CapEx Reserves for Rental Property: A Schedule, Not 5%

Why 5% of rent and $250 per unit are guesses. Build a component CapEx schedule for a rental property, with two fully worked reserves — 9.30% and 33.20%.

CapEx is not a percentage of rent. It is a schedule of dated component replacements, and the percentage is a proxy you use only until you have walked the building and written that schedule down.

The two properties below carry identical replacement costs on nine components and identical $1,900 rent. One derives a reserve of 9.30% of gross rent. The other derives 33.20%. Neither is 5%, and no rule of thumb could have told them apart.

A repair keeps a component running; a capital expenditure resets its clock

Maintenance is scheduled work that extends life without replacing anything — servicing, caulking, gutter cleaning, drain clearing. Repairs are unscheduled fixes that return a component to working order: a failed capacitor, a leaking supply line. Both are small and frequent enough to average, so a percentage of collected rent is a fair way to budget them.

Capital expenditures replace a component or materially extend its life — roof, furnace, service panel, kitchen. They are large, infrequent and lumpy, and lumpy costs cannot be averaged into a percentage without badly over- or under-funding.

The tax treatment splits along roughly the same line. IRS Publication 527 treats repairs that keep a property in ordinary operating condition as deductible in the year paid, while improvements are capitalised and recovered through depreciation over 27.5 years for residential rental property. CasaWise models annual depreciation as (Purchase Price x 80% + Rehab) / Depreciation Years on the standard 20/80 land-to-building split, published on the calculations page. Where a specific expenditure lands is a question for your CPA, not for a calculator.

One point that trips people: the reserve itself is not deductible. Setting aside $2,120 changes nothing on your return. The deduction arrives later, over 27.5 years, when the money is actually spent. A reserve is a cash-flow instrument, not a tax one.

The percentage method fails because a roof does not know what the rent is

The percentage exists because it is fast and needs no information. On a screening pass over forty listings it is defensible. As a funding number it breaks on one fact: CapEx is indexed to the building, and the percentage is indexed to the rent.

Take one physical house — 1,350 sq ft, three bedrooms, two baths — and put it in two markets. Assume shingle replacement at $11,500 on a 25-year cycle: $460 a year of roof consumed whether anyone is paying rent or not.

Same shingles, same square footage, same crew, same dumpster. The budget doubled. The roof did not. Across all nine components below, this building consumes $3,581.39 a year — 21.32% of rent at $1,400 and 10.66% at $2,800. The 5% rule under-funds the first by $2,741.39 a year and the second by $1,901.39. Over a ten-year hold: $27,414 and $19,014 of capital never budgeted.

The per-unit rule survives the test better. At $250 per unit per month — $3,000 a year — you land within 84% of $3,581.39, because at least it is indexed to a building rather than a rent roll. Still a guess. Better-shaped guess.

The component schedule: nine line items and the arithmetic that follows

List every component with a finite life. Assign a replacement cost and a remaining useful life. Sum every replacement that falls inside your holding period. Divide by the years.

The figures below are an illustrative model, not industry data. The costs are placeholders for a 1,350 sq ft single-family; the lives are planning assumptions. Replace every one with quoted pricing from your own contractors and remaining lives from an inspection of the actual building. That substitution is the exercise. The table is the structure, not the answer.

This is where a construction background does real work: CasaWise was built by a licensed general contractor with 30+ years and 54 construction projects behind him, which is why the model treats CapEx as a schedule rather than a slider — background here.

Whole-life consumption, before looking at any specific building:

That is $298.45 a month of capital this building consumes forever. Now apply remaining life over a ten-year hold.

Property A — built 2016, components original, exterior painted two years ago. Roof (15 yrs left), kitchen (15), baths (15), windows (25) and panel (30) all fall outside the hold and contribute nothing.

$21,200 ÷ 10 = $2,120/yr = $176.67/mo = 9.30% of gross rent.

Property B — built 1978, same size and spec, nothing replaced in two decades.

$75,700 ÷ 10 = $7,570/yr = $630.83/mo = 33.20% of gross rent.

Note the paint line: an 8-year cycle starting in Year 1 recurs in Year 9, so it funds twice. A percentage never catches that.

Identical rent, identical replacement costs, reserves 3.57x apart. Property A lands below its own steady-state $3,581.39 because the roof, kitchen and baths fall outside the hold — you hand those bills to the next buyer, and their schedule will show it in their offer. Property B lands far above steady state because it absorbs a full generation of deferred replacement inside one decade.

At 33.20% of rent, Property B does not have a reserve problem. It has a price problem. That $75,700 belongs in the offer or in the rehab line at acquisition — the same logic that governs holding costs on a flip.

