The First
Ninety Days
Two studies, one operating question
In Fall 2026 Activated Insights published the first editions of its Workforce Index for skilled nursing and for assisted living. Together they follow more than 236,000 employees through hiring, the first year and departure, and they pair employment records with what those employees said in new-hire, engagement and exit surveys.
This paper reads the two reports side by side and asks a question neither was written to answer: if the industry now knows who leaves, when, and what they say about why, what stops an operator from acting on it in every community at once?
The answer offered here is that workforce stability is an operating-record problem before it is an analytics problem. The evidence sits in systems that do not agree, the people with authority to act see it late, and the number means different things to different readers. SeniorCRE's doctrine for that condition is short: Coherence before Intelligence.
Whose work is whose. Every Index figure in this paper is Activated Insights' and is cited to the edition it came from. The comparison, the derived arithmetic and the governance reading are SeniorCRE's, drawn from its published workforce method and evidence register. Activated Insights did not commission, contribute to or review this paper.
How to read it. Sections 01 to 05 set out the evidence. Sections 06 and 07 make the argument and turn it into a ninety-day sequence. Section 08 is for owners. Section 09 states what this paper does not claim.
Contents
- 01Executive summary
- 02The same operator, ten points apart
- 03What the Index measured
- 04One role, one window
- 05What employees say
- 06Why this is an operating-record problem
- 07Six controls and the first ninety days
- 08For owners: set the standard, not the system
- 09Where SeniorCRE fits, and what this paper does not claim
- AAppendix: six working tools
- —Sources, method notes and important notices
Executive summary
The retention gap that matters most is inside the operator, and that makes it governable.
Communities of the same operator differ almost as much as operators do. In assisted living the typical spread in first-year retention between one operator's stronger and weaker communities is 9.7 points; the gap between top- and bottom-quartile operators is 9.9. In skilled nursing the figures are 11.4 and 16.5. Same pay bands, same benefits, same systems, same policies, different result. Whatever explains that difference is local, and it is the part of the problem an operator can govern.
The loss is concentrated in one role and one window. Annual turnover is 64.4% in skilled nursing and 71.9% in assisted living. Direct care turns over at 80.5% and 92.8%, two to 2.3 times the management rate. Four in ten direct-care hires are gone by day 90 and about one in five by day 30. Fewer than three in ten reach a first anniversary.
Employees attribute it to things a community manages. Leavers rate work/life balance 25 to 28 points below current staff, their supervisor 24 to 26 points below and recognition 22 to 23 points below. Three yes-or-no onboarding answers separate hires who stay from hires who leave by 12 to 25 points, and the first arrives at day 30.
The constraint is coherence. Each finding lives in a different system: separations in payroll, schedules in timekeeping, survey answers in an engagement platform, workload in the clinical record, cost in the ledger. Turnover has no single definition: the Index counts all departures and every role, SeniorCRE's published methodology counts voluntary separations among employees only, and CMS publishes a third figure for skilled nursing. An operator cannot manage a spread it cannot measure the same way twice, and cannot act on a day-30 signal that never reaches a person with authority over the schedule.
What to do. Apply to the workforce number the six controls a finance function applies to a ledger: shared definitions, source authority, reconciliation, decision lineage, human authority and repeatability. Section 07 runs SeniorCRE's published five-step workforce method against the first ninety days, on the systems an operator has today. Observable drivers come first; a predictive model is not required. None of it requires buying software, and none of it requires SeniorCRE.
The same operator, ten points apart
Retention is decided community by community. That is the finding that changes the question.
Between operators, one-year retention of new hires ranges from 29.4% to 45.9% at the quartile marks in skilled nursing and from 29.8% to 39.7% in assisted living. Inside a typical multi-site operator, the stronger and weaker communities differ by 11.4 and 9.7 points. In assisted living the two spreads are essentially the same size.
Communities of one operator share a wage scale, a benefits plan, an applicant-tracking process, a learning system and a handbook. If those were what determined retention, the spread inside an operator would be small. It is not. The Index's own reading is that the difference lies in what each location controls day to day: scheduling, supervision and how new hires are brought on.
