How this compares
Nobody in multifamily is choosing between eight vendors. You are choosing between keeping the work on your own people, adding a person to do it, and paying a consultancy to build you dashboards. I will tell you where each of them beats me.
Deals in this segment are almost never lost to a named vendor. They are lost to a spreadsheet that already works, to a hire that is easier to approve, and to a proposal from a BI shop. Those are the three costed below, and no competitor is named anywhere on this page.
| Path | What it costs | Where it wins |
|---|---|---|
| Keep doing it by hand | No invoice. About 0.18 staff-hours per unit per month of recurring work, measured task by task. | Under about 1,000 units, and any portfolio where the hours are not actually there. |
| Add people | One multifamily analyst, roughly $91,000 to $102,000 a year loaded.Note 01 | When what you need is judgment: underwriting, ad-hoc analysis, someone in the room during a deal. |
| Buy a BI consulting engagement | A fixed five-figure fee, plus per-seat licensing, plus change orders, plus an internal owner.Note 02 | Five specific reports you can name today that will not change much. |
| NOI Engine | $4.00 per unit per month, plus a one-time implementation of $45,000 to $75,000. | When the work needs watching every night, not reporting on request. |
Each of the first three rows has a stated case where it beats me, and each is written out in full below. They are not throat-clearing. If one of them describes your situation, that is the right answer, and I will say so on the call before you spend anything.
Alternative 01
It looks free, and that is the whole problem with it. It is not on a line in your budget, so it never gets compared against anything.
Here is the only honest way I know to size it. I measured the recurring work on a single operating portfolio in the size range this is built for, task by task: what gets done, how often, by whom, how long it takes. One portfolio is a sample of one, and I will say that before you do. Normalized, it comes to about 0.18 staff-hours per unit per month, which is roughly $82 a door a year of loaded labor. Multiply that by your own door count and you have a starting estimate.
Your number will be different, and I would rather you build your own than take mine. The derivation is published task by task, including the assumptions I think are weakest.
That is the visible cost. The invisible one is the misses, and the misses are where the real money is, because a miss never shows up as a line item. It shows up as a rent cap exceeded for eleven months, a landscaping contract that repriced fourteen percent because the notice window slipped, a turn that sat forty-one days while nobody priced the vacancy. Nobody writes those down. They just show up in NOI and get attributed to the market.
Complete control. No integration. No vendor. No new tool for anyone to learn, and no security review.
Your people know your portfolio, and an experienced regional will catch things no system would think to look for. That is worth more than most software vendors will admit.
Almost every operator I have talked to has one person who knows when things are due. That person is not backed up. When they are out, or when they leave, the checks do not fail loudly. They just stop, and nobody finds out until the consequence arrives.
Your team checks things when they are at work. Lease events, work orders and payments do not stop when your team goes home.
Going from 4,000 to 6,000 units does not add fifty percent to the work. It adds fifty percent to the work and takes the coordination from manageable to not, because the number of things one person can hold in their head is fixed.
The four-hundredth lease file gets a different level of attention than the first. That is not a criticism of anyone, it is just true of people, and it is exactly the argument opposing counsel makes.
If you are under about 1,000 units, keep doing it by hand. The coordination is still small enough for one competent person to hold, and at that size no software is worth the implementation. I will tell you that on the call rather than let you buy something you do not need.
Same answer if the hours are not actually there. Take one week and have three people log what they actually spend on compiling and chasing. If it is small, you have your answer and you should stop reading. I would rather you run that exercise and disqualify me than skip it and buy something that does not pay for itself.
Alternative 02
This is the version most operators land on, and it comes in two shapes: hire an analyst, or ask somebody already on staff to build it. They are the same decision.
The in-house version just pays for it out of a salary that already exists, which makes it look cheaper than it is.
Publicly posted salary data puts a multifamily analyst around $73,000 on average, in a $52,000 to $87,000 band, higher in coastal markets, before loaded cost.Note 01 Loaded, with benefits and taxes, call it roughly $91,000 to $102,000 a year for one. You hire these people and I do not, so if your number is different, yours is the one to use.
The in-house build version costs the same money plus a risk: the person doing it has another job, so the reporting gets built in the gaps, and the gaps close whenever something operational catches fire.
A person, which is more than any software gives you.
An analyst can answer a question nobody anticipated, build a one-off model for an acquisition you are underwriting, sit in a meeting and tell you the number is wrong before it goes in the deck, and be redeployed to whatever is on fire this month. Software does none of that.
