Robot Site Readiness Guide: Factory Rollout

Most robot delays start before the robot shows up. If you want a factory rollout to stay on schedule, I’d treat site readiness as a hard pass/fail gate before install, not a loose checklist.

In plain terms, I’d make sure five things are locked down first:

• Layout and flow: enough space, clear access, safe crossings, and defined material movement
• Machine interfaces: full signal lists, handshake timing, and clear ownership for controls and safety
• Utilities and OT: power, air, grounding, temperature, humidity, wired/wireless coverage, VLANs, firewall rules, and remote access rules
• Safety and task fit: risk review, stopping-distance checks, task stability, low manual intervention, and clear escalation paths
• Commissioning and release: FAT, SAT, endurance runs, training, spare parts, and hard go-live targets like 95%–98% uptime, 95%–105% throughput, and 98%–99% first-pass yield

A few numbers matter right away. I’d look for at least 36 inches of clear access around service areas, network latency below 100 ms for AMR routes, cycle-time spread within about 10%–15% for repeat work, and supervised validation runs of 24–72 hours before launch.

If even one of those items is vague, unsigned, or untested, the site is not ready. That’s the whole point: prove readiness with logs, drawings, photos, and sign-off before hardware arrives.

Review Layout, Material Flow, and Machine Interfaces

Walk the line with printed floor plans and a tape measure. A layout can look fine on paper and still hide tight corners, shared aisles, or missing buffer space once you're standing on the floor.

Check Workcell Space, Access Paths, and Congestion Points

Tape out the robot's reach, normal operating zone, and E-stop boundary on the floor. Then compare the surrounding area with the minimum clearance rule: require at least 36 inches of clear access on every serviceable side, more where hoists or lift trucks are used.

Use example cell sizes as a rough gut check, not a design rule:

• Single-station cells are often about 10 ft × 13 ft
• Double-station shuttle cells are often about 13 ft × 20 ft

The bigger problem is usually congestion. Watch at least one full shift before you lock in routes. Look for operator cut-throughs, pallet buildup, and AMR slowdowns. Build routes and barriers around observed traffic, not planned traffic.

Mark every crossing between pedestrian walkways, forklift lanes, and AMR paths on the updated floor plan. Then give each crossing a clear fix, such as rerouted traffic, painted visual lanes, physical guardrails, or revised SOPs.

Next, map how parts and people move through the cell.

Map Part Flow and Manual Touchpoints

Trace the material path from start to finish. Document every entry, queue, transfer, and exit point, including conveyors, chutes, pallets, totes, and bins.

For each entry point, define how the robot will know a part is there. That could be a sensor, vision, or operator confirmation. Then size buffer zones around actual demand swings, not just planned throughput. Otherwise, the robot can end up starved for parts or blocked by normal process variation.

Manual touchpoints need the same level of detail. List every place where an operator handles material inside the robot's process scope, including loading fixtures, visual inspection, rework, labeling, and exception handling. Each touchpoint should have:

• A defined physical location outside the robot's active motion area
• Clear ergonomic access
• Its own tools or reject bins

If these touchpoints aren't defined up front, people tend to create their own shortcuts after go-live. That's when unsafe crossings show up.

Document Machine Handshakes and Control Ownership

For each machine tied to the robot, build a signal list that covers six groups: cycle control, part presence and position signals, safety interlocks, fault and alarm codes, mode and recipe management, and reset and recovery logic.

That means documenting signals such as machine "cycle start", robot "ready to pick/place", and "cycle complete", along with guard door status, light curtains, and e-stops.

For every signal, record:

• Source
• Destination
• Data type
• Normal state
• Timing expectations, including handshake timeouts

Then assign ownership with an interface responsibility matrix. Use the matrix to set design, implementation, and test ownership for each signal or safety function.

Signal / Function	System Origin	Owner	Test Owner
Cycle start / cycle complete	Machine PLC	Controls Engineering	Controls Engineering
Part-present sensor	Machine PLC	OEM / Plant Engineering	Integrator
Guard door interlock	Safety PLC	Plant Safety	Safety + Controls
Robot fault / E-stop chain	Robot Controller	Robot Integrator	Integrator + Plant
Recipe / product selection	MES / SCADA	IT/OT	IT/OT + Controls
Fault reset and recovery logic	Robot + Machine PLC	Controls Engineering	Controls + Integrator

Without this matrix, fault resets and recipe changes tend to bog down commissioning. With it, each team knows which signals it owns before testing begins.