The gap is also a workload. Dividing one by the retention rate gives the number of hires needed to produce one employee who reaches a first anniversary. A bottom-quartile skilled nursing operator makes about 3.4 hires for each one-year stayer; a top-quartile operator makes 2.2. Every extra hire is recruited, oriented, scheduled and trained by the same supervisors whose relationship with staff is itself a retention factor. This arithmetic is SeniorCRE's, not the Index's, and the paper puts no dollar figure on it; Appendix Exhibit A5 gives a worksheet for an operator's own inputs.
What the Index measured
Two segments are converging on the same structure: a frontline, first-year problem.
The two reports draw on employment records and survey responses held in Activated Insights' retention product: 107 skilled nursing operators with 1,604 facilities, and 95 assisted living operators with 958 communities, through the third quarter of 2026. Turnover is departures over the trailing twelve months divided by average headcount. Retention follows each hire for 365 days. Both are computed from employer rosters and cover every role.
Skilled nursing has been flat, between 63.4% and 65.7% since late 2024, while assisted living has fallen 11.8 points from 83.7%. Management turnover is nearly identical in the two segments, at 39.6% and 39.9%. The difference between the sectors, and most of the movement over time, is frontline.
One detail in the skilled nursing series deserves attention. The gap between direct care and management narrowed to 40.9 points, the smallest in the series, but it narrowed because management turnover rose to its highest reading, not because direct care improved. The supervisor relationship is the second-largest difference between people who stay and people who leave, so rising turnover among supervisors is a warning and not a gain.
| Measure | Skilled nursing | Assisted living |
|---|---|---|
| Sample | ||
| Operators · locations · employees | 107 · 1,604 · 172K+ | 95 · 958 · 64K+ |
| Turnover, trailing twelve months to 2026 Q3 | ||
| All staff (the Index figure) | 64.4% | 71.9% |
| Direct care | 80.5% | 92.8% |
| Management | 39.6% | 39.9% |
| Median operator · lowest-turnover quartile at or below | 54.8% · 34.3% | 63.4% · 51.8% |
| First-year retention, hires October 2024 to September 2025 | ||
| All hires still employed at one year | 34.1% | 33.3% |
| Direct-care hires still employed at one year | 29.8% | 27.7% |
| Direct-care hires gone by day 30 · by day 90 | 21.3% · 41.0% | 20.0% · 39.3% |
| Share of all departures that occur within 90 days of hire | 39.9% | 40.7% |
| The workforce on September 30, 2026 | ||
| Staff with under one year of service · three years or more | 37.5% · 32.7% | 43.6% · 25.4% |
| Direct-care staff in their first 90 days · three years or more | 16.8% · 26.0% | 20.2% · 18.1% |
One role, one window
Four in ten direct-care hires are gone by day 90, and that share has not moved in two years.
One role. Direct care accounts for 50.1% of all first-year departures in skilled nursing and 40.0% in assisted living. Add licensed nursing and the two roles account for 72.3% of first-year exits in skilled nursing. A point of direct-care retention moves the whole organization's number more than a point anywhere else.
One window. Of the direct-care hires who leave within a year, 58.4% in skilled nursing and 54.4% in assisted living are gone before day 90. The share of all departures that fall inside the first 90 days has stayed between 39.1% and 42.5% in every quarter since 2024 Q3, in both segments. Assisted living's overall turnover fell by more than eleven points over that period; the early-departure share did not move. The first ninety days remain unsolved.
Two workforces. The result is an organization that runs on two populations at once. In skilled nursing 37.5% of staff have under a year of service while 32.7% have three years or more. In assisted living the figures are 43.6% and 25.4%. People who get past the early months often stay for years, and the role that spends the most time with residents carries the least experience at any moment: 20.2% of assisted living direct-care staff are in their first 90 days.
What employees say
The widest gaps are schedules, supervisors and recognition.
None of the three widest gaps is a compensation item, and none is set at the corporate office. Each is produced, shift by shift, by a schedule and a supervisor. Friendship at work shows the smallest gap, yet it turns out to be the earliest usable signal.
Two cautions apply, and both come from the source. The relationship is an association: the Index does not show that arranging a friendship causes a hire to stay. And the day-30 signal arrives after about one in five direct-care hires has already left, so the first month has to be designed and cannot wait for a survey.
