It is 9:52 PM and the property manager at the surgery center is texting photos of the same trash cans she photographed last Tuesday. Your crew never showed. Nobody on your side knows yet.
You will find out at 7:30 the next morning, when someone finally opens the timesheet export. By then you have lost the night's production, you will probably credit the invoice, and you have handed a competitor a talking point at renewal.
A real-time ops map is a live view of every scheduled site on one screen, color-coded by clock-in status. Its practical benefit is speed: a supervisor can see which of 40 buildings has no crew on site within 15 minutes of shift start, instead of finding out the next morning from a client email.
That is the whole value proposition. Everything below is the sequence for building one that actually produces that alert, in the order an operator has to do it.
Stage 1: What data has to be clean before a live map is worth anything
A map is a rendering of your location records. If the records are wrong, the map lies to you confidently, which is worse than having no map.
Input: your current site list, however it exists today, whether that is a spreadsheet, a whiteboard, or your billing system. Output: one file where every active site has a verified street address, a pinned coordinate, a scheduled start time, and a named crew.
| Data element | Where it comes from | What breaks if it is wrong |
|---|---|---|
| Street address | The service agreement, not the billing address | Pin lands at corporate HQ 30 miles away, every clock-in reads as out of zone |
| Map pin coordinate | Dropped on the service entrance, not the front lobby | Crew parks at the loading dock and falls outside the fence |
| Scheduled start window | The contract's access hours | Map shows red at 6:00 PM for a site the tenant does not release until 6:30 |
| Assigned crew | Current schedule, not last quarter's | Green dot for a person who quit, no alert for the vacancy |
| Site contact and door code | Account file | Supervisor sees the problem but cannot fix it at 9 PM |
Address geocoding is the piece most operators skip. A retail plaza with six suites often geocodes to the plaza centroid, which can sit 80 meters from the door your crew uses.
Gate to Stage 2: every active site has a pin you personally trust, and no site is listed twice under two spellings. Do not configure geofences on top of dirty addresses.
Stage 2: How big should the geofence radius be for each site?
The radius is the single setting that decides whether your map is useful or ignored. Too tight and honest cleaners get locked out, so they call you, so supervisors start waving people through, so the data becomes noise. Too loose and a green dot proves nothing.
Input: the pinned coordinate plus the parking reality at each site. Output: a radius per location, documented, with a reason attached.
Start from the physics. The U.S. government's published GPS performance standard commits to a signal-in-space user range error at the meter level, but a phone in the real world adds antenna quality, building reflection, and urban canyon multipath on top of that. Practical smartphone accuracy is good in an open suburban lot and considerably worse between two glass towers.
| Site type | Starting radius | Reasoning |
|---|---|---|
| Standalone suburban office, own lot | 100 to 150 m | Open sky, small footprint, tight fence still covers the whole lot |
| Multi-tenant strip center | 150 m | Covers rear service parking without swallowing the neighboring plaza |
| Downtown high rise | 200 to 250 m | Multipath error from surrounding buildings, plus garage entry a block away |
| Industrial or distribution | 250 m | Building footprint alone can exceed 150 m across |
| Multi-building campus | One fence per building at 150 m | You want to know which building, not just which campus |
Write the rule down as a formula your team can apply to new accounts: radius equals distance from pin to the farthest point of employee parking, plus roughly 50 meters of error margin, rounded up to the nearest 50.
Gate to Stage 3: run two weeks with the radii in place and count out-of-zone rejections. If more than a handful of sites generate repeat rejections from crews who were genuinely on site, fix those radii before you build alerting on top of them.
Stage 3: How long should the grace period be before a missed start counts as an exception?
This is where most ops maps die. Someone turns on alerts with zero tolerance, the supervisor's phone lights up 14 times on the first night for people who are three minutes late, and by Thursday the alerts are muted.
Input: your actual clock-in timestamps from the last 30 days. Output: a written grace period per shift type, and one number for detection lag.
Pull the last month of start times for a stable route and look at the spread. Most crews cluster within a few minutes of scheduled start, with a long tail. Set grace so that the normal cluster is not flagged and the tail is.