Once the layout and interfaces are mapped, confirm the site can power and connect the system.

Verify Utilities, Network, and OT Access

Next, make sure the site can support robot operations in the real world, at full production scale. That means checking both the physical setup and the digital setup. Power quality, compressed air, grounding, and network connectivity should be checked against the actual robot specs - not rough guesses.

Confirm Power, Air, Grounding, and Environmental Limits

Start with a load schedule for every device in the cell. Add up the worst-case simultaneous load, then compare that total with panel and transformer capacity using a 20%–30% margin. Put critical loads on dedicated circuits, and keep robot control circuits separate from noisy loads like welders or large motors.

Add surge protection, UPS backup for critical controllers, and power monitoring so you can deal with sags, transients, and harmonics before they turn into shutdowns.

Air matters too. Check point-of-use air pressure during a representative production cycle, not just at idle. Confirm air quality and grounding against the robot and tool specs, and bond all cabinets and bases to the plant ground grid.

Then watch the space itself. Log temperature and humidity for several days during normal production. If heat or moisture falls outside spec, fix it before installation. Record each item as pass or fail, along with measured values and photos or test logs.

Once utilities pass, move to the network and test coverage, latency, and segmentation along the robot’s actual route.

Test Network Coverage, Latency, and Segmentation

For fixed robots and workcells, wired Ethernet is the baseline. It’s deterministic and easier to secure. For mobile fleets like AMRs, verify that end-to-end latency stays below 100 ms across every route.

Run a full site survey before locking in access point placement. Measure received signal strength, throughput, latency, and jitter at multiple spots along robot paths, including corners, docking stations, and congested aisles. Simulate live telemetry and video traffic while robots are moving. That step helps you spot dead zones and trouble sources like metal racks, overhead cranes, and large machines before they cause field issues.

For segmentation, use VLANs with firewall rules and ACLs. A simple three-VLAN setup works well for many robot deployments.

VLAN	Traffic Type	Priority
Control & Safety	Robot controllers, PLCs, safety I/O	Highest - QoS enforced
Telemetry & Logging	Historian, MES, monitoring dashboards	Medium
Remote Diagnostics	Engineering workstations, VPN sessions	Controlled, logged

All inter-VLAN traffic should terminate at a firewall, not move freely between switches. Use whitelisting so only specific protocols and ports can pass between zones. RBAC should limit access by role, so operators, technicians, integrators, and remote support providers only get the permissions they need. Route remote access through VPN with multi-factor authentication.

Lock these settings before safety validation and commissioning.

Choose a Plant Network Design for Robot Operations

Pick the network design based on layout stability and mobility needs. Use wired Ethernet for fixed cells, and use wireless only where mobility makes it necessary. If the factory layout changes often, Wi-Fi 6/6E or private 5G can give you more room to work with over time. If the layout is fixed and electromagnetic conditions are rough, a wired-only design is simpler and easier to lock down.

Document the final architecture now so commissioning begins with fixed OT paths, not open design questions. Record the final network design, AP placement, VLANs, and firewall rules before commissioning.

Validate Safety, Task Fit, and Human Escalation

With utilities and network in place, the next step is simpler to say than to do: is this task safe to automate at this site, and can the robot run it with little day-to-day hand-holding? You need clear answers before commissioning starts.

Run a Standards-Based Safety Review

Start with a site-specific risk assessment for the exact workcell, the exact task, and every person who may interact with it.

Define the task, set the workcell boundaries, and list all human roles around the system: operators, maintenance technicians, cleaners, and supervisors. Then map hazards across normal production, recovery, maintenance, and cleaning. That includes crushing, shearing, entanglement, unexpected startups, collision, and loss of control. Use ISO 10218 for industrial robot applications and ISO/TS 15066 for collaborative or shared-space setups.

