Most underwriting models treat a renewal and a new lease as roughly the same event: a signature that keeps space occupied. They are not the same event. A renewal keeps income flowing across an expiration with little downtime and thin concessions. A new lease pays for months of vacant space, a full tenant improvement package, a leasing commission, and free rent to backfill the same square footage. Tenant retention is not a soft, tenant-relations line item. It is the single largest controllable variable in the re-leasing cost of a property, and models that assume a flat "re-leasing cost" per rollover systematically overstate what keeping a tenant costs and understate what losing one does.
The thesis is narrow and defensible: the cost gap between a renewal and a new lease is large, recurring, and knowable in advance, and any model that collapses the two into one number is mispricing rollover.
Key Takeaways
Tenant retention is the largest controllable input to re-leasing cost. A renewal and a new lease produce the same occupied square footage at very different net costs, and the gap is concentrated in downtime, tenant improvement allowances, leasing commissions, and free rent.
New office leases carried an average of 9.0 months of free rent and a $94.69 per square foot tenant improvement allowance in the first half of 2024, per CBRE. Renewals typically carry far less of both.
Multifamily turnover is not free either. The National Apartment Association places the cost of a single turn, including rent loss, starting around $1,000 and reaching $2,500 to $5,000 once capital replacements are counted.
Renewal probability is an underwriting input, not an afterthought. A model that assumes market re-leasing costs on every rollover ignores that a meaningful share of tenants renew at a fraction of that cost.
The correct question is not whether a tenant renews but what the net effective cost difference is between keeping and replacing them. That difference belongs in the model, tenant by tenant.
Why is tenant retention cheaper than a new lease?
Tenant retention is cheaper because a renewal skips the most expensive parts of re-leasing. There is little or no vacancy between the old and new term, the tenant improvement package is a refresh rather than a full build-out, no new leasing commission is owed on a direct renewal, and free rent is minimal. A new lease pays for all four.
Each of those four costs is real and separately underwritten. Downtime is lost rent for every month the space sits empty between the departing tenant and the paying replacement. Tenant improvement allowance is landlord capital spent to fit out space for a new tenant, and a fresh build-out costs far more than refreshing space a renewing tenant already occupies. Leasing commissions run a percentage of total lease value and are owed on new deals; a direct renewal often owes none. Free rent is months of occupied-but-unpaid space granted to win a new tenant. Stack the four together and the reason retention is cheaper stops being a matter of opinion. It is arithmetic.
The market data confirms the direction and rough magnitude. Per CBRE's analysis of 3,900 lease transactions across 12 major U.S. markets, new office leases in the first half of 2024 carried an average of 9.0 months of free rent and a tenant improvement allowance of $94.69 per square foot. Those concessions are what a new lease pays and a renewal largely avoids. This is the core of the net effective rent calculation: the gap between face rent and what the landlord actually nets after concessions.
What does the renewal versus new-lease cost stack look like?
The cost stack is the itemized difference between renewing a tenant and replacing them. A renewal typically carries minimal downtime, a light refresh allowance, no new commission, and little free rent. A new lease carries months of vacancy, a full build-out allowance, a full commission, and multiple months of free rent. The same occupied space, two very different costs.
The table below lays out the four cost lines as a representative range for a mid-size office tenant. Treat the figures as illustrative ranges, not fixed benchmarks; every market, tier, and asset differs. The point is the structure, not the specific dollar.
Cost line | New lease | Direct renewal |
|---|---|---|
Downtime (vacant months) | 6 to 12 months typical on full-floor space | Near zero; term rolls continuously |
Tenant improvement allowance | Full build-out, into the $90s per sq ft in major markets (CBRE, H1 2024) | Light refresh, a fraction of a full build-out |
Leasing commission | Full commission on total lease value, often 4 to 6 percent | Often none on a direct renewal |
Free rent | Multiple months; averaged 9.0 months in new office leases (CBRE, H1 2024) | Minimal to none |
Every one of these lines compounds with lease term and space size. On a large full-floor requirement, downtime alone can run 6 to 12 months, which means half a year or more of rent collected from no one while the build-out proceeds. That is why concentration of expirations matters as much as the expirations themselves, the subject of rollover risk. A single tenant lost is a leasing cost. A cohort of tenants lost in the same year is a capital event.
How should renewal probability enter the underwriting model?
Renewal probability is the odds that a given tenant renews rather than vacates at expiration, and it belongs in the model as an explicit input on every rollover, not a blanket assumption. A model that applies full market re-leasing costs to every expiring lease assumes a zero percent renewal probability, which no real rent roll matches. Weighting each rollover by its renewal odds is the correction.