Reporting is retrospective and it is triggered by someone asking. Monitoring is continuous and it is triggered by a condition. Your AMI cap exposure, your notice windows and your covenant headroom need the second one, and no human provides it, because providing it means checking every lease event every night forever.
The same failure as doing it by hand, now with a salary attached.
Eighteen months in, the whole thing lives in one head and in a workbook one person can safely modify. When that person takes a better offer, you are not down an analyst, you are down the reporting.
I say this as someone who did it, so take it as a list of what is under the waterline rather than as discouragement.
None of that is impossible. It is four to six months of somebody’s real attention, and it was not in the estimate.
Hire the analyst if what you actually need is judgment: underwriting support, ad-hoc analysis, someone in the room during a deal. That is not what this platform is for and I would not try to sell against it. Several of the operators who most need an analyst also need this, and the combination is better than either.
Build it in-house if you already employ someone with real data engineering experience who has the time, and if your reporting is genuinely unusual. If you have that person and that time, you should build it, and you will end up with something better fitted than anything you can buy.
Alternative 03
This is the proposal that started NOI Engine. The dashboards were not the problem. Some of them would have been good.
A BI shop put a five-figure proposal in front of an operator I was working with: five static dashboards, refreshed on a schedule, with a change order waiting behind every question anyone would think to ask after seeing them.
The fee is the part you see. The rest of it is not in the proposal.
Potentially good dashboards, built faster than you would build them, by people who do this for a living.
If your problem is genuinely that nobody can see the numbers, this solves it, and it solves it in weeks. I am not going to pretend otherwise.
Five of them are five pictures. Nothing in that proposal watches an AMI cap, counts down a notice window, prices a stalled turn, or drafts anything. When the engagement ends, the dashboards do exactly what they did on day one, and they will keep doing exactly that in three years.
The failure mode of dashboard-only reporting is not that the number is wrong. It is that the number was right and nobody opened the tab that week. Monitoring pushes. Dashboards pull.
Your business changes faster than a change-order cycle does. Two years in, the dashboards describe the portfolio you had.
They built the model. Six months later nobody at your company can safely modify it, so small changes become engagements.
This is the one that gets underweighted. A consulting engagement delivers an artifact and ends. If a figure is off in month seven, that is your problem, and finding out why means paying someone to go read a model they did not write.
If you need five specific reports, you know exactly what they are, they will not change much, and nobody needs anything watched: buy the engagement. It is cheaper than this and it is the right size for that problem. That is a real scenario and it is not rare.
All three alternatives are ways to report. None of them monitors, and none of them drafts. That is the gap this was built for.
| Portfolio | Platform, per year | Year one, with implementation | Roughly equivalent to |
|---|---|---|---|
| 2,000 units | $96,000 | $141,000 to $171,000 | About one loaded analystNote 01 |
| 6,000 units | $288,000 | $333,000 to $363,000 | About three |
| 10,000 units | $480,000 | $525,000 to $555,000 | About five |
Implementation is $45,000 to $75,000, fixed at signature, half at signing and half at go-live. The full pricing, published
The recurring work I measured normalizes to about $82 a door a year of loaded labor. The platform is $48 a door a year, plus a one-time implementation you divide by your own door count and add to year one only. That is a real spread, and it is not the ten-to-one return you get pitched by vendors who have not measured anything.
Here is where that spread is thinnest, before you find it yourself. At 2,000 units the labor I measured is about $164,000 a year, and year one here is $141,000 to $171,000. On the labor argument alone that is a wash, or worse. If the catches below are not real for you at that size, the honest read is that you would be buying this for year two onward, and that belongs in your approval memo in those words rather than behind a payback slide. At 6,000 units the same arithmetic is about $492,000 against $333,000 to $363,000, and the gap widens from there.
More importantly: that labor number is not cash you bank. You are not going to lay off six people, and I am not going to pitch you that you should. Nobody in this business is overstaffed right now. What the labor number honestly tells you is how much work is happening, which is the size of the exposure if the person doing it leaves, and the size of the capacity you get back for leasing, residents and the judgment calls.
Any one of those is a five-figure event on a portfolio in this range. Hours are the baseline. The catches are the point.
What you give up by buying instead of building. You do not own the code, you are dependent on a vendor’s roadmap for anything genuinely novel, and you have a security review to get through that a spreadsheet does not require. Those are real. Source escrow, documented schemas and one-click export in open formats exist because those are real, not because they sound good.
Open the live demo Eleven views. Fictional portfolio. No form.Every operator asks this, you should ask it, and you should ask it of every overlay vendor who pitches you. Here is the honest version, including the three cases where waiting is the right call.