New-hire scores also decline during exactly this period, among people who are still employed: from 78.9 at day 15 to 75.6 at day 90 in skilled nursing and from 82.7 to 79.1 in assisted living. Therapy in skilled nursing is the one role whose score rises, which is evidence that the decline is not inevitable. Each of these reports has a counterpart in records the operator already keeps; Appendix Exhibit A1 pairs them.
Why this is an operating-record problem
The evidence, the definition and the authority to act sit in three different places.
Nothing in the Index is a secret to a good executive director. What an operator lacks is the ability to see it consistently across communities, early enough to act, in a form that reaches someone with authority over the schedule. Three structural conditions stand in the way, and none is solved by a better dashboard.
1. The evidence lives in systems that were never asked to agree
A day-30 survey answer sits in an engagement platform. The schedule that may explain it sits in timekeeping. The supervisor's name is on a roster, the workload is in the clinical record, and the agency invoice that follows a departure reaches the ledger weeks later. Each system is correct about what it records. No one of them holds the sequence.
Agreement between systems is not proof either. SeniorCRE's published example is a team member whom payroll and HR both show as active because a termination was never processed, while scheduling shows the person gone. HR had copied payroll, so the two agree because one copied the other. Even the hours disagree: scheduled, worked and paid hours are three quantities, and care coverage is a fourth.
Payroll and the schedule agree at 1,140 hours and are both wrong about coverage: forty of the paid hours were leave, employees worked 1,100, agency staff added 60, and 100 hours of assumed demand went uncovered. An executive reading the payroll total would see a fully staffed week.
2. The number has no single definition
The Index is careful about its own definitions, and its care illustrates the problem. SeniorCRE's published workforce methodology makes a different choice on nearly every line.
| Element | Activated Insights Workforce Index | SeniorCRE published workforce methodology |
|---|---|---|
| Source records | Employer rosters and survey responses | HRIS, scheduling, time and payroll, reconciled before any measure is trusted |
| Who is counted | Every role on the roster | Employees, one row each; agency workers kept out of the employee table and the retention denominator |
| Separations counted | All departures | Voluntary separations only; involuntary exits, reductions in force and internal transfers excluded |
| Rehires | A person who leaves and returns counts as two hires | Rehires within twelve months flagged separately |
| Denominator | Average headcount over the trailing twelve months | Average active headcount, annualized by months in the period |
| Early-tenure measure | One-year retention; departure within 90 days of hire | 90-day new-hire survival by closed hire-date cohort |
Neither set of choices is wrong. But an operator that measures voluntary turnover among employees and compares it with the Index's 64.4% or 71.9% is comparing unlike numbers, and will look better than it is. Skilled nursing operators carry a third figure: Payroll-Based Journal reporting remains mandatory each quarter, CMS computes a turnover measure from it, and Activated Insights states that this figure is not directly comparable with its own.
The same holds for every labor ratio built on census. SeniorCRE's published illustration gives one community three figures for labor hours per resident day: 3.9 on payroll hours and EHR census, 4.1 on payroll hours and billed days, 3.6 on scheduled hours and budget census. Each is arithmetically correct and answers a different question. An operator that reports “turnover” or “hours per resident day” without stating which choices it made is reporting a number nobody else can reconcile, including its own regional leaders.
3. The signal and the authority are in different places
The employee tells a survey tool. The person who can change the schedule is a unit supervisor who may never see that tool. The person accountable for the turnover number is a regional leader who sees it a quarter later, aggregated, after 40% of the departures it counts have already occurred. Insight without a route to authority is a report. SeniorCRE's published labor design calls the route the Operator Authority Chain™: six links, each a precondition for the next.
Exhibit 9 walks one signal along it, and its fifth link is deliberately empty. SeniorCRE's own workforce method says that observable drivers come first and a predictive model is not required, and that a fitted turnover score needs roughly 12 to 18 months of separations per role family before it should be attempted. A turnover-risk model trained on a roster that disagrees with payroll, scored against a definition that changes by community and delivered to no one in particular, will produce confident output and no change.
