{ "name": "SeniorCRE Workforce Intelligence methodology", "publisher": "SeniorCRE, LLC", "url": "https://seniorcre.com/workforce-intelligence", "version": "2026-09-17", "lastReviewed": "September 17, 2026", "license": "https://creativecommons.org/licenses/by/4.0/", "evidencePosture": "Everything described here is built. It is not accompanied by published operator-production outcome data, and no savings figure or payback period is claimed. Industry ranges shown are published third-party ranges used to size exposure and to model a case from an operator’s own history.", "resources": [ { "href": "/workforce-intelligence/best-practices", "title": "Retention analytics best practices", "question": "How do senior housing & care operators use analytics to improve retention?", "desc": "Metric definitions, a 13-week workforce forecast example, human-reviewed retention signals, data requirements, measurement plans, and downloadable methodology.", "url": "https://seniorcre.com/workforce-intelligence/best-practices" }, { "href": "/workforce-intelligence/roi-calculator", "title": "Retention ROI calculator and model", "question": "How do you build a defensible retention ROI and agency-labor savings case?", "desc": "The interactive model plus the structure behind it: components, source authority for each line, frozen-baseline rules, attribution discipline, and the limitations that invalidate a result.", "url": "https://seniorcre.com/workforce-intelligence/roi-calculator" }, { "href": "/workforce-intelligence/payroll-forecasting", "title": "Payroll forecasting: method, checklist, and variance review", "question": "How do you forecast payroll and labor cost for a senior housing portfolio?", "desc": "Seven forecast components held separately, five accuracy measures, a readiness checklist across data, model, controls, and governance, and the weekly variance review that keeps a 13-week forecast honest.", "url": "https://seniorcre.com/workforce-intelligence/payroll-forecasting" }, { "href": "/workforce-intelligence/agency-labor-reduction", "title": "Agency labor reduction", "question": "How do operators reduce agency labor without losing coverage?", "desc": "Where premium labor originates, how invoices are reconciled against worked hours, and the internal-fill sequence that has to work before agency hours come down.", "url": "https://seniorcre.com/workforce-intelligence/agency-labor-reduction" }, { "href": "/workforce-intelligence/benchmarks", "title": "Workforce benchmarks and the Senior Care Workforce Risk Benchmark", "question": "What workforce benchmarks exist and how are they measured?", "desc": "Frozen measures, system-derived collection, suppression rules, participation floor, published field dates, and stated sample limitations. No edition published yet.", "url": "https://seniorcre.com/workforce-intelligence/benchmarks" }, { "href": "/workforce-intelligence/methodology", "title": "Measurement methodology", "question": "How are the workforce numbers defined and produced?", "desc": "Definitions, hours reconciliation, human-confirmed intervention, frozen baselines, and honest accuracy reporting, each with the test for whether it is actually in place.", "url": "https://seniorcre.com/workforce-intelligence/methodology" } ], "modelInputs": [ { "signal": "Overtime density (OT hours / scheduled hours, rolling 4 weeks)", "source": "Payroll and time-and-attendance", "rationale": "Sustained OT concentration on the same employees is the most consistently observable precursor to voluntary exit in shift-based care work.", "cadence": "Daily" }, { "signal": "Consecutive-day and clopen patterns", "source": "Scheduling system", "rationale": "Recovery-time compression predicts call-offs first and exits second.", "cadence": "Daily" }, { "signal": "Shift-swap, no-show, and late-clock behavior", "source": "Scheduling + time clock", "rationale": "Disengagement usually appears as schedule friction weeks before a resignation.", "cadence": "Daily" }, { "signal": "Earnings variance vs. the employee’s own trailing 8-week average", "source": "Payroll", "rationale": "Falling or erratic take-home pay is a stronger driver than absolute wage level for caregivers and aides.", "cadence": "Per pay period" }, { "signal": "Acuity-weighted workload per worked hour", "source": "Clinical / care-plan data reconciled to census", "rationale": "Headcount can be flat while the work itself intensifies; unweighted ratios hide that entirely.", "cadence": "Daily" }, { "signal": "Early-tenure isolation (days 1–90: orientation completion, preceptor continuity, unit consistency)", "source": "HRIS + scheduling", "rationale": "A large share of caregiver separations occur inside the first ninety days, where the causes are onboarding-operational rather than compensation.", "cadence": "Weekly" }, { "signal": "Primary-caregiver continuity on assigned residents", "source": "Scheduling + assignment data", "rationale": "Continuity breakdown correlates with both resident-experience decline and caregiver exit.", "cadence": "Weekly" }, { "signal": "Open-shift exposure on the employee’s unit and shift", "source": "Scheduling", "rationale": "Chronic vacancy on a unit transfers load to the people who stay, which is how one exit becomes three.", "cadence": "Daily" }, { "signal": "Agency mix on the employee’s shift", "source": "Scheduling + agency