Per unit
Lease events, work orders, turns, AR aging, income recertifications, leads, deposit dispositions. Work that arrives with a door.
Scales on unitsMethodology · the portfolio model
The savings number, with the math shown.
Every vendor in this category asserts a savings figure and almost none publish the derivation. Here is mine: the 51-automation table, the four drivers each line scales on, the hourly rate and where it came from, and the three assumptions I think are weakest. Then you run your own portfolio through the same model.
Go straight to the model Runs in your browser. Nothing is sent anywhere.
Read-only into Entrata
Reconciled 06:00 ET
One portfolio, task by task, then cut by forty to seventy percent.
The raw total across eight domain lenses was about 2,400 hours a month. A feasibility reviewer and a simulated owner-operator went at it line by line. What survived is about 1,110. The source document calls its own result directional. So does this page.
51Automations counted
~1,100Hours returned / month
0.18Hours / unit / month
~$500KLabor value / year
The study started with roughly ninety candidate automations proposed across eight domain lenses: leasing, maintenance, compliance, accounting, reporting, lease-up, vendor management and governance. Around twenty were killed outright. Seven duplicate families were collapsed into one line each. Every surviving hour claim was then challenged and deflated, most of them by 40 to 70 percent, until a reviewer whose job was to say no stopped saying it.
What is left is 51 automations with a per-item hour figure, a suite, an availability tier, and an assignment to the driver that actually generates the work. That table is the entire basis for every number on this site. There is no second, friendlier model behind it.
Read the pricing those hours are measured against on the pricing page, and what the integration is contractually allowed to touch on trust and security.
What is published here, and what is not. The measurement came from one operating portfolio. Which operator it was, and where, is not published on this page, will not be published anywhere, and does not come out on a call either. That one is not mine to give away, and a vendor who trades one operator's numbers for a sale will eventually trade yours. What is publishable is the normalized result: about 0.18 hours per unit per month, or roughly 180 hours a month per 1,000 units, across all 51 automations. Multiply that by your own door count for a back-of-envelope figure. The model below does the same arithmetic properly, driver by driver, and lands lower for a large portfolio and higher for a small one.
These are modeled estimates from a published methodology. They are not guaranteed results and they are not a quote. Your organization is not the one that was measured. Change the inputs until the model reflects how you actually run, then judge it.
Run your own portfolio through it.
Nothing is sent anywhere: no request leaves your browser, no analytics event, no storage. Your inputs are encoded in the address bar once you change something, so the exact scenario you ran is a link you can send to your CFO.
4,000 units across 20 communities (200 units each) · 2 in lease-up over the next 12 months · 50% income-restricted (2,000 units) · loaded ops hour $38 · capacity realized as budget relief at 100% · 173-hour FTE-month · system of record Entrata · no third-party management · counting available now (466 hrs) plus phase 2, months 4 to 9 (209 hrs).
Capacity returned
Modeled estimate676Hours / month
3.9FTE equivalent
Staff hours currently spent compiling, chasing, checking and remembering, that a machine can do overnight. That is 0.169 hours per unit per month at your size. It is capacity absorbed, not a headcount reduction plan, and the source study is explicit that it must never be sold as one.
The money case
$4.00 / unit / month- Capacity returned, valued at $38/hr676 hrs × 12 months
- $308,065
- NOI Engine subscription4,000 units × $4.00 × 12
- −$192,000
- One-time implementationFixed at signature, not time and materials. Billed half at signing, half at go-live.
- −$61,200
- Net, year one
- $54,865
- Net, year two onwardNo implementation fee after year one
- $116,065
- Payback on the implementation feeImplementation divided by monthly net
- 6.3 months
- Three-year netOne implementation, three years of subscription
- $286,996
- Labor case breaks even atThe share of recovered capacity you must genuinely convert to budget relief
- 62%
Read that last line carefully. You would need to convert 62% of the recovered capacity into actual budget relief for this to pay for itself on labor alone. That is the honest hurdle. If you would redeploy the time rather than remove the cost, drag the realization dial down and watch these numbers change. They should. The source study is explicit that this is capacity, not a headcount reduction plan.