| 1 · EvidenceWhat each labor system asserted, dated and attributedSurvey: “no” to the day-30 friendship question. Schedule: four changes in month one. HR: active, day shift. | → | 2 · DisagreementWhere scheduling, payroll and HRIS conflictHR lists days; the schedule shows nights. The conflict stays visible and is not averaged away. | → | 3 · AuthorityWho may decide which labor number governsA standing rule, written before the conflict: the schedule governs assignment; payroll governs paid hours. |
| 4 · Operating recordThe governing determination per employee and purposeOne reconciled entry for this hire: nights, unstable schedule, inside the first 90 days. | → | 5 · IntelligenceExposure scoring on governed contextNot needed here. The observable drivers are enough. A score, where one exists, is a flag and not a conclusion. | → | 6 · ExecutionCoverage and intervention workflows with a named ownerThe unit supervisor confirms the case, sets a fixed schedule and a peer partner; the action and day-90 outcome are logged. |
Six controls and the first ninety days
Apply to the workforce number what finance already applies to the ledger, then run it for one quarter.
SeniorCRE's published standard for governable operating truth has six controls, each with an analogue a finance function already runs. They describe the operating record, not a product. Exhibit 10 applies them to the workforce number and ties each to the Index evidence that makes it necessary.
| Control | Financial analogue | Applied to the workforce number | Index evidence it answers |
|---|---|---|---|
| Shared definitions | Chart of accounts | Freeze period, employee identity, role, unit, voluntary separation, transfer, average headcount and closed-cohort rules before measuring anything. Keep agency workers out of employee retention denominators. | Roster-based and CMS turnover are not comparable; a rehire counts as two hires. |
| Source authority | System-of-record designation | Assigned per attribute, not per system: payroll governs paid hours, the schedule governs assignment, the clinical record governs acuity, the ledger governs posted cost. | Role families come from job titles, which differ by system and community. |
| Reconciliation | Account reconciliations | Join schedule, time, payroll and agency invoices by worker and shift. Tie scheduled, worked and paid hours. Retain unresolved mismatches; never average them. | Records post late and revise the newest quarter, mainly upward. |
| Decision lineage | Audit trail | Record each signal, the determination, who made it, the rationale and what followed. Log signals that were deferred or rejected as well. | Three onboarding answers flag hires far less likely to reach a year. |
| Human authority | Delegation of authority | A named manager confirms, defers or rejects each exception. HR authorizes employment decisions. A score never triggers an action on its own. | The widest gaps are schedules, supervision and recognition, all managed by people. |
| Repeatability | Standardized close | The same definitions and cadence in every community: intervention coverage weekly, voluntary separation monthly, 90-day survival per hire cohort. | A typical operator's communities differ by 9.7 to 11.4 points. |
Two carry most of the weight. Repeatability turns the within-operator spread from an anecdote into a managed number: until every community measures retention the same way on the same cadence, an operator cannot tell a ten-point gap from a ten-point difference in bookkeeping. Human authority keeps the response proportionate. The Index is explicit that its early signals describe groups, not individuals. A manager who uses a signal as a reason to have a conversation is acting on it correctly. A system that scores an employee and acts on the score is not.
The first year as an operating calendar
SeniorCRE publishes a five-step workforce method. It can be run on the systems an operator has today, inside the same ninety days.
Design the first month; do not wait to predict it. Step 3 looks at conditions the operator sets on day one: a stable unit, a consistent preceptor and a schedule that does not move. SeniorCRE's published article on turnover indicators offers starting thresholds for weekly review, among them overtime above 20% for three consecutive pay periods, more than three shift changes in a week, and two unscheduled absences in thirty days. They are starting points for an operator to test, and they have not been validated.