invoices", "rationale": "High contract-labor mix changes both workload distribution and perceived pay equity on the floor.", "cadence": "Weekly" }, { "signal": "Manager-span and supervisor-change events", "source": "HRIS", "rationale": "Supervisor change is a discrete, dateable risk event that is trivially available and frequently unused.", "cadence": "Event-driven" }, { "signal": "Commute-band and internal transfer history", "source": "HRIS (coarse, non-address band)", "rationale": "Coarse distance bands carry signal without storing home address in the model.", "cadence": "Monthly" }, { "signal": "Role-and-market wage position vs. BLS OEWS for the same role and MSA", "source": "BLS OEWS (public) reconciled to payroll", "rationale": "Public wage benchmarks make a pay-position finding defensible to an owner or lender.", "cadence": "Annual (BLS release) with quarterly reconciliation" } ], "excludedInputs": [ { "input": "Race, ethnicity, national origin, sex, religion", "reason": "Protected characteristics are excluded from features and from any derived interaction term. Adverse-impact monitoring is run on outputs instead." }, { "input": "Age and date of birth", "reason": "Excluded as a feature. Tenure is used instead, which is the operationally relevant variable." }, { "input": "Pregnancy, disability, leave type, and any health information", "reason": "Excluded. Leave presence may be used only as an availability fact for scheduling, never as a risk feature." }, { "input": "Union status and protected concerted activity", "reason": "Excluded categorically. Prediction must not touch protected activity." }, { "input": "Home address, precise geolocation, and personal device data", "reason": "Excluded. Only coarse commute bands are used." }, { "input": "Survey sentiment as a primary driver", "reason": "Not excluded, but never a standalone trigger: response bias and cadence make it unusable as the basis for an individual-level action." }, { "input": "Immigration status and visa detail", "reason": "Excluded from the model entirely; handled only in compliance workflows." } ], "interventionWorkflow": [ { "window": "Days 0–7 (confirm before you act)", "objective": "Establish that the score reflects reality on the floor before anyone is contacted.", "owner": "Community executive director with the regional workforce lead", "actions": [ "Review the ranked flight-risk list with the driver contributions shown per employee — never a bare score.", "Mark each name confirmed, not-confirmed, or insufficient-information. Not-confirmed is a first-class outcome and is retained for model review.", "Suppress any name where the driver is a data artifact (missed punch batch, mis-mapped unit, agency worker in the employee table).", "Record the confirmation decision, the decider, and the timestamp in the audit trail." ], "exitCriteria": "Every high-risk name has a dated human decision. Nothing is auto-actioned." }, { "window": "Days 1–30 (remove the operational cause)", "objective": "Act on the driver, not the score.", "owner": "Scheduler and unit manager, tracked by the ED", "actions": [ "Overtime-density driver: rebalance the next two schedule cycles and cap consecutive days for the named employees.", "Earnings-variance driver: correct the schedule shortfall or the differential error before any conversation about pay.", "Early-tenure driver: assign a named preceptor, hold unit and shift constant for the remainder of the ninety days, and complete the day-14 and day-30 check-ins.", "Continuity driver: restore primary-caregiver assignment consistency on the affected residents.", "Log each action against the employee record with the driver it addresses." ], "exitCriteria": "Each confirmed name has at least one logged action mapped to a specific driver, with a due date." }, { "window": "Days 31–60 (verify the driver actually moved)", "objective": "Check that the operational condition changed, independent of whether the employee stayed.", "owner": "Regional workforce lead", "actions": [ "Re-measure the driver metric for each named employee against its value on the confirmation date.", "Escalate any name where the driver is unchanged after two schedule cycles — that is a scheduling-capacity problem, not a retention problem.", "Review the unit-level pattern: three or more confirmed names on one unit and shift is a staffing-structure finding, not twelve individual conversations.", "Feed suppression reasons back into the data-quality backlog." ], "exitCriteria": "Driver movement is recorded per name; unit-level findings are opened as structural items." }, { "window": "Days 61–90 (attribute honestly)", "objective": "Close the loop at the ninety-day milestone with an attribution a CFO can defend.", "owner": "Finance owner with the regional workforce lead", "actions": [ "Report retained / separated / transferred for every confirmed name, alongside the not-confirmed cohort as a comparison group.", "Attribute only where a driver moved and an action was logged; everything else is reported as unattributed.", "Publish the model’s precision at the operating threshold, plus the count of suppressed and not-confirmed names.", "Re-freeze the baseline for the next quarter and version the model." ], "exitCriteria": "A dated ninety-day cohort report exists with attributed, unattributed, and suppressed counts