One property cannot self-insure a $14,000 event

Reserves are a pooling instrument, and a single property is a pool of one.

Property A accrues $2,120 a year. An unscheduled $14,000 event in Year 3 — sewer lateral, foundation drainage, something the schedule missed — meets a balance of $6,360. You are $7,640 short, funded from a HELOC, a card, or savings, at whatever that costs.

Eight doors on the same schedule accrue $16,960 a year. By the end of Year 3 the pool holds $50,880, pays the $14,000 and keeps $36,880 working. Same expected cost per door, radically different exposure to timing.

Pooling only works if the vintages differ. Buy eight identical 2016 tract houses in one year and the schedules synchronise: every HVAC comes due in Year 8 together. Cumulative accrual through Year 8 is $135,680, cumulative spend on water heaters, flooring and paint before then is $101,600, leaving $34,080 against a $68,000 bill. Short $33,920 in a single year. A pool of eight identical assets is a pool of one, eight times over.

Where the reserve sits, and why CapEx gets its own worst-case modifier

CasaWise subtracts the reserve inside NOI, before any return metric exists:

NOI = Effective Gross Rent - (Fixed Costs + Revenue-Based Costs + CapEx Reserve)

Cap rate, cash-on-cash, DSCR and IRR all inherit it. The formulas are published in full on the calculations page.

Property A at $189,000, 25% down, 6.67% on a 30-year fixed (the average for the week of 13 August 2026), vacancy at the Q2 2026 national rate of 7.3%:

Effective gross rent  $22,800 x (1 - .073) =  $21,135.60
Fixed costs           tax 1.1% + ins 0.45% =   $2,929.50
Revenue-based         8% mgmt + 5% maint   =   $2,747.63
CapEx reserve         from the schedule    =   $2,120.00
NOI                                        =  $13,338.47
Annual debt service   $141,750 @ 6.67%     =  $10,942.35
Annual cash flow                           =   $2,396.12  ($199.68/mo)
Cap rate 7.06%   DSCR 1.219   Cash-on-cash 4.53%

Delete the reserve and the same property reports an 8.18% cap rate, 1.413 DSCR and 8.53% cash-on-cash. That is the seller's pro forma. Nothing about the building changed.

Note the rent ratio this deal required: 1.005%. That is not an endorsement of the 1% rule as a screen — it is the output of solving for a property that clears DSCR 1.20 at 6.67% money with an honest reserve already deducted, which tells you how little room the current cost of debt leaves.

Put Property B's $7,570 reserve on the same $189,000 price and NOI falls to $7,888.47 — DSCR 0.721, cash flow negative $254.49 a month. To clear DSCR 1.20 honestly, Property B has to be bought near $127,000. The $62,000 gap is the deferred capital, and it is not a negotiating position. It is the price.

The three-scenario engine gives CapEx a modifier of its own — x1.4 worst case against x1.2 for operating expenses:

The separation is not cosmetic. Roll CapEx into operating expenses at x1.2 and worst-case NOI reads $10,655.24 at a DSCR of 0.974 — $424 and 0.039 of coverage that do not exist. On Property B's reserve the same shortcut overstates worst-case NOI by $1,514. Capital work overruns wider than routine operating costs because a tear-off finds rotted decking and a panel swap finds aluminum branch wiring. The modifier is wider because the variance is wider.

Scheduling the event instead of averaging it

A level reserve is a compromise. A schedule has dates on it, and dates change IRR. Ultimate AI Investors supports up to two planned capital expenditure events, each deducted in its specified year, alongside additional income streams such as ADU rent — tier detail on the pricing page.

Model Property A's flooring in Year 2 and HVAC in Year 8 as events and the level reserve drops to the remaining $6,200 over ten years, or $620 a year. Base cash flow rises to $3,896.12 — then Year 2 lands at negative $2,603.88 and Year 8 at negative $4,603.88. Total capital spent is identical at $21,200 either way. What changes is when, and IRR is the only metric that prices when. Do it one way or the other, never both, or you fund the roof twice.

A deal that only works if you skip the reserve is a deal that does not work

Property A's cash-on-cash moves from 4.53% to 8.53% by deleting one line. The line is not optional — the water heater fails in Year 1 regardless of what the spreadsheet says.

Every reserve is provisional until you have walked the building. Write the schedule, price it locally, and treat the percentage as what it is: a placeholder you are trying to replace. Then run the deal three ways before you commit, as you would for any acquisition, and check the result against thresholds calibrated to today's cost of money.

Frequently Asked Questions