Where the hours come from
8 suites| Suite | Hrs / mo | FTE | Value / yr |
|---|---|---|---|
| Platform and GovernanceApproval queue, notification router, reconciliation harness, counsel-owned template library. | 17 | 0.1 | $8K |
| Lease-Up Command CenterThe 229-task pre-opening checklist, contract abstracts, renewal watchdog, collateral drafting. | 70 | 0.4 | $32K |
| Reporting and AlertingThe weekly ops packet per community, exception engine, revenue leakage, delinquency, covenant headroom. | 116 | 0.7 | $53K |
| Compliance GuardRent-cap guarding on every lease event, set-aside mix, recert deadlines, file audit, COI and permit calendars. | 98 | 0.6 | $45K |
| Maintenance and Turn MoneyTurn watchdog and the down-unit lost-rent meter, SLA breach escalation, work-order triage, preventive maintenance. | 113 | 0.7 | $51K |
| Money-Side AuditsLate-fee consistency, concession approval trail, tax and insurance renewal exposure. | 19 | 0.1 | $9K |
| Leasing Funnel and RenewalsRenewal watchdog, application pipeline, deposit disposition deadlines, draft-for-approval prospect work. | 183 | 1.1 | $83K |
| Financial Narrative and ReportingVariance commentary with GL citations, after-hours drafts, invoice duplicate detection, owner and LP reporting. | 61 | 0.3 | $28K |
| Total | 676 | 3.9 | $308K |
The Lease-Up Command Center scales on lease-up count, not door count. You entered 2, which drives 59 hours a month of the total. Set it to zero and watch most of that suite go. About 66 hours a month of the total exists only because you carry 2,000 restricted units, and scales on that count rather than your full door count.
What is available when
Stated honestly, because it will not survive diligence otherwise| Wave | Automations | Hrs / mo | Counted here |
|---|---|---|---|
| Available now | 36 automations | 466 | Counted |
| Phase 2 · months 4 to 9 | 13 automations | 209 | Counted |
| Later · 12 months and beyond | 3 automations | 55 | Not counted |
Presenting all 51 as shipping today would not survive your diligence and would deserve not to. The three in the last wave each need either counsel work you control, a reconciliation record that has to be earned, or write access I will not ask for. None of them can be scheduled.
Hours are the baseline. The catches are the point.
Four events, each priced by arithmetic you control, each compared against what the platform costs you with implementation amortized over three years. Every input below is editable and none of them is a statistic about anyone. The defaults are deliberately small: one error, one contract, one unit, one waiver, priced at the low end. And the caveat first: NOI Engine does not guarantee it catches any of these. It runs the check nightly and puts a named human in front of the answer.
One over-cap AMI lease
Cap errors are almost never one-off. One wrong AMI tier, one stale utility allowance or one missed annual limit update applies to every unit in that floor plan until somebody catches it, usually at the annual file audit. The rent-cap guard re-checks every lease event against the current cap table nightly. Deterministic, read-only, no model anywhere in the pass or fail.
$/mo over cap×units on the same wrong cap×months before discovery
$9,000against $212,400 a year of platform cost, amortizing implementation over three years. That is 4% of a year of the platform.
Restitution arithmetic only. It does not price the agency finding, the corrective action plan, the re-audit, or the conversation with your equity partner about why the compliance file was wrong.
And the default above is a small one, found early. Cap errors usually are not. A wrong tier attaches to a floor plan rather than to a lease, and it surfaces at an audit rather than at signature. Put $150, 60 units and 24 months into the three fields above, which is one floor plan caught at the second annual file audit, and this single catch comes to $216,000, or 1× a full year of the platform at your portfolio size. That is the arithmetic behind the sentence on the home page.
One missed auto-renew notice window
Service contracts carry 120, 90, 60 and 30-day notice windows, and at most operators they are tracked in somebody's memory and a folder. Miss the window and you are committed for another full term at a rate you had already decided to rebid. The contract watchdog reads the abstract on intake and starts the countdown the day the PDF lands.
$/yr contract you meant to rebid·% you would have saved on rebid
$3,600a year you would have kept, locked inside a $30,000 commitment you can no longer exit for another twelve months. That is 2% of a year of the platform.
The smallest of the four on a single contract, and the one that is never a single contract. Multiply by however many service agreements sit across 20 properties. Landscaping, trash, pest, elevator, pool, fire, snow, laundry. The $30,000 you can no longer exit is the figure the source study means when it calls one missed notice window a five-figure catch. The $3,600 is the narrower number, and it is the one shown above, because the other three blocks show losses and this one should be comparable to them.
One bleeding turn
A down unit costs you rent every single day and nobody gets an alert on day nine. The turn watchdog runs a per-unit lost-rent meter and escalates when a make-ready stalls, so the sentence a regional actually reads is not “unit 214 is late.” It is “unit 214 has cost you $3,389 so far.”