The major PMS vendors have shipped substantial native AI capability in the last year and are moving quickly.Note 03
Any vendor who tells you your PMS will never catch up is selling you something.
Their AI work is pointed at the resident lifecycle: leasing conversations, tour scheduling, renewal communications, resident service. That is the highest-volume, most universal surface they have, it sits on data they own end to end, and they will do it well. If that is where your pain is concentrated, wait for them. Buying an overlay to solve a problem your PMS is about to solve natively is a bad trade and I will say so.
Two other cases where waiting is right. If you are currently evaluating whether to change PMS at all, do not layer an overlay on a system you might replace inside a year. And if your team cannot absorb another tool this quarter, wait a quarter. Implementation with a distracted team is how software gets bought and never used, and that outcome is worse for me than a slipped deal.
A PMS builds for the median of thousands of customers. Your covenant test, your AMI table, your JV waterfall, your chart of accounts and your owner reporting are not the median, and they are where the compliance and money-side exposure lives. That work tends to arrive late in a platform roadmap because it is specific, and specific does not scale to thousands of customers.
The other thing worth saying plainly: the exposure clock does not pause for a roadmap. A rent cap exceeded next March is exceeded whether or not a feature ships in June.
NOI Engine is a read-only overlay with a full export and a real exit. It is not a system of record and it never will be. That makes waiting a cheap decision to reverse in both directions.
That asymmetry is the actual argument. Not that they will never build it. They might. The question is what it costs you to be wrong, and here it costs a week.
When you should not buy this
Five situations where I will take myself off the table on the call. Saying no in public is the only bandwidth filter a solo founder has that costs nothing.
The implementation does not pay for itself. Between 1,000 and 2,000 it only pays if the catches are real for you, which is a conversation and not a spreadsheet. Below 1,000, keep doing it by hand.
Entrata is what the connector supports. Yardi is next. If you are on something else we can talk about timing, but I am not going to sell you a connector that does not exist yet.
Some of what operators want from a platform is two weeks of PMS configuration and a conversation with their CSM. If that is your situation I will tell you on the call and you can have the twenty minutes back.
That is a different product with different requirements, and this is not it today.
Every send needs a named human. If no one has capacity to be that person, the automation suites do not work, and you would be paying the $4.00 platform rate for the analytics, which is published on its own at $1.50. Buy that instead, or buy nothing.
This list is not modesty. Every one of those is a deal I would rather lose in August than unwind in March. An implementation that does not fit is worse for me than one that never started, because the only thing I have instead of references is the record of what I said before you signed.
Two questions, one week, no vendor involved. The answers decide this, not a demo.
Have three people log a week of what they spend compiling, chasing and remembering. Multiply. Then ask your controller how many deadlines are currently tracked in a spreadsheet or in somebody’s head.
If both answers are small, none of the four options on this page are worth your time and you should not buy anything.
If they are not small, open the demo. It is a full eleven-view platform on a fictional portfolio and it will answer more in five minutes than I will in twenty.
Open the live demo See the pricing
No form. No call. There is no sales team.
Or write to michael@michaelgaff.com and we will find twenty minutes. Bring your worst reporting problem.
Everything else on this page is either something I measured or a price I publish. These three are somebody else’s, so each one is set out separately with how much weight it can carry.
About $73,000 on average, in a $52,000 to $87,000 band, higher in coastal markets, before loaded cost. That is compiled from publicly posted listings for the role, not from a survey I ran. The loaded figure of $91,000 to $102,000 is that band plus benefits and taxes, which is my arithmetic on somebody else’s number. You hire these people. If your number is materially different from mine, yours is the right one and mine is the one that is wrong.
Use your own marketBusiness intelligence platforms bill per user per month for everyone who needs to open a report. I have not quoted a rate anywhere on this page, because published rates change and a stale one inside a cost comparison is worse than none at all. Take it from the vendor’s own pricing page, multiply by every person who will need to open the report, and add it to the fee in the proposal, because the proposal will not.
No rate published hereThe claim that the major PMS vendors have shipped substantial native AI in the last year and are moving quickly. That comes from their own public announcements, which is the only place I will take it from, and nothing on this page characterizes any specific vendor’s plans. Getting a platform’s public roadmap slightly wrong on my own website is expensive and entirely avoidable.
Their announcements, not mineNo operator, unit count, market or client is identified anywhere on this page, and none ever will be. If you want the numbers that are mine, the hours derivation is published task by task.