For each hazard, score the severity, how often people are exposed, and how likely it is that someone could avoid harm. After that, apply protective measures in the right order: inherently safe design first, physical safeguards next, then warnings and training. In practice, most cells rely on layered controls such as guarded access, presence detection, safety-rated stops, and speed or zone limits. A common setup includes fenced guarding with interlocked gates, light curtains or laser scanners, and safety-rated controllers that enforce speed and zone limits.

Stopping-distance validation has to happen on site. It is not just a spreadsheet exercise. Measure the robot’s actual stopping time at worst-case speed and load, then verify the safety distance with the formula S = K × T + C, where S is safety distance in inches, K is approach speed, T is total reaction time, and C accounts for intrusion depth. Record the stop time, speed, load, and engineer sign-off in the validation log.

The final safety file should include the hazard list, risk ratings before and after mitigation, protective device specs, stopping-distance test results, residual risk notes, and sign-offs from safety, engineering, and operations. Make sure it lines up with OSHA expectations and your internal EHS audit process before commissioning begins. Once the cell is safe, the next check is whether the process itself is steady enough to automate.

Check Whether the Task Is Stable Enough for Automation

A robot does best when the process is repeatable, measurable, and light on exceptions. If the line keeps throwing curveballs, automation tends to bog down.

Collect at least several days of live production data before making the call. Log cycle times, scrap rates, rework, and manual interventions across multiple shifts. A practical threshold is this: if cycle-time spread stays within 10%–15% across a series of identical parts, and manual stops stay low, the task is a fair candidate for automation. Record at least 20 to 30 consecutive parts and confirm that the sequence repeats the same way before approving robotization.

Then look at part variability. Measure dimensions, weight, surface finish, and orientation variation. Watch how operators pick parts now. Do they grab them cleanly, or do they keep regripping and adjusting? Also count exceptions such as jams, misfeeds, defective parts, and label issues. If those problems pop up every few minutes, the cell will spend more time recovering than producing.

Readiness Factor	Automation-Friendly	Poor Fit for Early Automation
Process stability	Repeatable steps, narrow cycle variation	Frequent changes or unclear step sequence
Part presentation	Consistent orientation and feed	Variable orientation, mixed containers
Exception rate	Rare, well-defined, recoverable	Frequent edge cases, unpredictable recovery
Safety	Clear safeguarding, validated stopping	Unresolved hazards, incomplete validation

If the task fails these checks, document the no-go call and set a re-evaluation date, usually six months out. Tie that review to specific fixes, such as part presentation standardization or fixture redesign. If it passes, the next job is to define exactly what happens when autonomy stalls.

Define Intervention and Teleoperation Paths Before Launch

At go-live, every stop needs an owner and a response time. Parts move. Sensors drift. Odd objects show up. That’s normal. What matters is whether your team knows who responds, how fast, and under what rules.

Build a responsibility matrix that ties each event type to a named role and target response time. Operators should handle simple recoverable stops. Cell technicians should own repeated faults. Maintenance should take hardware issues. Engineering or vendor support should take software and integration problems. Set response targets based on production criticality. For example, operator acknowledgment within 2 minutes and technician arrival within 15 minutes during staffed shifts. Document those expectations as site-level SLAs.

It helps to think of escalation as a four-level flow:

• Level 1: Local operator intervention
• Level 2: On-site automation technician
• Level 3: Remote support or teleoperation
• Level 4: OEM or integrator escalation

Build these paths into HMI screens, andon boards, or your MES so alarms route to the right people without delay. Any intervention that involves physical access must follow OSHA lockout/tagout rules.

For Level 3 remote support, set up a teleoperation path for vetted remote operators. Use it when the robot stalls, runs into unfamiliar objects, or triggers a safety alert. During rollout, the robot may run on its own for standard cases and call for remote help on edge cases. For U.S. operations, remote operator authority limits and on-site EHS rules need to be defined before any remote-controlled action is allowed.

Commissioning the System and Set Go-Live Criteria

Once safety steps and escalation paths are in place, commissioning answers a simple question: can this cell run at production speed on the plant floor? That’s why FAT, SAT, and on-site validation should be treated as formal gates, not last-minute bring-up.