- Single-site crew, drives from home: 10 minutes of grace. Traffic variance is theirs to manage, but you do not need a call over a red light.
- Route cleaner, second or third stop of the night: 15 minutes. Their start depends on how the previous building went.
- Day porter with a fixed badge-in time: 5 minutes. The client is watching this one directly.
- Access-restricted site: grace runs from the door release time, not the schedule time.
Now the arithmetic that produces the promised outcome. Detection lag equals grace period plus sweep interval. A 10-minute grace with a supervisor who glances at the board every 30 minutes gives you a worst case of 40 minutes. A 10-minute grace with a deliberate sweep at start plus 15 gives you a 15-minute alert, every time.
Gate to Stage 4: the number of exceptions your grace period generates on a normal night is small enough that a human will actually work the list. If your 6:00 PM wave produces 20 red sites, the problem is your schedule, not your map.
Stage 4: Who watches the map, and at what three times of night?
A live map with nobody assigned to it is a screensaver. The map does not create accountability. A named person working the exception list at a named time does.
Input: the exception rules from Stage 3. Output: a written sweep protocol with an owner per shift wave.
Nightly supervisor sweep protocol
- Sweep 1, start plus 15 minutes: filter the map to sites with no clock-in. Call the cleaner first, the crew lead second. Log the reason code before you hang up.
- Sweep 2, mid-shift: look for the inverse problem, sites clocked in for far longer than the budgeted hours, and sites clocked in with fewer people than scheduled.
- Sweep 3, 30 minutes after the last scheduled out time: catch open shifts nobody closed. An 11-hour open shift on a 3-hour job is a payroll correction and a coaching conversation.
- Handoff note: one short message to the morning office with every unresolved exception and the client-facing action needed.
Three sweeps, five to ten minutes each. That is the entire labor cost of the system, and it replaces the habit of calling every crew lead every night to ask whether they made it in.
Gate to Stage 5: every exception now has a reason code attached by a human. "No clock-in" is data. "No clock-in because the cleaner's daughter is in the ER" is a decision.
Stage 5: The dispatch decision: recover the night, or credit the invoice?
The map's value only converts to money at this stage. You found a hole at 6:15 PM instead of 7:30 AM. Now do the arithmetic that a supervisor can run in 90 seconds.
Input: the exception, the site's nightly billing value, and the drive time from your nearest available body. Output: dispatch, reschedule to a same-week catch-up, or notify the client and credit.
Work a named example. Brookfield Medical Plaza is 22,000 square feet, cleaned five nights a week by two cleaners at three hours each, starting at 6:00 PM. Assume, for illustration, a fully loaded labor cost of $22 per hour and a monthly contract of $5,280 across 22 service nights.
- Nightly revenue: $5,280 / 22 = $240 per night.
- Normal nightly labor: 6 hours x $22 = $132.
- Recovery option: one floater, solo, needs roughly 4 hours to cover what two people do in three, plus 25 minutes of paid drive. That is about 4.4 hours x $22 = $97, plus mileage.
- Net position if you dispatch: you spend $97 instead of the $132 you did not spend, you keep the $240 billable, and the client never learns there was a problem.
- Net position if you do nothing: $240 at risk, plus a credit conversation, plus a documented service failure in the renewal file.
Write the decision rule once and let supervisors apply it without calling you: dispatch if a qualified person can be on site before the client's access window closes and the recovery labor cost stays under the nightly billing value. Otherwise notify the client that night, not the next morning, and offer the catch-up date.
For estimating how long a solo recovery actually takes, use published task times rather than a guess. ISSA's cleaning times standards give per-task production rates you can total for a partial scope.
Gate to Stage 6: the outcome of every exception is recorded, including the ones where you chose to do nothing. Unrecorded decisions cannot be reviewed.
Stage 6: Turning last night's map into a morning exception report
The live map is a tactical instrument. It goes stale at dawn. The reason the whole build pays for itself is the record it leaves behind.
