According to the 2026 NSI National Health Care Retention and RN Staffing Report, 22.7% of newly hired registered nurses left their employer within the first year. That is the number worth sitting with, because it is not a recruiting failure. Those nurses were recruited successfully. They were interviewed, screened, credentialed, oriented, and assigned to a unit, and then within twelve months, roughly one in five of them was gone. The same report puts the average cost of replacing one staff RN at $60,090.

Most health systems respond to a number like that by spending more on recruitment, which is understandable and also mostly wasted, because the leak is downstream of the offer letter. The period that decides whether a new clinician stays is the stretch between week three and month nine, after orientation has finished and before the person has built any real relationships on the unit. That stretch is rarely staffed with anything, since preceptorship covers the first few weeks and then ends.

A mentorship program is the cheapest structural thing you can put into that gap. It is also, in my experience of watching these launch and then quietly die, the initiative most likely to be announced at a town hall and abandoned within eighteen months because nobody owned the operational detail. What follows is the sequence that separates the programs that survive from the ones that do not.

Mentorship, Preceptorship, Residency, and Coaching Are Four Different Things

The words in this category get used interchangeably by people who should know better, and the confusion causes real damage during planning, so it is worth being literal.

A mentorship program is a structured, longer-term developmental relationship between a more experienced person and a less experienced one, focused on career direction, judgment, professional identity, and navigating the organization. It is not tied to a competency checklist and it does not usually involve evaluation.

A preceptorship is a short, role-specific clinical training assignment in which an experienced clinician supervises and signs off on a new hire's practice. It is evaluative, time-boxed to the orientation period, and exists to establish safe independent practice.

A nurse residency is a formal, structured transition-to-practice program, normally running for a full year, that combines classroom content, clinical support, and cohort activity for newly licensed nurses. The Vizient/AACN Nurse Residency Program is the best-known example, and its published figures put first-year retention for participants at 89.0% against a national comparison figure of 76.2%.

Coaching is shorter, more targeted, and performance-focused. A coach works on a specific behavior or skill over a defined period, and then the engagement ends.

These are complementary rather than competing. Preceptorship gets a clinician safe, a residency carries that clinician through the first year, and mentorship takes the person from year one to year five and onward into charge, educator, or management roles. If your organization already runs a residency, the mentorship program should start where the residency ends rather than duplicating it.

Step One: Decide What the Program Is Actually For

Almost every failed program I have seen failed here, at the very beginning, because the stated purpose was some version of "supporting our people," which is a sentiment rather than a purpose. A real purpose is specific enough that you could tell, twelve months later, whether it happened.

Workable purposes look like these:

• Reduce first-year voluntary turnover among newly licensed nurses on medical-surgical and telemetry units.
• Prepare a defined group of experienced staff nurses for charge-nurse and unit-educator roles within eighteen months.
• Transfer procedural and institutional knowledge from a cohort of clinicians who are within five years of retirement.

Pick one, or two at the absolute outside. A program that tries to serve every cohort at once produces a mentor directory that nobody can navigate and a set of outcomes that nobody can measure.

The purpose also determines who owns the program. A retention-focused program belongs with nursing professional development or the CNO's office, because those are the people with unit-level visibility, whereas a leadership-pipeline program tends to sit better with talent management. A program with no clear owner drifts into whichever inbox is least busy, which is how programs die.

Step Two: Choose the Cohort Where Attrition Actually Costs You

Workforce pressure is not distributed evenly across a health system, and the temptation to launch enterprise-wide is a trap. Run the numbers by unit and by tenure band first.

The wider context supports being selective. The U.S. Bureau of Labor Statistics projects approximately 180,800 openings for registered nurses each year on average over the 2025 to 2035 decade, with most of those openings coming from replacing workers who leave the occupation rather than from net new positions. The labor market is not going to solve your unit- level vacancy for you.