| Published step | Owner | What it means inside the first ninety days | Index finding it answers |
|---|---|---|---|
| 1 · Establish baseline definitions | HR and finance | Write and freeze the definitions in Exhibit 8. Pull one-year retention and 90-day survival by community for the last four closed hire cohorts and compute the spread between the strongest and weakest community. | Within-operator spread of 9.7 to 11.4 points |
| 2 · Reconcile payroll and scheduling | Payroll and scheduler | Join schedule, time, payroll and agency invoices by worker and shift for one closed period at two or three communities. Keep the mismatches on the page. | Work/life balance: the widest stayer-leaver gap |
| 3 · Identify exposure | Operations and clinical coverage | For every hire inside 90 days, review orientation completion, preceptor continuity, unit consistency and the day-14, day-30 and day-90 milestones. Review overtime concentration, open shifts and pay errors. Treat a supervisor change as a leading indicator. | 41.0% and 39.3% of direct-care hires gone by day 90 |
| 4 · Assign human review | A named manager; HR for employment decisions | Route each material exception, including a “no” on the day-30, day-60 or day-90 question, to a named person who confirms, defers or rejects it. Clinical leaders accept any coverage change. | 12- to 25-point gaps on three onboarding answers |
| 5 · Measure interventions | Operations; HR per cohort | Log action completion and 90-day survival for each closed cohort. Record whatever else changed at the same time. Before-and-after movement alone is not causation. | The Index reports association, not cause |
Read agency use as a symptom. Every early departure opens shifts, and the most expensive way to fill them should be the last resort. SeniorCRE's published displacement sequence runs through six internal options before agency labor, which is treated as a documented exception with a reason code (Appendix Exhibit A4). The reason codes, read by unit and shift, show where step 3 should look next.
What good looks like at day 90
- Any executive can state the operator's definition of turnover and of 90-day survival, and two executives give the same answer.
- The operator knows its own spread between communities and can compare it with the Index's 9.7 and 11.4 points, on a stated definition.
- One closed period reconciles across schedule, time, payroll and agency invoices, with every variance explained or logged.
- Every new direct-care hire has a named person who sees the day-30 answer within a week of it being given.
- One decision made in the quarter can be traced from signal to action to outcome, and nothing above depends on a particular employee, spreadsheet or vendor.
For owners: set the standard, not the system
Ask for the definition, the rule and the spread. Never mandate the system.
Owners and capital partners in senior housing & care have a direct interest in workforce stability. It shapes the experience on the floor, the reliance on agency and overtime, and the predictability of the operating result. It is also one of the few leading indicators available: the first-year retention of a hire cohort is visible months before its effects reach a financial statement.
The principle SeniorCRE sets out in The Institutional Owner applies without modification. An owner should expect governed operating truth and should never mandate an operator's systems, including SeniorCRE's. Mandating a system transfers accountability to the owner and does not solve the problem. The leverage is in the expectation. Six questions carry it for the workforce:
| # | Question an owner can ask | What a governed answer looks like |
|---|---|---|
| 1 | How do you define turnover and early retention? | A written definition, the same in every community, stating treatment of voluntary and involuntary exits, rehires and agency staff. |
| 2 | When HR, payroll and scheduling disagree, which governs? | A rule per attribute that existed before the disagreement did. |
| 3 | What is the spread in first-year retention across the communities in this portfolio? | A number, by community, for the last four closed hire cohorts. |
| 4 | What happens, and who acts, when a new hire signals risk at day 30? | A named role, a time limit and a record of what was done. |
| 5 | Which census and which hours sit behind your labor ratios, and how do agency invoices reconcile to shifts worked? | A declared basis on every KPI and a documented reconciliation with variances explained. |
| 6 | Could these answers be reproduced next quarter without the person who prepared them? | Yes, and here is last quarter's. |
An operator that can answer only with a software purchase has not understood the standard. An operator that can answer all six with documents it already keeps has given its owner something more useful than a favorable turnover figure: a reason to believe the next one. Owner reporting on labor should be aggregate; individual risk flags belong with the supervisor and HR and nowhere else.
Retention has an economic case, and owners will ask for it. Appendix Exhibit A5 gives the arithmetic as a worksheet with no numbers in it, because the numbers are the operator's to supply. A change in net operating income can be carried into a debt-coverage or capitalization calculation, and the result will look precise. It is division on assumed inputs and it is not evidence. The more durable gain for an owner is confidence in the labor information behind underwriting, asset management and forecasts.
Where SeniorCRE fits, and what this paper does not claim
Designed or built on synthetic data, published, and not yet proven with an operator. The status is stated exactly.
SeniorCRE is operator-controlled operating infrastructure for senior housing & care. Its differentiating construct is the Operator-Controlled Operating Record, designed to sit read-only above the systems an operator retains and to apply definitions, source authority, reconciliation and lineage to what they report. For workforce purposes it does not require replacing an EHR or an HR system: the published design works from scheduling, time, payroll, agency and census inputs. An operator can govern the systems it keeps, or later choose to run selected functions on SeniorCRE.