Input: clock-in and clock-out timestamps, geofence results, reason codes, and dispatch outcomes. Output: four numbers per week, per supervisor or per region.
| Metric | Formula | What it tells you |
|---|---|---|
| On-time start rate | Starts inside grace / total scheduled starts | Whether your schedule matches reality or your grace period is doing the work |
| Exception rate by site | Exceptions at a site / service nights at that site | Which single building is generating a disproportionate share of your late nights |
| Mean time to detection | Alert time minus scheduled start, averaged | Whether the sweeps are actually happening |
| Recovered vs. lost nights | Dispatched recoveries / total exceptions | The dollar value the map returned this month |
Export it, keep it, and bring the site-level exception rate into your quarterly business review with the client. A property manager who sees that you detected and recovered three missed starts before she noticed is a property manager who stops shopping the contract.
Use the same file when you build next year's labor budget. Your loaded hourly cost should reflect current wage data for your metro, which the Bureau of Labor Statistics publishes by occupation and area.
What a real-time ops map cannot tell you
Be precise with your team and your clients about what this technology is. A site-status map shows presence at a point in time, captured when someone clocks in and again when they clock out. It is not a moving dot following a cleaner through a building.
Continuous location tracking between punches is a different product, a different cost, and a different conversation with your workforce. Most commercial cleaning operations do not need it, because the question they actually need answered is binary: is anyone at this building right now?
Before you roll out any location capture, check your state law. Several states, including Connecticut and New York, require employers to give written notice of electronic monitoring, and California has specific rules around location tracking devices. Notice costs you nothing and prevents an expensive surprise.
Frequently asked questions
What geofence radius should I use if my cleaner parks in a public garage two blocks away?
Do not stretch one fence to cover the garage. Two blocks in a downtown core can be 250 meters or more, and a fence that large will accept a clock-in from a coffee shop. Instead, pin the site at the building entrance, keep the radius at 200 to 250 meters for urban canyon error, and instruct the cleaner to punch in at the lobby door, not in the car.
Can a cleaner clock in from home if the geofence is set too wide?
Yes, and that is the real cost of a lazy radius. A 500-meter fence in a dense residential area can easily cover apartments across the street. Keep the fence just large enough to cover employee parking plus error margin, and review any site where the same person is repeatedly punching in at the outer edge of the zone.
How do I handle sites where the building has no cell signal or Wi-Fi?
Basements, mechanical rooms, and older concrete structures kill both signal and GPS lock. Set the expectation that the punch happens at the entrance before the crew descends, and flag those specific locations in your site notes so supervisors do not chase a false exception. If a site fails repeatedly, switch it to a supervisor-confirmed start rather than pretending the data is reliable.
Is a 15-minute alert realistic if all my sites start at 6:00 PM?
Only if the exception list is short enough to work. If 40 buildings start at once, one supervisor cannot call 12 no-shows in 15 minutes. Either stagger start times across a 6:00 to 7:00 PM window, which most commercial clients will accept, or assign sweeps by region so two or three people each work a list of five or fewer.
What reason codes should supervisors use when logging an exception?
Keep the list under eight so people actually use it. A workable set: called out sick, no call no show, vehicle issue, site access denied, schedule error on our side, client cancelled, phone or app issue, and started late but on site. Those codes are what turn a pile of red dots into a coaching plan or a schedule fix.
Where CleanTrack360 fits
If you want to build the sequence above without stitching together a spreadsheet, a mapping tool, and a group text, CleanTrack360 covers the pieces in one place. Geofenced GPS clock-in and clock-out runs in the phone browser, with a default radius of 150 meters that you can configure per location, so Stage 2 is a setting rather than a project. Scheduling is drag-and-drop with recurring shifts and work orders, bulk CSV import gets your location list in, and reports with CSV export give you the Stage 6 numbers. Note that location is captured at clock-in and clock-out only, not continuously between punches, and crews work in the phone browser today because the native app is still in development.
Plans start at $99 per month for Starter, up to 5 team members, then $199 for Pro, up to 20, and $249 for Business, up to 50. Pricing is per plan rather than per user. There is a 14-day free trial and no credit card required, which is enough time to run the two-week radius calibration described in Stage 2 on your real routes.