Multi-site attendance controls for construction and field operations

Single-site attendance is a solved problem. One office, one geofence, one team, one shift pattern. Pick any modern workforce tool and you can be up and running in an afternoon.

Multi-site is where the assumptions break. A construction crew works on three sites in a week. A maintenance technician covers ten retail branches across a region. A nurse rotates between two hospitals in a single shift. The simple model — one employee, one location, one expected schedule — does not survive contact with how field work actually happens.

This is a guide to the decisions that matter when you are designing multi-site attendance for crews that move.

The four questions to answer up front

Before you turn anything on, the design has to answer four questions clearly.

1. Who decides where someone is supposed to be today?

In a single-site organisation, the answer is implicit: they are supposed to be at the office. In a multi-site setup, the answer is a deliberate choice. The options are:

• Rostered. A scheduler assigns the worker to a site in advance. Check-ins anywhere else are exceptions.
• Pool-based. The worker can check in at any site within a defined pool (e.g. "any retail branch in the North region"). Anything outside the pool is an exception.
• Self-declared. The worker selects from a list at check-in time. The system records the declaration and validates against the geofence they are actually inside.

Different teams need different models. Construction tends toward rostered. Maintenance tends toward pool-based. Inspection roles tend toward self-declared. Pick deliberately. Mixing models inside one team makes the exception report incoherent.

2. What counts as "on site"?

Construction sites are large. Hospitals are bigger. A 50-metre geofence around the office reception will reject a foreman who spends the day in the far corner of a building plot. The geofence has to match the operational reality.

For large sites, the right default is multiple overlapping zones rather than one big circle. A construction site might have:

• A "site arrival" zone at the gate, used for check-in.
• A "works area" zone covering the active build, used for safety-critical roll calls.
• A "compound" zone covering offices and welfare, used for break tracking.

Each zone has its own purpose. Each one produces different evidence. The worker does not have to think about which is which — the system selects based on the action.

3. What happens when there is no signal?

Field work happens in basements, in lifts, behind concrete, in tunnels, in rural areas with poor coverage. An attendance system that fails closed in these environments will be hated. One that fails open without controls will be exploited.

The right pattern is deferred capture with integrity binding:

• Allow the worker to perform the action (check-in, check-out) without network.
• Bind the action to the device's signed local time, GPS reading at that moment, and any integrity signals captured.
• Queue the event for sync. When the device next has network, the queued event is uploaded with all of its original metadata intact.
• Flag any event with a sync delay above a threshold (say, two hours) for review.

This preserves the experience for the worker, preserves the evidentiary value of the record, and gives supervisors a clear filter for anomalies.

4. Who reviews exceptions when the supervisor is not on site?

In a single-site setup, the supervisor and the workers share a location. In multi-site, the supervisor might be at headquarters while the team is spread across four states.

The review path has to be explicit:

• Tier 1. Local lead at the site (foreman, charge nurse, store manager) reviews same-day exceptions for their site.
• Tier 2. Regional supervisor reviews exceptions that local leads escalate, plus cross-site patterns (e.g. one employee with exceptions at multiple sites).
• Tier 3. Compliance or HR reviews exceptions that involve discipline, payroll dispute, or audit response.

Build the queue with explicit routing. Without it, exceptions either drown the wrong person or fall through.

Team-by-site assignments

For most field organisations, the unit of policy is not the employee — it is the team-by-site assignment. The same worker can be:

• On Site A on Monday, where the shift pattern is 07:00–15:00 and meal breaks are auto-deducted.
• On Site B on Tuesday, where the shift pattern is 06:00–14:00 and breaks must be clocked.
• On Site C on Wednesday, where they are supervising rather than working, with a different pay rate.

The attendance system has to know, at the moment of check-in, which assignment is in force. That comes from the roster. If the roster is missing or wrong, the system cannot tell which policy to apply.

The design implication: rostering and attendance are the same product, not two separate ones. Teams that buy them separately end up with reconciliation problems by week two.

What to capture at each check-in

For multi-site reliability, every check-in event should carry:

• Employee identifier
• Assignment identifier (team + site + role + shift)
• Geofence ID matched at the time of action
• Device fingerprint and integrity signals
• GPS coordinates with accuracy radius
• Local device time, server time, and the delta
• Capture mode (online / offline-deferred) and queue delay if applicable
• Selfie or photo evidence if the policy requires it

This is more data than a single-site system needs. It is exactly what a multi-site programme requires to stay defensible.

Three patterns that catch problems early

Once the foundation is in place, three reporting patterns surface most of the operationally interesting signal.

Cross-site presence anomalies

An employee with check-ins at two sites more than 30 minutes apart on the same day, separated by less than the realistic travel time, is either rostered wrong or the data is wrong. Either way, you want to see it.

Site-level integrity drift

If one site's exception rate climbs steadily over two weeks while the others stay flat, the cause is local — a supervisor who stopped reviewing, a device issue in the area, a policy update that did not propagate, or a team that has discovered a workaround. Surface the trend at the site level, not just the employee level.

Roster vs. reality

A weekly comparison of "rostered attendance" against "actual check-ins" tells you which sites are running over- or under-staffed against plan. This is where attendance data starts paying for itself outside of compliance.

What good multi-site looks like

A mature multi-site attendance programme has five properties:

1. Every employee, on every day, has a known assignment that the system uses to validate check-ins.
2. Sites have geofence configurations that match their actual operational footprint, not just a circle around a postcode.
3. Offline capture is the default behaviour, with sync-delay flags driving the exception review.
4. Exception routing is tiered by site, region, and severity.
5. Cross-site reports surface patterns that single-site reports cannot.

When a team can describe all five, multi-site stops being a source of friction. It becomes the part of the workforce programme that actually generates leverage — because field work is, by definition, where the value gets produced and where the data has historically been weakest. Closing that gap is the highest-yield piece of work in most operations functions.