Practically, the cohorts worth starting with tend to be new graduates on high-acuity units, experienced staff who have just moved into a first formal leadership role, and clinicians in specialties where your internal posting-to-fill time is longest. Run a pilot of forty to eighty pairs rather than four hundred, because a pilot that works produces internal demand, and internal demand is the only thing that reliably gets a program funded in year two.

Step Three: Design the Match, Because This Is Where Programs Break

If a program is going to fail after launch, the match is usually the reason. Pairs that were assigned badly stop meeting after the second session, and once a meaningful share of pairs have gone quiet, the program loses credibility across the unit and never recovers it.

The default approach is to match on job title and seniority, mostly because job title is the field that already exists in the HR system and nobody has to do additional work to use it. It is a poor predictor. Being a nurse manager for nine years tells you very little about whether that person can help a second-year nurse decide between an educator track and a nurse practitioner program.

A competency-based approach pairs people on demonstrated capability instead. You define the small set of things the program is meant to develop, such as clinical decision-making under time pressure, difficult conversations with families, or preparing for a first leadership role, and then you match on those attributes rather than on organizational position. In clinical settings, this has a second advantage that is easy to miss, which is that it lets you pair across specialties and across sites, and that widens the pool of available mentors considerably. A cardiac ICU nurse in one city can genuinely mentor an oncology nurse in another on leadership and career navigation, even though neither could precept the other.

You also have to decide who controls the pairing. Administrator-assigned matching gives you oversight and prevents the popular mentors from being oversubscribed while everyone else sits idle. Self-directed matching, where mentees browse profiles and request a mentor themselves, tends to produce stronger early commitment, because people invest more in a choice they made for themselves. Many programs end up running a hybrid, with self-selection as the default and administrator intervention for anyone who has not matched within a set window.

Putting any of this into operation is where most hospital program owners stall, because a competency-based directory with matching rules, scheduling, and reporting is not something a spreadsheet does past about thirty pairs. A category of vendors now sells exactly this as mentoring software for healthcare organizations, covering matching, agreements, in-platform scheduling, and program reporting for hospitals and health systems as a managed service rather than an internal build. MentorCity is one of the providers that has run the competency- based model with a healthcare client: its work with the AO Foundation produced the AO Access mentorship program, which was built deliberately on non-clinical competencies rather than on specialty or seniority, and which paired medical professionals across multiple global regions and clinical specialties. The company reports that 81% of participants maintained their mentoring relationships through to completion, and that the same proportion, 81%, reported professional growth. Participants rated platform usability as good or very good in 85% of cases, which is a more important figure than it appears, given that clinicians tend to abandon anything that takes more than a minute to open between patients.

Step Four: Give Every Pair an Agreement and a Cadence That Survives the Schedule

Two clinicians on rotating shifts will not find each other by accident. This is the most underestimated operational problem in healthcare mentoring, and it is the reason programs that work perfectly well in a corporate office fall over on a hospital unit.

A written mentoring agreement fixes most of it. The agreement should state the purpose of the relationship, the expected duration (six to twelve months is the usual range), the meeting frequency, and what both people have committed to. It gives either party a legitimate reason to raise the subject when meetings stop happening, which people otherwise find socially awkward and therefore avoid.

On cadence, be realistic rather than aspirational. Monthly is the honest floor for shift-based staff, and anything scheduled more often than that will be missed. A pair that misses three sessions in a row usually stops trying. Build asynchronous options into the design rather than treating them as a fallback, because a recorded voice message left at the end of a night shift is a real mentoring interaction and it is often the only one available that week.

One more rule, and it is not negotiable. The mentoring record is a personnel record rather than a clinical one. Patient identifiers and case details do not belong in mentoring notes, messages, or session summaries, and saying so explicitly in the agreement prevents most of the conversations you would otherwise have with your privacy office six months in.

Step Five: Build the Leadership Bench Deliberately Rather Than by Accident

Health systems promote clinically excellent people into leadership roles and then provide very little preparation for the actual work of leadership, which is scheduling, budgets, performance conversations, and absorbing pressure from above without transmitting all of it downward. The predictable result is that a strong clinician becomes a struggling manager and then leaves the role or the organization.