$ average monthly rent, which is $53 a day×days past target×stalled turns a year
$19,066against $212,400 a year of platform cost, amortizing implementation over three years. That is 9% of a year of the platform.
The source study used a $53 per day meter, which is a $1,610 average rent. The default above reproduces it exactly. One stalled turn at your numbers is $1,589. This is also the one catch that lands directly in occupancy, so your asset manager will already believe it.
One covenant surprise
DSCR gets tested on a date somebody else picked. Finding out you will miss it on the test date leaves you negotiating. Seeing it 60 to 90 days out leaves you managing the quarter. The covenant projection runs headroom forward off the trailing ledger every week.
$ loan balance×bps waiver or amendment fee
$50,000against $212,400 a year of platform cost, amortizing implementation over three years. That is 24% of a year of the platform.
This prices the waiver fee and nothing else. Not the default-rate step-up, not the cash sweep, not the lender relationship, and not what a technical default does to your next refinancing. Those are the expensive parts and none of them is in this number.
These are deliberately not added together. Summing four hypotheticals into one headline is how vendor ROI calculators earn the discount every CFO already applies to them. So read what the four percentages above actually say at the defaults, which is that no one of these small single events pays for the platform by itself. The claim is narrower than that and it is the only one worth making: these are the same kind of money as the platform rather than a rounding error beside it, they recur, they are the events nobody has a report for, and not one dollar of them is counted anywhere in the hours above. Scale any one of them to the size the event usually arrives at, which the first block works through, and the arithmetic changes character.
Every assumption, on the page.
Savings do not scale linearly with door count, because the work does not. A weekly ops packet is produced once per community whether it has 90 units or 400. A board one-pager is produced once for the portfolio. A cap check runs once per lease event. So every one of the 51 automations is decomposed across four drivers and each driver is scaled on its own.
Per community
Weekly ops packets, permit and license calendars, preventive maintenance schedules, property P&Ls, profile listings, insurance certificate rosters.
Scales on propertiesPortfolio fixed
The board one-pager, covenant reporting, the LP packet, the fair-housing template library. One organization, one copy, and it grows far slower than the portfolio does.
Fourth root of unitsPer lease-up
The 229-task pre-opening checklist, day-zero provisioning, lease-up against pro forma, opening collateral. A discrete project with a fixed task list.
Strictly linearFour drivers, five scalars
hours(i) = base(i) × (1 − a) × (u·sU + c·sC + f·sF + l·sL)
+ base(i) × a × sA
u + c + f + l = 1 on every item. a is the share of that
item that exists only because restricted units exist.
sU = damp( units / ref )
sC = damp( (properties / ref) × (avg units per property / ref)0.25 )
sL = lease-ups / ref
sF = ( units / ref )0.25
sA = damp( restricted units / ref )
damp(x) = x if x ≤ 1, else x0.90
The reference denominators are the basis the item table is expressed against: a unit of measure, not a claim about anybody. Two of them, the lease-up count and the restricted share, are modeling assumptions rather than measurements, and both are flagged as weak further down. They sit in this page as ordinary constants rather than behind a login, because a derivation you cannot run is a brochure. Whose portfolio produced them is the part that stays private.
Why each is shaped the way it is
Damping above the reference, never below. A 15,000-unit operator has usually already built a shared service center, standardized a reporting pack and absorbed some of this through process. They are not 2.5 times as manual as a mid-sized shop, so the 0.90 exponent takes roughly ten percent off at that size. Below the reference the model scales strictly linearly and does not apply the mirror-image uplift, even though smaller operators demonstrably carry more manual work per door. Crediting a 1,200-unit shop for being less automated would inflate the number in exactly the segment where the price case is weakest. So it does not. If you run a small portfolio and tell me the model understates you, you are probably right, and I will say so.
Why the per-community term carries a property-size factor. A 50-unit property still generates its own weekly packet, its own permit calendar and its own maintenance schedule, but that packet is genuinely less work than a 400-unit property's. Without the fourth-root size term, a scattered-site portfolio of 300 small properties scores absurdly high.
Why lease-ups are strictly linear. A lease-up is a discrete project with a fixed task list. Two lease-ups is twice the work of one. There is no economy of scale to credit.
