Offline Time Tracking in Basements: Your 1-in-100 Punch Ceiling

Every time-tracking vendor leads with accuracy: "GPS accurate to within 5 meters." That figure is worthless to a cleaner mopping a sub-basement corridor, because accuracy describes a satellite fix the phone never gets down there.

The number that actually matters below grade is not meters. It is the share of punches that reach your payroll record at all.

Offline time tracking saves a cleaner's punch locally on the phone when there is no signal below grade, then uploads it as soon as the device reconnects. The benefit is a complete, defensible record: the punch keeps its original timestamp instead of drifting to whenever the crew walked back up to daylight.

This article gives you the numbers to check yourself against, where each one comes from, and the arithmetic that tells you when a missing-punch rate has stopped being annoying and started costing real money.

Why does GPS clock-in fail in a basement?

Two separate radio links have to work for a browser-based, geofenced clock-in. The phone needs a position fix from GPS satellites, and it needs a data connection to send the punch. Basements break both, for different reasons.

The GPS civil signal is astonishingly weak by the time it reaches the ground. The published minimum guaranteed received power for the L1 C/A signal is about -158.5 dBW, which is roughly -128.5 dBm, below the ambient noise floor. Receivers only pull it out of the noise through correlation gain.

Put a slab of reinforced concrete, a layer of earth, and a steel-framed floor plate between that signal and the phone, and there is nothing left to correlate. Rebar mesh behaves like a partial Faraday cage. The ITU's work on building materials documents penetration losses in concrete that climb steeply with thickness and frequency.

Cellular data fails for the same structural reason, which is why the International Fire Code has required in-building emergency responder radio coverage in new construction, with below-grade areas specifically among the hardest to cover. If firefighters need a distributed antenna system to talk in your client's sub-basement, your cleaner's phone is not going to reach a tower from the same spot.

What happens to a punch when the phone has no signal?

There are only four possible outcomes, and knowing which one your system produces tells you most of what you need to know.

• The punch is queued locally and uploaded later: the record keeps the device timestamp from the moment the cleaner tapped the button. This is what offline time tracking means.
• The punch is stamped on arrival at the server: the cleaner taps at 10:42 PM in the sub-basement, the phone reconnects on the sidewalk at 10:58 PM, and payroll shows 10:58 PM. Sixteen minutes of paid time vanish.
• The punch fails silently: the page spins, the cleaner assumes it worked, and there is no record at all. This is the expensive one, because you do not find out until you run payroll.
• The page never loads: a browser-based clock-in needs the page in memory. If the cleaner closed the tab before descending, there is nothing to tap.

That last case is the one operators miss most often. Offline capability in a web-based tool depends on the page having been loaded while signal was available. Force-closing the tab in the elevator undoes it.

One more consequence follows from the physics: a punch recorded below grade cannot carry a real GPS coordinate, because there was no fix to record. Depending on the system it will carry nothing, or a last-known position that may be minutes and a hundred meters stale. Treat the timestamp as trustworthy and the coordinate as advisory.

Below-grade time tracking benchmarks worth measuring against

Some of these numbers are law. Some are physics. Some are operating targets you set yourself, because no trade association publishes a punch-capture benchmark for janitorial work. The table marks which is which, and you should read that column before you quote any of these to your team.

How to measure your own punch-capture rate in one pay period

You cannot manage this until you count it, and most operators have never counted it because the missing punches get quietly patched by a supervisor before payroll closes.

1. Export one full pay period of punch data to CSV. You need cleaner, location, punch type, timestamp, and whether the record was created by the cleaner or by an admin.
2. Count expected punches: shifts scheduled multiplied by 2, plus any mid-shift breaks your state requires you to record.
3. Count punches that arrived with no human intervention. Divide by expected. That is your capture rate.
4. Count admin-created and admin-edited records separately. Those are your edits per 100 punches.
5. Sort the failures by location, then by punch type. The pattern is almost never random. It clusters on out-punches at buildings where the last task of the night is below grade.

Do this for two consecutive periods before you draw conclusions. One period tells you the number. Two tell you whether it is stable or drifting.

What a missing basement punch actually costs: a 62,000 sq ft example

The following is an illustrative example with stated assumptions, not a survey result. Run it with your own numbers.

Take Harborview Medical Office: 62,000 sq ft across three floors plus a sub-basement holding the trash room, the loading dock corridor and two mechanical rooms. A crew of five cleans it five nights a week. Last task of the night is the trash haul to the sub-basement compactor, so the out-punch happens where there is no signal.

Five cleaners over five nights is 25 shifts, or 50 punches per week. Assume 6% of punches fail, which is three per week.

• Unverified paid time: when a punch is missing, the default is usually to pay the scheduled end time. Assume that adds 12 minutes of unverified time per incident. Three incidents is 36 minutes per week, about 31 hours per year.
• Wage assumption: BLS Occupational Employment and Wage Statistics reports a median hourly wage of roughly $17 for janitors and cleaners (May 2023). Load it at 1.25 for payroll taxes and workers' comp and you are near $21 per hour. Thirty-one hours is about $650 a year.
• Supervisor chase time: six minutes per incident to text the cleaner, confirm what happened, and enter the record. Three per week is roughly 16 hours a year. At $25 loaded, about $400.