The mechanics are straightforward. For each expiring lease, the expected re-leasing cost is the renewal-path cost times the renewal probability plus the new-lease-path cost times the probability of loss. If a tenant has a 65 percent chance of renewing, the model should carry 65 percent of the cheap renewal path and 35 percent of the expensive new-lease path, not 100 percent of the expensive one. Applying the worst case to every rollover inflates modeled re-leasing reserves and can make a stable, sticky rent roll look riskier than a churning one that merely reports similar occupancy today. Retention shows up in durability, not in a single snapshot, which is why weighted average lease term is a better proxy for cash-flow quality than occupancy alone. The tenant retention rate is the historical anchor for the probability you assign forward.
The quotable version: renewal probability is not a market assumption, it is a property fact, and a model that ignores it is not conservative, it is wrong in a specific and costly direction.
Worked example: the net cost of losing versus keeping a tenant
The net cost of losing a tenant is the full new-lease stack minus what a renewal would have cost. Below is a worked example for a single 20,000 square foot office tenant. Every figure is a stated input; the result is derived, not asserted.
Assume face rent of $40 per square foot per year, so annual rent is $800,000, or roughly $66,667 per month. Inputs:
Space: 20,000 sq ft
Face rent: $40 per sq ft per year ($800,000 annual)
Downtime if lost: 9 months
New-lease tenant improvement allowance: $90 per sq ft
New-lease free rent: 9 months
Leasing commission: 5 percent of a 5-year lease value
Renewal refresh allowance: $15 per sq ft, no commission, no free rent, no downtime
New-lease path, cost to backfill:
Downtime: 9 months x $66,667 = $600,000
Free rent: 9 months x $66,667 = $600,000
Tenant improvement: 20,000 x $90 = $1,800,000
Commission: 5% x ($800,000 x 5 years) = 5% x $4,000,000 = $200,000
New-lease total: $3,200,000
Renewal path, cost to keep:
Refresh allowance: 20,000 x $15 = $300,000
Downtime, free rent, commission: $0
Renewal total: $300,000
The net cost of losing this tenant rather than renewing them is $3,200,000 minus $300,000, or $2,900,000. On $800,000 of annual rent, the difference between keeping and replacing a single tenant is roughly 3.6 years of that tenant's rent. Now weight it by renewal probability. At a 65 percent chance of renewal, the expected re-leasing cost is 0.65 x $300,000 plus 0.35 x $3,200,000, which equals $195,000 plus $1,120,000, or $1,315,000. A model that ignored retention and assumed the full new-lease path would have reserved $3,200,000 for the same rollover, overstating the cost by nearly $1.9 million on one lease. That is the mispricing the thesis names.
Frequently Asked Questions
Is tenant retention actually cheaper, or does a renewal just defer costs?
It is cheaper, not merely deferred. A renewal avoids downtime, a full build-out, a new commission, and free rent outright; those costs are not pushed to a later date, they are never incurred because the space never turns. A renewing tenant may negotiate a modest refresh allowance or a rent adjustment, but that is a fraction of the full new-lease stack.
What is a reasonable renewal probability to assume in underwriting?
There is no universal figure, and inventing one would be a mistake. Renewal probability depends on tenant type, space size, lease structure, and market. The correct practice is to anchor the assumption in the property's own historical tenant retention rate and adjust for known factors, then apply it per lease rather than as a single portfolio number.
Does this apply to multifamily as well as office?
Yes, though the magnitudes differ. Multifamily turnover costs are smaller per unit but occur far more often. The National Apartment Association places a single turn, including rent loss, starting around $1,000 and reaching $2,500 to $5,000 with capital replacements. Across hundreds of units and annual turnover, retention compounds into a major controllable expense.
Why do models get this wrong so often?
Because a flat per-rollover re-leasing cost is simpler to build than a tenant-by-tenant renewal-probability weighting. The simplification is convenient and directionally conservative-looking, but it prices every tenant as a certain departure, which overstates reserves on sticky rent rolls and hides the value of retention.
Conclusion
Tenant retention is cheaper than your underwriting model assumes because the model, in most cases, was never built to price it. A renewal and a new lease are not two versions of the same event. One keeps income flowing across an expiration for the cost of a refresh; the other pays for downtime, a full build-out, a commission, and months of free rent to reproduce the occupancy you already had. The gap between them is large, recurring, and knowable before the lease expires. For the asset manager, the discipline is to stop modeling rollover as a single blended cost and start modeling it as two paths weighted by renewal probability, tenant by tenant. Do that, and retention becomes what it always was: the biggest lever you control over the true cost of holding the asset.