separated." } ], "metricDefinitions": [ { "metric": "Annualized voluntary turnover rate", "formula": "(voluntary separations in period / average active headcount in period) × (12 / months in period)", "numeratorDenominator": "Numerator excludes involuntary terminations, RIFs, and internal transfers. Denominator is average active headcount, not period-end headcount.", "commonError": "Mixing voluntary and involuntary separations, or using period-end headcount, which inflates the rate in a shrinking portfolio." }, { "metric": "90-day new-hire survival", "formula": "hires still active on day 90 / hires in the cohort", "numeratorDenominator": "Cohort is defined by hire date, not by month of report. Rehires within twelve months are flagged separately.", "commonError": "Reporting a rolling snapshot instead of a closed cohort, which makes the number move for reasons unrelated to retention." }, { "metric": "Regrettable turnover", "formula": "voluntary separations flagged regrettable / total voluntary separations", "numeratorDenominator": "Regrettable requires a documented manager designation at separation, not a retrospective judgment.", "commonError": "Backfilling the flag after the fact, which makes the metric unauditable." }, { "metric": "Agency share of worked hours", "formula": "agency worked hours / (agency worked hours + employee worked hours)", "numeratorDenominator": "Agency hours come from invoices reconciled to the schedule, not from the schedule alone.", "commonError": "Using scheduled rather than invoiced agency hours, which understates share whenever cancellations are billed." }, { "metric": "Overtime density", "formula": "OT hours / total worked hours, by role and shift", "numeratorDenominator": "Computed at role-and-shift level; portfolio averages hide the concentration that drives exits.", "commonError": "Reporting a single portfolio OT percentage." }, { "metric": "Vacancy duration (time-to-fill)", "formula": "median days from requisition open to first worked shift", "numeratorDenominator": "Ends at first worked shift, not at offer acceptance.", "commonError": "Measuring to offer acceptance, which omits the no-show and orientation-attrition window." }, { "metric": "Primary-caregiver continuity", "formula": "resident-days covered by the top-two assigned caregivers / total resident-days", "numeratorDenominator": "Measured per resident, then rolled up. Agency coverage counts against continuity.", "commonError": "Measuring at the unit level, which averages away the residents with the worst continuity." }, { "metric": "Cost-to-fill an open shift", "formula": "incremental premium (agency rate differential + OT premium + guarantee + travel/orientation) for one covered shift", "numeratorDenominator": "Incremental cost only — base hours the operator would have paid regardless are excluded.", "commonError": "Charging the full agency invoice as incremental, which overstates avoidable cost." }, { "metric": "Fully loaded cost per separation", "formula": "separation + vacancy coverage premium + recruiting + onboarding + ramp productivity loss", "numeratorDenominator": "Built from the operator’s own payroll, invoice, and recruiting cost lines; industry ranges are used only to sanity-check.", "commonError": "Adopting a published per-separation figure as if it were the operator’s own cost." } ], "roiFormulas": [ { "name": "Avoided-separation value", "expression": "AS = (B_sep − M_sep) × C_sep_loaded", "variables": [ "B_sep = separations in the frozen baseline period, same roles, same communities, seasonally aligned", "M_sep = separations in the measured period", "C_sep_loaded = fully loaded cost per separation, built from the operator’s own cost lines" ], "guardrail": "Only counted where a driver moved and an action was logged. Report the rest as unattributed." }, { "name": "Avoided agency premium", "expression": "AA = Σ (H_agency_baseline − H_agency_measured) × (R_agency − R_internal_loaded)", "variables": [ "H = invoiced agency hours by role and shift", "R_agency = effective invoiced agency rate including guarantees and cancellations", "R_internal_loaded = internal wage plus burden for the same role and shift" ], "guardrail": "Rate differential only. Hours the operator would have paid internally are not savings." }, { "name": "Avoided overtime premium", "expression": "AO = (H_ot_baseline − H_ot_measured) × 0.5 × R_base_loaded", "variables": [ "H_ot = overtime hours by role and shift", "R_base_loaded = base wage plus burden" ], "guardrail": "Only the 0.5 premium portion counts. Counting the full OT hour double-counts base labor." }, { "name": "Census-protection value (reported separately, never blended)", "expression": "CP = Δ occupied units × net revenue per unit per month × months", "variables": [ "Δ occupied units attributable to staffing capacity, evidenced by a documented admission hold released" ], "guardrail": "Reported as a separate, clearly-labelled scenario line. Never added into the retention ROI total without a documented hold." }, { "name": "Net program value", "expression": "NPV_program = (AS + AA + AO) − (platform cost + internal effort cost)", "variables": [ "internal effort cost = loaded hours spent by EDs, schedulers, and finance on the workflow" ], "guardrail": "Internal effort is never excluded. A model that omits operator labor is not a model an owner will