes, competitive dynamics, and company performance; an attribution analysis with no alternative factors considered is not credible
- Reviewer who focuses only on the process rather than the causal logic — a reviewer who asks only 'how did you implement the stay interviews?' rather than 'why do you believe the stay interviews changed retention rather than simply being correlated with a period of low attrition?' did not meet the adversarial standard