Run FAT, SAT, and On-Site Validation in Order

FAT happens at the supplier's or integrator's facility before the system ships. The goal is to check function, safety, and documentation before anything leaves the building. Nothing should ship until FAT is signed off and backed by a formal issue log.

SAT checks the system in the plant where it will actually run. This is where the factory floor has to prove the setup works with real grounding and cabling, actual utilities, links to upstream and downstream equipment, production parts, cycle-time checks against takt time, and fault recovery tests. Those tests should cover power loss, communication dropout, mis-picked parts, and blocked conveyors.

On-site validation pushes testing into day-to-day operating conditions. Operators run the cell under supervision during supervised operator runs. Endurance tests should target 24–72 hours of cumulative operation at planned takt time. Cross-shift trials make sure every shift - day, swing, and night - can start up, run, and hand off the cell the right way.

Compare FAT and SAT Responsibilities

Use FAT to prove the build. Use SAT to prove the plant integration.

Attribute	FAT	SAT
Location	Supplier or integrator facility	Customer's plant floor
Primary objective	Verify design spec and function before shipment	Confirm integration and performance in the real environment
Stakeholders	Integrator, robot OEM, customer engineering	Plant operations, maintenance, safety, quality, and sometimes workforce reps
Test depth	Functional and interface testing, simulated I/O	Live utilities, real materials, upstream/downstream integration
Pass criteria	Meets specification, no critical faults in test routines	Meets cycle time, uptime, and quality targets
Common failure modes	Incomplete I/O mapping, software bugs, missing docs	Layout fit issues, firewall or segmentation conflicts, safety-device nuisance trips

Set Go-Live Thresholds and Handoff Requirements

After validation, turn test results into hard release criteria. Go-live thresholds need to be quantitative and agreed on before SAT begins. If that sounds strict, good - it should be. A vague launch target is how teams end up arguing on the floor at 2:00 a.m.

Required release items include:

• Throughput at 95–105% of planned units per hour
• First-pass yield above 98–99% for high-volume lines
• Sustained uptime of 95–98% or higher during validation runs
• All safety validation items formally closed, with no high-severity findings left open
• Documented training completed for every operator and maintenance technician on every shift
• Critical spare parts physically on-site: end-of-arm tooling wear parts, safety relays, and key controller modules
• Named escalation contacts with defined response-time expectations
• Remote-support path tested and documented before go-live

The final go-live review should present performance and safety data from the validation runs, record open risks and mitigation owners, and formally transfer ownership from the project team to plant operations. Only then should ownership move to operations.

FAQs

What can delay a robot rollout most?

The biggest delays usually come from factory-readiness gaps, not the robot itself.

Before deployment, teams need to check the basics:

• layout
• machine interfaces
• network capacity
• safety
• task fit

Delays also show up when escalation paths, commissioning steps, and go-live criteria aren't clearly defined. Evlo.ai helps by connecting systems, enabling human-in-the-loop intervention, and helping teams learn from exceptions so future performance gets better.

How do I know if a task is a good fit for automation?

A task makes sense when the site is ready and the scope is clear. Before moving ahead, teams should check the layout, machine interfaces, network and edge readiness, safety boundaries, and whether the task can be measured in a pilot.

They should also set clear escalation paths, test all the way through commissioning with logged interventions, and go live only after a few things are in place: safety and commissioning gates, stable uptime and MTTR, and alert and recovery loops that work as expected.

What should be signed off before go-live?

Before go-live, teams need sign-off that the factory is ready in the places that matter most: layout and workcell fit, machine interfaces and data flow, network readiness, safety and governance, task fit, escalation paths, and commissioning.

The go-live decision should be based on measured performance, not gut feel. That means checking reliability, MTTR, OEE, and first-pass yield. It also means making sure the full alert-to-work-order-to-feedback loop is working end to end, with human oversight and incident handling in place.

Related Blog Posts

• Why Smarter Robots Alone Won't Fix Your Factory
• Meet the Intelligence Layer That Never Sleeps on Your Factory Floor
• Why Every Deployed Robot Is Really an Infrastructure Problem
• Humanoid Robot Safety Checklist for Factory Floors