accept." } ], "sensitivityTreatment": [ { "variable": "Fully loaded cost per separation", "treatment": "Run low / base / high from the operator’s own cost lines. Publish all three; lead with low." }, { "variable": "Attribution share", "treatment": "Default to attributing only the confirmed-and-actioned cohort. Show a 0% attribution floor case so the downside is visible." }, { "variable": "Agency rate differential", "treatment": "Use the trailing three months of invoices, not the contracted rate card." }, { "variable": "Seasonality", "treatment": "Align the baseline and measured periods to the same calendar quarters, or state the adjustment." }, { "variable": "Portfolio change", "treatment": "Exclude acquisitions, dispositions, and units in lease-up from both periods." } ], "limitations": [ { "limitation": "No operator-production workflow is published as of September 29, 2026.", "consequence": "Every capability statement is at the built level. Ask for a back-test against your own history rather than a reference outcome." }, { "limitation": "Prediction cannot establish causation.", "consequence": "A flight-risk score identifies conditions that precede exits. It does not prove an intervention caused a retention. Attribution is reported, not asserted." }, { "limitation": "Model quality is bounded by the worst input feed.", "consequence": "Unreconciled agency invoices, missing punches, or mis-mapped units degrade scores faster than any modeling choice improves them." }, { "limitation": "Small communities produce unstable rates.", "consequence": "Below roughly 40 FTEs in a role, quarterly turnover rates swing on single events. Use counts and rolling twelve-month windows instead." }, { "limitation": "Cold-start requires history.", "consequence": "Roughly twelve to eighteen months of separation history per role family is needed before scores are worth acting on; before that the workflow still runs on driver metrics alone." }, { "limitation": "Labor markets shift underneath the model.", "consequence": "Wage compression, a new competitor opening, or a rate-card change invalidates a stale baseline. Re-freeze quarterly." }, { "limitation": "Published industry cost ranges are not operator results.", "consequence": "They size exposure. They are never presented as savings achieved, and no payback period is published." }, { "limitation": "Adverse-impact risk is real and must be monitored.", "consequence": "Protected characteristics are excluded as inputs, but outputs are still monitored for disparate distribution, and monitoring results are part of the audit record." } ], "caseStudyFormat": [ { "section": "1. Operator profile", "requirement": "Portfolio size, care types, states, bed/unit count, and payer mix. No named operator without written consent." }, { "section": "2. Frozen baseline", "requirement": "Named period, exact metric definitions used, and the data sources reconciled. Baseline is frozen before the first intervention and never restated." }, { "section": "3. Data readiness at start", "requirement": "Which historical exports, representative records, sandbox feeds, or manual inputs were used, and which reconciliation exceptions were open at the start." }, { "section": "4. Intervention log", "requirement": "Confirmed names, drivers addressed, actions logged, and dates. Includes suppressed and not-confirmed counts." }, { "section": "5. Measured period result", "requirement": "The same metrics, same definitions, seasonally aligned, with attributed / unattributed / suppressed reported separately." }, { "section": "6. Model performance", "requirement": "Precision and recall at the operating threshold, model version, and drift notes for the period." }, { "section": "7. Economics", "requirement": "ROI formulas applied with the operator’s own cost lines, low/base/high sensitivity, and internal effort cost included." }, { "section": "8. Confounders and limitations", "requirement": "Wage actions, market events, acquisitions, leadership changes, and anything else that could explain the result." }, { "section": "9. Sources and verification", "requirement": "Public sources cited with access dates. Operator-supplied figures marked as operator-reported and reviewed by a named operator signatory." }, { "section": "10. Evidence class", "requirement": "Every figure tagged: A public primary source, B vendor public materials, C operator-reported, D design target not yet operator-proven." } ], "sources": [ { "label": "BLS Occupational Employment and Wage Statistics (OEWS)", "url": "https://www.bls.gov/oes/", "use": "Wage position by role, state, and MSA." }, { "label": "BLS Job Openings and Labor Turnover Survey (JOLTS)", "url": "https://www.bls.gov/jlt/", "use": "Sector-level quits and openings context for the baseline period." }, { "label": "CMS Payroll-Based Journal (PBJ) staffing data", "url": "https://data.cms.gov/provider-data/", "use": "Reported nurse-staffing hours and turnover measures for skilled nursing." }, { "label": "CMS Care Compare / Five-Star staffing measures", "url": "https://www.medicare.gov/care-compare/", "use": "Publicly reported staffing and turnover measures used for peer comparison." }, { "label": "BLS Current Employment Statistics — healthcare and residential care", "url": "https://www.bls.gov/ces/", "use": "Employment trend context for the market in the measured period." } ] }