A mentorship program can address that, provided it is designed for it rather than assumed to produce it. In practice that means identifying charge-nurse and manager candidates one to two years before you expect the vacancy, then pairing them with someone who is one or two rungs ahead rather than five.

Knowledge transfer is the other half of the same problem. When an experienced clinician retires, a great deal of institutional knowledge leaves with them, including how a particular escalation pathway really works and which colleague to call at three in the morning, and almost none of it is written down anywhere. Pairing people approaching retirement with mid- career staff, and giving those pairs a specific brief rather than a general one, is the only reliable way I have seen organizations capture it.

Step Six: Keep the Measurement Set Small

Program dashboards tend to expand until they measure everything and prove nothing. Keep the set small and keep it comparative.

• Retention of participants against a matched non-participant group in the same units and tenure bands. Without the comparison group, the number is not evidence of anything.
• Relationship survival, meaning the share of pairs still meeting at three, six, and twelve months. This is the early-warning indicator, and it moves months before retention does.
• Progression, meaning the share of participants who moved into a target role or took on additional responsibility within the program window.

Collect meeting frequency and goal completion passively wherever the platform allows it, because asking clinicians to fill in a monthly reporting form is a reliable way to lose them. Qualitative material is worth gathering as well, but treat it as illustration rather than evidence.

Be careful about what you claim. The 2024 National Nursing Workforce Study from the National Council of State Boards of Nursing found that 39.9% of registered nurses reported an intent to leave the workforce within the next five years. A mentorship program does not fix a number that size on its own, and saying that it will is how program owners lose credibility with a finance committee. What it can move is the portion of that intent that comes from feeling unsupported and unclear about a career path, which is one of the few portions a program owner actually controls.

What Usually Goes Wrong in Year Two

Year one is easy, because there is enthusiasm, a launch, and a cohort of volunteers who genuinely wanted to be there. Year two is where programs are lost, and the cause is nearly always operational rather than conceptual.

Mentor supply runs out. The first cohort recruits the willing and the second has to recruit the merely available. Build mentor recognition in from the start, and count mentoring hours toward professional development and clinical ladder requirements where your structure permits it.

The administrator leaves. A program that lives in one person's head and one person's spreadsheet does not survive that person moving roles, which in healthcare happens constantly. Document the process and keep the participant data somewhere institutional.

Nobody reported on it. A program that produced no numbers at the end of year one becomes an easy line to cut when budgets tighten.

The Short Version

• First-year turnover is an onboarding and support problem rather than a recruiting problem, and the 22.7% first-year RN turnover figure reported by NSI for 2026 is where the argument for a program starts.

• Be literal about definitions, because preceptorship, residency, mentorship, and coaching solve different problems at different points in a clinician's first five years.

• Write down a purpose specific enough to be measured, give the program a named owner with unit-level visibility, and start with one cohort where attrition is genuinely expensive.

• Pilot at forty to eighty pairs rather than launching enterprise-wide.

• Match on demonstrated competencies rather than job titles, and decide deliberately whether pairing is self-directed, administrator-assigned, or a combination of the two. Use a written agreement, set a monthly cadence for shift-based staff, and treat asynchronous contact as a legitimate part of the design.

• Identify leadership candidates one to two years ahead of the vacancy, and give pre-retirement pairings a specific knowledge-transfer brief.

• Measure participant retention against a matched comparison group, relationship survival at three, six, and twelve months, and progression into target roles.

• Plan for year two before year one ends, because mentor supply, administrator turnover, and missing numbers are what actually kill these programs.

None of this is complicated, and that is rather the point. The organizations that get value out of mentoring picked a narrow problem, matched people carefully, made the meetings possible given how clinical schedules actually work, and then kept reporting on it long enough for the numbers to mean something.