After-hours maintenance calls: what they cost and how to fix the process

Why do after-hours maintenance calls go wrong?

Check which of these four failures you have before you change vendors or buy software.

• Fuzzy tiers. If the person answering cannot tell a dripping faucet from an active leak, every call becomes a page, and technicians learn to ignore pages.
• A list instead of a policy. "Try Mike, then Carlos" leaves the real question open: how long do you wait for Mike?
• Messages instead of records. A voicemail or an email has no owner, no status and no timestamp the day team can trust.
• Nobody measures. Without time to acknowledge and time on site for each emergency, you cannot tell a bad night from a bad process.

Sort every call into four tiers

Write four tiers per property and train everyone who answers on them: staff, the answering service or the software. The matrix below shows the shape. The questions in the middle column matter as much as the tier names, because they are how a tired caller's description becomes a decision.

Climate and code move the lines. An AC failure in July in Phoenix belongs in the emergency row, and some cities set minimum heat requirements in their housing codes; your lease and local code decide. Other common emergency entries: a gas smell (life safety), sewage in the unit, no heat in cold weather, a broken exterior lock and an elevator entrapment. A refrigerator that stops cooling or a garage gate stuck open usually waits for morning.

Write the escalation policy on one page

A call list becomes a policy when every step names a person, a channel and the window before the page moves on. Adapt the sample below and keep the current version in one shared place. If the rotation lives in a group text, someone will page the technician who swapped weekends.

Set acknowledgement and response targets

Measure acknowledgement and arrival separately. A slow acknowledgement is a paging problem; a slow arrival is a staffing, distance or vendor problem.

Treat these as starting points to adapt. Extend the on-site target for scattered homes and long drives. Track the median and the worst case every month. The worst case is the one an owner, insurer or board will ask about.

Capture the same fields on every ticket

Whoever answers, the ticket should look the same in the morning, and it should answer the question an owner or insurer asks later: what happened at 11:47 PM on a Sunday.

Every field on the example above belongs on the minimum set, whoever answers. Two are easy to skip at night: whether the problem has happened before (repeat calls within 72 hours are a flag), and what the resident was told, and when. Photos sent by a text link beat a verbal description. In privatized military housing, federal law adds its own requirements, such as telling the family the repair time frame when the request goes in; the MHPI Tenant Bill of Rights work-order guide lists the fields that follow from them.

Which calls should always reach a person?

Some calls should never end with an answering service or an AI system alone, however good the script. Either can take the details, give the 911 instruction and page someone. A person makes the decision.

• Injury, a death, a threat or a crime in progress. After the 911 instruction, the manager is notified and a person follows up.
• A distressed caller. Panic, grief or anger gets a person on the phone as fast as your staffing allows.
• Shutting down building systems. Closing a riser, impairing a fire system or taking an elevator out of service is a staff decision.
• Entry and identity disputes. "Let me into my ex's apartment" is never an intake question. Routine lockouts can follow a written policy; the student housing move-in week playbook has a sample.
• Money and policy exceptions. Fee waivers, emergency relocation, hotel costs and anything with legal weight.
• Anything the policy does not cover. The safe default is a page with a note.

Print this setup checklist

You do not need new software to start. You need about a month of attention and the people who answer the phone in the room for the second week.

A worked cost model

After-hours calls cost money in three places: intake (answering-service minutes or a staff member's night), dispatch (overtime, callout minimums and on-call pay) and misses (the real emergency that waited until morning).

On intake, published per-minute answering-service plans we reviewed range from about $1.40 to $5.00 a minute, depending on vendor and plan size (PATLive, Ruby). On dispatch, the Bureau of Labor Statistics puts the median pay for general maintenance and repair workers at $23.84 an hour (2025) and notes they "may need to be on call outside normal business hours, such as during evenings and weekends, for emergency repairs" (BLS). Under the federal FLSA, an employee on call at home is "not working (in most cases)," but "additional constraints on the employee's freedom could require this time to be compensated" (DOL Fact Sheet #22; see also 29 CFR 785.17). Talk to employment counsel before you redesign on-call pay. Managers carry the next morning: BLS lists their 2025 median at $33.65 an hour and notes they "may need to respond to emergencies during off-duty hours" (BLS).

Here is a model for a hypothetical 1,000-unit apartment portfolio. Every input is an assumption; replace each one with your own phone logs and payroll data.

What tighter triage changes. Suppose written tiers cut unnecessary callouts from eight to two, and complete tickets mean only 20 calls need a morning follow-up. Six fewer callouts come to $519.12 and 40 fewer follow-ups to $112.17: about $631 a month recovered, or $7,575 a year.

The line you cannot model cleanly matters most. Across 2019 to 2023, water damage and freezing claims on homeowners policies averaged $15,400 paid per claim, according to ISO data published by the Insurance Information Institute (Triple-I). That is homeowners data rather than apartment data, but it shows the scale. One leak caught an hour earlier can be worth more than the whole table, which is why the policy matters more than the rate you pay per minute. To compare answering options on that basis, see AI concierge vs. answering service. To run your own call volume and wages, use the ROI calculator.

Where TowerLine AI fits in the night process

TowerLine AI does not write your tiers or set on-call pay. Once the tiers are written, the AI concierge in TowerLine AI is set up per property to ask the triage questions and use the contacts above. On a maintenance call it is designed to give the 911 instruction when it applies, open the work order during the call with priority, location, notes and permission to enter, text a link for photos, and page the on-call person your policy names. Who gets paged, and in what order, comes from the escalation policy we set up with you; what happens when a page goes unanswered is something we scope with you during implementation. The transcript is attached to the ticket, and the call log in the Command Center can show the page and whether it was acknowledged. See how the AI concierge handles a maintenance call, or how this looks for apartment and multifamily teams.

Frequently asked questions