Call it $1,000 a year on one mid-size account, before you count the argument you will eventually have with a property manager who claims the crew left at 10:15 and you have no punch to prove otherwise. Multiply across a portfolio of twelve buildings with below-grade work and the picture changes.

That arithmetic is also how you justify the 1-in-100 ceiling. At 1%, the same building produces about half a missing punch per week, the chase cost drops into the noise, and no single incident is old enough that the cleaner has forgotten the night.

Where should cleaners punch in a building with a sub-basement?

Offline queuing is the safety net. The procedure is the fix. In practice, the operators with clean below-grade timecards did one thing: they moved the punch point instead of trying to force a signal where there is none.

• Designate one punch point per building, above grade, where both data and GPS are reliable. Usually the dock apron, the vestibule inside the glass, or the freight lobby.
• Make it the same point for in and out. Mixed punch points produce inconsistent coordinates and confuse anyone reviewing the record later.
• Set the geofence radius from that point, not from the building centroid. If the reliable-signal spot is 60 m from the pin, a 50 m radius rejects every punch and your crew will start calling the supervisor instead.
• Sequence the route so trash and mechanical rooms are not dead last where you can. If the compactor run has to be last, the punch happens at the dock door on the way out, which is where the cleaner is standing anyway.
• Load the punch page before descending and leave the tab open. Cover this in onboarding, not in a group text at 11 PM.

What the FLSA expects when a punch uploads an hour late

Nothing in the Fair Labor Standards Act requires a particular timekeeping technology. 29 CFR 516.2 requires that you keep the hours worked each workday and the total hours each workweek, and that the records be accurate. A punch that was captured at 10:42 PM and uploaded at 11:15 PM satisfies that, as long as your record shows 10:42 PM.

A punch that was captured at 10:42 PM and recorded as 11:15 PM does not. That is a 33-minute overstatement of hours, and it is your record that is wrong, not the cleaner's memory.

Rounding is permitted under 29 CFR 785.48(b) to the nearest 5 minutes, tenth of an hour, or quarter hour, provided the practice does not consistently favor the employer over time. Rounding is not a license to swallow a systematic 15-minute upload delay, because that delay only ever runs one direction.

Also remember that walking from the sub-basement to the exit is time within the continuous workday, not commuting. Under 29 CFR 785.38, travel from one work point to another during the workday counts as hours worked. Your punch procedure should reflect the moment work actually ends.

Frequently asked questions

Does an offline punch still count as a valid time record under the FLSA?

Yes. The FLSA cares about accuracy and retention, not about how the data traveled. 29 CFR 516.2 requires accurate daily and weekly hours; 29 CFR 516.6 requires you keep the underlying time records for two years. An offline punch carrying its original device timestamp satisfies that. Keep the audit trail showing when it was captured and when it uploaded.

Can a cleaner fake an offline punch by changing the phone's clock?

If the punch inherits the device clock, yes, in principle. Two controls reduce it: require automatic network time in device settings at onboarding, and review any punch whose upload gap and timestamp do not match the known walk-out time for that building. Do not assume your system detects tampering unless the vendor states plainly that it does.

What if the phone dies in the basement before it ever reconnects?

The queued punch is gone with the battery, so you need a fallback in writing. A supervisor-entered record with the cleaner's stated end time, the supervisor's initials, and the reason is defensible if it is rare and documented. If it is happening weekly at one building, the problem is a charging habit or a route sequence, not the software.

Is a 150 m geofence radius too wide for a downtown building?

It can be. On a dense block, 150 m may reach two adjacent buildings and a parking garage. On a suburban campus it may not even reach the far dock. Size the radius to the distance between your pin and the designated punch point, add a margin for GPS scatter near tall buildings, and set it per location rather than portfolio-wide.

How many missing punches are too many?

Use cost, not feel. Multiply your missing punches per week by the unverified minutes you pay plus the supervisor minutes you spend, price both at loaded rates, and compare to the effort of fixing the punch point. Most operators find the arithmetic turns painful somewhere between two and five missing punches per week per account.

Where CleanTrack360 fits

CleanTrack360's geofenced clock-in and clock-out runs in the phone browser, with a default radius of 150 m that you can configure per location. That per-location control is the practical lever for below-grade work: set the pin and radius around the above-grade punch point your signal survey identified, rather than accepting one radius across every building. Location is captured at clock-in and clock-out only, so the timestamp is what your record hangs on. There is no native app yet, and crews work in the phone browser today.

Reports with CSV export let you run the capture-rate calculation described above instead of guessing at it. Plans are Starter at $99 a month for up to 5 team members, Pro at $199 for up to 20, and Business at $249 for up to 50, priced per plan rather than per user, with a 14-day free trial and no credit card required.