Why portfolio-fixed work uses a fourth root. It is not truly fixed. A larger operator has more entities, more lenders and more capital partners, so the board packet and the covenant reporting are genuinely bigger, but nowhere near proportionally bigger.
The hourly rate, and why the default is $38
It is derived, not chosen. The published claim pairs about 1,110 hours a month with about $500,000 a year of loaded labor value. $500,000 divided by 13,344 annual hours is $37.48, which rounds to $38. Running it back the other way, 1,112 hours times 12 months times $38 is $507,072, which is the same $500,000 claim. One derivation reproduces both published headline figures, which is the point of using it rather than picking a flattering number. A CFO can check the two claims against each other and find they are internally consistent.
What $38 represents: a blend across community managers, leasing consultants, maintenance technicians and the corporate roles that touch this work, fully loaded with payroll taxes, benefits and overhead. It sits at the conservative end for a blended multifamily operations rate.
Income-restricted work is scaled separately
Some of the work exists only because restricted units exist, and it scales on the count of restricted units rather than total doors. Six automations carry a restriction weight: rent-cap guarding, set-aside reporting and recert deadlines at 100 percent, lease-file audit sampling at 25 percent, the renewal-offer compliance sweep at 40 percent, and the renewal decision-support worksheet at 15 percent. Turn the affordable switch off and every one of those hours leaves, and the over-cap catch is replaced with the words “not applicable to you.” It is the single largest catch in the product and it is worth nothing to a pure market-rate operator.
Sanity-check the implied per-file rate, because this is where a compliance-literate buyer will push. The recert line works out to roughly eight minutes per file of deadline tracking, notice templating, checklist management and 30-day escalation. The rent-cap line works out to under nine minutes per lease event to check contract rent plus utility allowance against the current cap table. Anyone who has run a recert queue will tell you both are conservative, and it is a machine doing it.
Pricing and implementation
- Subscription
- Units × $4.00 × 12. The published combined rate: analytics plus all 51 automations.
- Implementation
- $45,000 + $30,000 × clamp((units − 1,000) / 14,000, 0, 1)0.40. Fitted to the published $45,000 to $75,000 band. Fixed at signature, half at signing, half at go-live.
- 1,000 units
- $45,000, the stated floor
- 3,000 units
- $58,775
- 8,000 units
- $67,736
- 15,000 units
- $75,000, the stated ceiling
The $1,500 a month platform minimum is not modeled, because $4.00 per unit clears it at every portfolio size in the target band. Analytics on its own is $1.50 per unit per month. It is available, and it is the least valuable way to buy this: dashboards do not catch the over-cap lease or write the Monday packet. It is deliberately not offered as a column here.
The number I would rather you heard from me.
Recovered hours are capacity, not cash. They become dollars only if a role is removed, a planned hire is avoided, or somebody moves onto revenue work. So the model always prints the share of that capacity you would have to genuinely convert into budget relief for the subscription to pay for itself on labor alone.
| Units | Communities | Lease-ups | Hrs / mo | Breakeven, now plus phase 2 | Breakeven, available now only |
|---|---|---|---|---|---|
| 1,200 | 6 | 1 | 239 | 53% | 73% |
| 3,000 | 15 | 2 | 532 | 59% | 85% |
| 8,000 | 40 | 4 | 1,275 | 66% | 96% |
| 15,000 | 74 | 7 | 2,199 | 72% | 105% |
Read the last column. The 36 automations available today do not pay for themselves on labor alone at published pricing in year one at the top of the band. That is not a rounding artifact, it is the finding. It is also exactly why the conclusion in the source study is not "sell hours." It is sell decision speed, vendor displacement and the catches.
And read the bottom of the band. At 1,200 units the labor case in year one is essentially breakeven and goes negative at any realization rate under 100 percent. The economics at the small end only work if the catches carry the decision on their own. That is a legitimate way to buy this and it is also a legitimate reason not to.
Per-door sanity, which any CFO can check in one division. Across that same band the model returns $91 down to $67 of modeled labor value per door per year against a price of $48 per door per year. A ratio of 1.4 to 1.9 times. That is modest, and modest is what makes it checkable. A vendor calculator that returns eight times should be assumed to be measuring something it invented.
The three assumptions I consider weakest.
Flagged here before you find them. A vendor who marks his own soft numbers is the only kind this audience has any reason to believe.
The reference share of income-restricted units
The weakest input in the model. The source study describes a mixed book but never states the split, so the reference share behind the compliance scaling is my assumption rather than a measured figure. The whole Compliance Guard suite hangs off it. If the true reference share is much higher, a heavily restricted prospect is understated here. Mitigation: the share is your field, not mine. Set your own and the suite rescales.