SeniorCRE publishes its evidence status on a four-level scale: Designed, Built, Operator deployment and Production evidence. “Built” on that scale rests on synthetic or de-identified demonstration data and is not validation with an operator. As of the register recorded November 13, 2026, which is self-reported, nothing in the workforce area has reached operator deployment or production evidence.
| Component | What it is designed to do | Published guardrail | Published status |
|---|---|---|---|
| Operator-Controlled Operating Record | Read retained systems read-only, attach lineage to each value and let the operator declare which source governs. | Conflicts keep both values and raise an exception; nothing is silently overwritten. | Designed. Not implemented in any community. |
| Workforce Retention Intelligence Engine (WRIE) | Six capabilities, including coverage exposure, 30-, 60- and 90-day flight-risk tiers, and 90-day survival reporting. | A named manager confirms every case. Individual scores go only to the direct supervisor and HR. Protected-class, sentiment, biometric and credit data are excluded. | Built on synthetic data. Not validated with an operator; not in operator deployment or production. |
| Labor Cost Intelligence & Forecasting System (LCIFS) | Forecast labor cost from the schedule, census, overtime and agency use, on a rolling 13-week basis. | No forecast output changes staffing or pay on its own. Each version is frozen and reconciled to actual payroll. | Design targets, not results. |
| Workforce Retention Predictor and Staffing-to-Acuity Balancer agents | Team-level retention hypotheses for review; coverage-gap scenarios by shift and skill mix. | No employment or staffing decisions. No training on operator, resident or workforce data. | Designed. Not built. |
| Ten workforce workflows | Scheduling and shift fill, onboarding, turnover risk review, agency reduction, credentialing, labor variance, payroll forecast, daily stand-up, staffing-rule tracking and PBJ submission. | Each has a named primary owner. The turnover workflow is human-in-the-loop; a score does not affect pay, scheduling or termination. | Designed. Not implemented in any community. |
| Supervisor effectiveness scorecard, scheduling stability index, outcome-linked recognition | The first three items in the published workforce roadmap queue. | No delivery dates; no retention gain is claimed for any of them. | Not built. |
The last row deserves a note. The three items at the head of SeniorCRE's workforce roadmap queue are a supervisor effectiveness scorecard, a scheduling stability index and outcome-linked recognition. They correspond to the three widest gaps the Index found between people who stay and people who leave. None is built, and no retention effect is claimed for any of them.
A bounded first step. SeniorCRE's published proof of concept for workforce is one community, one region or one contested labor measure, over one closed period with source records intact. The operator signs the definitions and authority rules; the period is reconciled; each exposure case is routed to a named owner. No duration, savings figure or accuracy figure is published for it. The first proof does not need a working turnover model. It needs to show that workforce information from the operator's existing systems can be made governable. Appendix Exhibit A6 proposes the evidence an operator should ask for.
Stewardship rules for workforce intelligence
- No autonomous employment, disciplinary, compensation or termination decision. A score does not affect pay, scheduling or termination.
- No employee is penalized because a model flags a possible departure. A flag is a reason for a supportive conversation.
- Data is minimized and purpose-limited. Protected-class attributes, social sentiment, personal-device, biometric and credit data are excluded.
- Individual information stays with the direct supervisor and HR. Executives, owners and capital partners see aggregates.
- Bias assessment, privacy review and prospective evaluation come before any predictive claim. Financial optimization never overrides resident safety or clinical judgment.
What this paper claims, and what it does not
It claims that the Index figures are reported accurately and attributed; that the derived arithmetic follows from them; and that the governance reading is a reasoned interpretation by SeniorCRE.
It does not claim that governance causes retention. The Index reports associations, and so does this paper.
It does not claim that any SeniorCRE system has reduced turnover, agency use or labor cost at any operator. No outcome, savings, payback or accuracy figure is offered, and none should be inferred.
It does not claim that the Operating Record, WRIE, LCIFS or any workforce workflow is implemented at an operator. WRIE is built on synthetic data and has not been validated with one.
It does not claim that the Index describes any single operator. Activated Insights states that its figures describe each segment as a whole, that these are first editions, that the newest quarter may revise, and that its turnover is not comparable with CMS figures.
It does not claim endorsement by, or affiliation with, Activated Insights.
