Four requests a day is what 300 doors produces

At 7:52 in the morning a tenant texts that her toilet is running and has been since Sunday.

That is the first of about four requests your 300 doors will produce today. Not a bad day. The average one.

We don't have a customer we can name running every tool in this article, and inventing one would be the easiest thing on this page to fake. So this Tuesday is assembled out of published numbers instead, and every figure in it links to where it came from. Where nobody has measured something, we say that too, because those gaps turn out to be most of the day.

The arithmetic that makes this a normal day rather than a bad one

Hemlane analyzed more than 193,000 maintenance requests from its own platform between August 2018 and May 2026. Single family rentals generated 3.40 requests per unit per year.

Run that against 300 doors and you get about 1,020 requests a year, which is roughly four every working day. Before anybody calls about a lease, a late payment, an owner statement or a showing.

You can also guess what they will be. In that same data plumbing was 25.4% of everything, HVAC 12.3% and appliance repair 12.2%. Half of what lands on you is one of three things, which is the first useful fact about automating any of it.

8:15am, and the toilet has been running since Sunday

The running toilet is the ordinary case. It is plumbing, so it belongs to the quarter of your volume that is plumbing. It isn't an emergency. It has already waited two days because she didn't want to bother anybody on a weekend.

What happens next is the whole argument about this technology, and it's smaller than either side says. Somebody has to read the text and decide it is a flapper rather than a supply line. Then work out whether your handyman covers that building, check whether she is home Thursday, message the vendor, message her back, and write it down.

None of that is skilled. All of it takes attention. It is the part software has actually got good at, and it is worth being precise about why: every step there is reading a sentence, matching it against something you already decided, and sending a message. Nobody has to be persuaded of anything.

10:30am, the call that takes two minutes

The second request comes by phone, because plenty still do.

Property Meld reports in its 2025 annual benchmarking report that the average maintenance request call handled by its voice product runs 132 seconds, and that the heaviest call volumes are no heat, active water leaks, clogged plumbing and lockouts. Requests coming in that way scored 4.55 out of 5 with residents.

Two minutes twelve. Hold that against what the same call costs you when it lands on a person who is already doing something else, and then gets written up later from memory.

The saving is not the two minutes. It is that the call never interrupted anybody, and the write-up happened while the tenant was still talking.

Latchel publishes a figure of 23% of requests solved at intake, meaning somebody walked the tenant through it and nothing got dispatched. Treat that as the vendor's own scoreboard: their page gives no method, no period and no sample, and the adjacent numbers on it do not agree with each other. But the shape of the claim is checkable on you. Roughly one in four or five of these ends without a van.

1:15pm, the one where a person is not optional

The third request is from a tenant whose last request you turned down.

Brian Teeter, who manages property in Little Rock, described exactly this on BiggerPockets after putting an AI maintenance system into his own business.

We had one case where a tenant whose request had been previously denied just called in work order via Vendaroo and had it re-entered and dispatched.

Brian Teeter, Turnkey Property Management, Little Rock, Arkansas

His fix was one sentence long, and it is the most useful sentence any operator has published about this: since then, we've made human approval a required step before any dispatch.

Notice what failed there. The software understood the request perfectly. It did the job it was asked to do. What it didn't have was the history and the reason. A tenant worked out that gap faster than the vendor did.

4:40pm, and the first request has not moved

Here is the part of the day that no software has fixed, and the industry's own numbers are blunt about it.

Property Meld's benchmark for speed of repair, from resident submission to invoicing, was 6.81 days across 2025. Only 39.4% of repairs finished inside seven days. Their figure got slower than 2024, not faster.

So the toilet from 7:52 this morning will most likely still be running next Tuesday. Not because nobody answered, and not because nothing was logged. Because a person has to physically arrive with a part, and the earliest that works for both of them is next week.

Every minute saved today was saved in the first hour of that request. The other six days belong to scheduling, parts and a vendor's calendar. Any pitch that implies otherwise is selling you a faster start on the same wait.

6:20pm, and it is probably not an emergency

The fourth one arrives after hours, which is the request everybody sells against.

It's worth knowing how thin the evidence under that pitch is. No credible published figure exists for how many after-hours calls a portfolio of a given size receives, or what share turn out to be real emergencies. We looked hard both times we have written about this.

What exists is operators saying so in public. Nathan Gesner, who manages almost 350 rentals, wrote on BiggerPockets that he gets fewer than four emergency calls a month, and defines an emergency as anything posing an immediate danger to a person or the property. Fire, a broken water pipe, no heat in winter. Not a clogged toilet, not lost hot water, not a noisy neighbor.

One operator on a forum is not a study. But it's one more real number than most of the people selling after-hours coverage will give you, and it suggests the problem is being answered rather than the calls being many.

What the day actually cost, and what a person was needed for

Four requests. One closed on the phone in about two minutes. Two turned into scheduling. One needed somebody who knew a history the software did not have.

The judgment calls took a few minutes each. The chasing took the rest of the day, and the chasing is what changed. That is a smaller claim than the brochures make and a more durable one, because it is the part that repeats four times tomorrow.

Count your own four

Pull last month's work orders and divide by the working days. If the number is near four per 300 doors, this Tuesday is roughly your Tuesday, and you already know which of the four cost you the afternoon.

It was not the toilet.