DisclosedThe reference lease-up count
Also not in the source. The Lease-Up Command Center is a meaningful share of the total, so this assumption moves roughly six percent of the number. Setting it too high in the reference makes the per-lease-up rate too low, which understates a developer-operator with many deliveries and overstates nobody. A fully stabilized owner loses about eight percent of the total and should.
DisclosedThe driver mix on the four largest line items
Four items are about fifteen percent of the total between them and their per-unit against per-community splits are reasoned, not measured. The exception alert engine in particular could defensibly sit anywhere from 30/60/10 to 70/20/10, which moves the total by roughly two percent at extreme portfolio shapes. Bounded, small, and small relative to the 40 to 70 percent deflation already applied upstream.
DisclosedAlso honest about two smaller ones. The go-live verification bot scales on total community count as a proxy for new-property count, which slightly overstates it for a stabilized owner. Restriction-driven hours all scale on restricted-unit count, including the set-aside reporter which is partly per-property, which slightly understates a portfolio of many small restricted properties. Both effects are under one percent and they run in opposite directions.
What the model deliberately leaves out.
All of it stated before you ask, because every item on this list is one you were about to raise.
- All revenue lift. No occupancy gain, no loss-to-lease recapture, no bad-debt reduction, no renewal-rate improvement. These are real and very probably larger than the labor line. They are also unmeasured, so they are worth zero here.
- Your own implementation time. Roughly twelve hours across three weeks: one read-only API user provisioned by IT, two ninety-minute mapping sessions with your accounting lead, one template review with counsel, an hour of training per role. Not netted out of any figure above.
- Any addition of the catches to the labor value. The two arguments are kept structurally apart so you can reject either one without rejecting both.
- Work you are not doing today. Automating a task nobody currently performs returns no hours. It may still be worth buying. It is not worth counting.
- Vendor displacement. The study valued one recurring report against a real competing quote. That figure belongs to the operator that was measured and is indirectly identifying, so it appears nowhere in this model and it will not be quoted in any public setting.
And one refusal that matters more than any of them. NOI Engine does not guarantee it catches the over-cap lease, the notice window, the stalled turn or the covenant miss. It runs each check nightly and puts a named human in front of the answer before anything leaves the building. Overclaiming detection on a compliance-sensitive product is how a vendor acquires liability it never priced, and you would be right to discount anyone who does it.
The scope of what the integration can and cannot touch is a contract term, not a policy page. It is written out on trust and security.
Six challenges, and the answer to each.
These are the ones I actually get. A buyer who bothers to attack the math is a buyer who is considering it.
Where did 1,100 hours come from?
A task-by-task study of one operating portfolio across eight domain lenses, then a feasibility reviewer and a simulated buyer who cut the raw claims by 40 to 70 percent. Raw was about 2,400 hours a month. The honest figure is about 1,110. The suite-level breakdown is in the model above, and the itemized rows sum two hours higher than the summary table rounds to. I use the itemized sum, because that is the one you can check line by line.
That $38 rate is convenient.
It is derived, not chosen: $500,000 divided by 13,344 hours. If your loaded rate is $52, put in $52. The field is right there, and the number will go up, which is not an argument in my favor.
You are just multiplying by our door count.
No. Fifty-one line items, four drivers, five independent scalars. Your community count moves this as much as your door count does. Set lease-ups to zero and watch most of a suite leave the total.
So we fire six people?
No, and the source study explicitly forbids that framing. It is capacity, not headcount. Which is why the model prints the realization rate you would have to hit to convert it into actual budget relief, and why the realization dial defaults to the optimistic end rather than hiding there. Drag it down and look at what happens.
These vendor calculators always show a big number.
Then set realization to fifty percent and watch this one go red. Or uncheck phase 2 and look only at what is available today. At the top of the target band it does not clear on labor in year one, and I would rather you hear that from me now than work it out in month four.
What is not in here?
All revenue lift, your own implementation hours, and any work you are not already doing. The full exclusion list is on this page, above. I did not model the things I cannot evidence.
The fastest way to judge the math is to watch it run.
The demo is the whole platform on a fictional portfolio, with no form and no call. If you would rather argue with the model first, that is the better use of an afternoon and I will not be offended.
Michael Gaff · michael@michaelgaff.com · (317) 985-2446