Governance first. Intelligence second. Execution last. SeniorCRE publishes evidence when it has it rather than before. Operators and owners who want to test their own workforce number against the six controls can request a 30-minute assessment at seniorcre.com.
Six working tools
Each can be used with the systems an operator has today. All are proposals or illustrations, not results.
| What employees report | What an operator can observe in its own records | Proposed management response |
|---|---|---|
| Work/life balance | Overtime concentration, schedule changes, consecutive days, short-turnaround shifts, open shifts | Schedule and workload review |
| Supervisor relationship | Supervisor changes, span of control, completed check-ins | Supervisor coaching and support |
| Recognition and appreciation | Check-ins completed; recognition of observed work logged by the supervisor | A recognition routine owned by the supervisor |
| A friend at work · day 30 | Preceptor continuity and unit consistency in the first weeks | An assigned peer partner |
| Job met expectations · days 60, 90 | Role, unit and shift actually assigned against what was offered | Role clarification; a realistic preview before hire |
| Understands the mission · day 90 | Orientation completion; the day-90 check-in | A development conversation tied to the mission |
| Metric | What must match before two communities are compared |
|---|---|
| Annual turnover | Which separations count, average headcount, reporting period |
| First-year retention | Closed hire cohorts, treatment of rehires and transfers |
| Agency share | Agency hours matched to shifts, total eligible worked hours |
| Overtime | Hours versus premium dollars; jurisdiction and employee classification |
| Hours per resident day | Eligible staffing categories and the census behind the denominator |
| Labor cost per resident day | Accrual treatment, cost components, census definition |
| Staffing to acuity | Care setting, workload method, competency requirements |
| Forecast accuracy | Horizon, frozen version, actuals and expense classification |
| Field | Illustrative entry |
|---|---|
| Finding | Sustained overtime concentration among direct-care staff |
| Scope | Community A, night shift, four consecutive weeks |
| Evidence | Schedule, timekeeping and payroll for the period, each dated and attributed |
| Definition in force | Overtime hours divided by worked hours, by role and shift |
| Source authority | Payroll governs paid hours; the schedule governs assignment |
| Unresolved conflict | One agency invoice not matched to a worked shift; shown as open, not as zero |
| Decision owner | Regional operations leader, by name |
| Approvals required | Clinical leadership for any coverage change; HR for any employment decision |
| Verification date | The date the evidence was last reconciled |
| Proposed next action | Review schedule stability, qualified internal capacity and workload distribution |
| Line | Formula, using the operator's own inputs | Rule that keeps it honest |
|---|---|---|
| Avoided separations | Change in retention × separations by role × fully loaded cost per separation | Build the cost from the operator's own cost lines. Not every avoided exit is cash. |
| Agency premium | Agency hours displaced × (agency rate − internal rate) | Premium only. Displacement is proven by reconciling invoices to worked shifts. |
| Overtime premium | Overtime hours avoided × the half-time premium | Only the premium counts; the base hours are still worked. |
| Net | Sum of the three − program cost − internal effort | No double counting. Census effects are reported separately, never added in. |
| Area | Evidence an operator should see before accepting the record |
|---|---|
| Data | Traceable source exports and a completeness assessment |
| Definitions | A metric register the operator has signed |
| Reconciliation | Material disagreements preserved, reviewed and dispositioned |
| Authority | Named decision owners and approved rules |
| Lineage | Any reported figure reproducible from source |
| Access | Role-based permissions verified; individual flags limited to supervisor and HR |
| Boundaries | Tests showing that no output triggers an action without authorization |
| Forecasting and prediction | Back-testing on the operator's history, then prospective evaluation, before any accuracy claim |
| Outcomes | Closed cohorts, appropriate comparisons and a finance-approved savings method |
Sources, method notes and important notices
Sources
Activated Insights, Workforce Index: Skilled Nursing — What Employees Say Drives Turnover & Improves Retention, Fall 2026. Data through 2026 Q3; 107 operators, 1,604 facilities, 172,000+ employees.
Activated Insights, Workforce Index: Assisted Living — What Employees Say Drives Turnover & Improves Retention, Fall 2026. Data through 2026 Q3; 95 operators, 958 communities, 64,000+ employees.
SeniorCRE, LLC: Coherence Before Intelligence (seniorcre.com/book); The Institutional Operator (seniorcre.com/the-institutional-operator); The Institutional Owner (seniorcre.com/the-institutional-owner); Operator Authority Chain (seniorcre.com/operator-authority-chain). Workforce pages on seniorcre.com: /workforce-intelligence and its methodology, best-practices, benchmarks, payroll-forecasting, proof-of-concept, journey and roadmap pages; /workflows/workforce and its ten workflow pages; /ai-systems/wrie; /seniorcre-ai/agents/workforce-retention and /staffing-to-acuity; /pbj-staffing-compliance; /senior-living-data-integration; the workforce and labor questions at /questions; and the article on predicting staff turnover. Pages read November 14, 2026.
Method notes
Combined total. The figure of more than 236,000 employees is the sum of the two editions' cover counts. Operators that run both settings may appear in both editions.
Index definitions. A hire is one period of employment. One-year retention is the share of hires with no separation recorded within 365 days. Early departure is a separation within 90 days of hire. Survey scores convert five-point items to a 0 to 100 scale; day-90 scores come only from hires still employed at day 90.
Limits stated by the source. Both reports are first editions. Movement under about two points is within normal variation. Records post with a lag, so the newest quarter is the most likely to change, mainly upward. Operator comparisons cover operators with at least 150 hires in the window. Early-signal results show association, not cause.
SeniorCRE arithmetic. Hires per one-year stayer equals one divided by the one-year retention rate. The between-operator gaps (16.5 and 9.9 points) are top-quartile minus bottom-quartile thresholds. The share of direct-care hires gone by day 365 is 100% minus one-year retention. Exhibits 7, 9 and A3 are illustrations and contain no operator data.
Important notices
This paper is published by SeniorCRE, LLC for general information. It is not legal, employment, clinical, financial, investment or regulatory advice. Operators remain responsible for their own employment practices and decisions, staffing levels, regulatory compliance and resident care. Nothing here should be used as the basis for an employment decision about any individual.
Outputs of analytical and artificial-intelligence systems are probabilistic and require human review. SeniorCRE components described as designed, proposed or built on demonstration data are not validated with, or in production at, any operator. Statements about what a system is designed to do are forward-looking and are not commitments about timing, availability or results.
Activated Insights' figures are reproduced as individual data points with attribution; the Workforce Index reports are © 2026 Activated Insights, and readers should consult the original reports for full charts, context and methodology. SeniorCRE is not affiliated with, endorsed by or sponsored by Activated Insights. Other names are the property of their respective owners; their mention does not imply integration, partnership or endorsement.
SeniorCRE® and Operator Authority Chain™ are marks of SeniorCRE, LLC. © 2026 SeniorCRE, LLC. All rights reserved.
Workforce companion materials
Supplied supplementary materials, separate from the original whitepaper and its PDF.
Anatomy of a Resignation
The 90-Day Retention Playbook
The original supplied PowerPoint, with thirteen slides on early-tenure workforce signals, supervisory review, and proposed interventions.
Unreviewed supplementary presentation: some comparisons mix within-operator spreads with between-operator quartiles. Prediction thresholds, causal statements, and cost-avoidance figures are unverified—not validated SeniorCRE outcomes or implemented capabilities. Human review and source verification are required before reliance.
Download the playbook · PowerPoint · 13 slidesWhy Senior Living Caregivers Really Quit
An audio deep dive on early-tenure turnover in senior housing and care, drawing on the Fall 2026 Activated Insights Workforce Index and SeniorCRE workforce papers.
Keep exploring workforce intelligence
Ten places to go next on workforce and the Workforce Retention Intelligence Engine.
- Workforce IntelligenceThe hub for every workforce page, guide and tool.
- Turnover PredictionHow departures are anticipated before they happen.
- Agency Labor ReductionMoving from agency dependence to internal capacity.
- WRIE MethodologyHow the retention engine reads your workforce data.
- Retention ROI CalculatorEstimate what lower turnover is worth to you.
- Payroll ForecastingProjecting labor cost before the payroll run.
- Workforce BenchmarksHow your staffing numbers compare.
- Proof of ConceptThe evidence behind the workforce approach.
- WRIEThe Workforce Retention Intelligence Engine, explained.
- The WRIE StoryWhy the retention engine was built, in plain language.