AMR and AGV ROI: how to calculate payback for your plant

Who this is for

This guide is for plant and warehouse project leads who must take an AMR or AGV investment through a budget meeting. You already know automation is on the shortlist. What you need is a payback and ROI story finance can interrogate: every major cost listed, every saving tied to payroll or avoided hiring, and assumptions written down before anyone debates the robot brand.

It is not a guide to which load mechanism to buy. For low-lift pallet trucks, autonomous forklifts, and lifting AMRs under racks or carts, use the pallet handling selection guide (/insights/amr-agv-pallet-handling-selection/) first, then bring the chosen job family back here to size the money. It is also not an NPV or IRR textbook. If your finance team requires discounted cash flow, apply their discount rate to the same cash flows this page builds. The formulas here match the simple payback and five-year net benefit the homepage estimator uses, so plant and finance stay on one model.

Typical readers sit in PCB, EV and battery, tire, appliance, electronics, solar, auto parts, food, or warehouse logistics operations. The method is the same across those plants: fix the scope of flows, price the full cost stack, and only then compare vendors.

Payback vs ROI: the two numbers finance asks for

Use two figures. Payback period in months equals total upfront investment divided by annual net savings, times 12. ROI over a chosen horizon equals cumulative net savings over that horizon minus the upfront investment, all divided by the upfront investment. Net savings means gross savings minus recurring costs such as service and software. Upfront investment should include robots, dispatch software, integration, site work, commissioning, training, and internal project time—not only vehicle list price. Count only labor hours that will actually leave payroll or avoid hiring you would otherwise need.

In symbols: payback (months) = upfront ÷ annual net savings × 12. Period ROI = (cumulative net savings − upfront) ÷ upfront. Align the horizon with your equipment policy. Five years is a common choice on industrial projects, and it is the horizon the payback estimator on our homepage (/) uses for its five-year net benefit figure.

Gross savings without subtracting subscriptions and service overstate the case. Net savings are what finance will defend. If savings ramp after go-live, either model Year 1 separately or state the average annual net savings you use in the payback denominator so the ramp is visible.

Full upfront cost (not just the robots)

A credible AMR or AGV ROI calculation includes every cost needed to put the system into production. List each line even when some turn out small for your site:

• Robots: fleet sized for peak throughput, including charging time and spare capacity—not only average hourly moves.
• Dispatch and fleet software: licenses or subscriptions for task assignment, traffic, and charging.
• Integration: MES, WMS, ERP, or PLC links, plus automatic doors, lifts, and conveyors in scope.
• Site work: Wi-Fi along routes, floor repairs, charger power, and fixtures for racks, carts, or pallets.
• Project engineering and commissioning: survey, simulation where used, mapping, testing, and go-live support.
• Training: operators, shift leads, and maintenance.
• Internal project time: your project lead, IT, and production staff during design and cutover.
• AGV guidance infrastructure (when applicable): magnetic tape, floor markers, reflectors, or similar—and a reserve for future re-routes when the layout changes. For how guidance cost differs from AMR mapping, see AMR vs AGV (/insights/amr-vs-agv/).

Recurring costs and a credible baseline

Recurring costs (OPEX) usually include service contracts, software subscriptions, spare parts, energy, and—if you lease or use robots-as-a-service—monthly fees. Subtract them from gross savings to get net savings. Leaving them out is one of the fastest ways a board challenge fails.

Build the baseline from what you spend today on the flows the robots will take over: fully loaded labor × people per shift × shifts, overtime premiums, forklift lease or maintenance on those routes, and damage or error costs tied to handling. Do not load the baseline with roles or routes outside the automation scope. Only flows the robots truly own belong in the comparison.

If you already lease forklifts or pay agency labor on the same routes, decide which of those costs disappear when robots take the moves. Keep a short note in the budget pack so auditors can trace each baseline line to a cost center or invoice class.

Savings you can defend in a budget meeting

Direct labor is usually the largest line, but only when the hours leave payroll or avoid hiring you had already planned. Overtime that disappears when peaks are covered by the fleet is often easier to prove than permanent headcount change. Turnover and hiring costs for hard-to-fill handling roles—especially nights—belong in the model when the project shrinks that hiring need. Damage to products, racks, and doors, plus wrong or late deliveries that cause rework, can be material if you have history.

Throughput and fewer line stops from waiting on material are real, but harder to measure. Keep them out of the base case or park them in a sensitivity so finance does not treat them as cash. Redeployment counts as a saving only when it avoids a hire you would otherwise make; moving someone to a role you would have filled anyway is a saving, moving them to idle time is not.

• Counts as savings: roles removed from the handling roster; avoided new hires for growth or attrition; overtime and agency hours that stop; documented damage and error reduction on automated routes.
• Does not count (or needs a separate note): people kept on the same payroll with no avoided hire; theoretical busy time that never left the shift plan; double-counting the same gain as both labor and throughput.

Board / finance assumption checklist

Print or paste this list into the budget pack. Every row should have an owner and a source (payroll extract, time study, vendor quote, or simulation). Finance will ask these questions whether or not you write them down first.

• Labor capture: which roles leave payroll or avoid hiring, and in which month after go-live?
• Fully loaded cost: which burden rates are included (benefits, premiums) and which overheads do not change?
• Ramp months: how many months until net savings reach the steady-state figure used in payback?
• Utilization source: peak moves per hour from which data—WMS logs, time study, or vendor default?
• Fleet sizing: charging, traffic wait, and spare capacity included, or assumed 100% busy?
• AGV infrastructure: guidance install and future re-route cost included when routes may change?
• Recurring OPEX: service, software, parts, energy (and lease/RaaS if any) subtracted from gross savings?
• Integration and site scope: MES/WMS/PLC, doors/lifts, Wi-Fi, floor, and charger power in the upfront total?
• Internal time: project, IT, and production hours valued or at least listed?
• Redeployment rule: saved only when an avoided hire is named; otherwise excluded from base case.
• Horizon: payback in months plus ROI over the same five-year window finance expects for equipment.
• Scenario label: which job family (finished-goods forklift, process-transfer lift, and so on) this model covers.

Worked example (illustrative only)

The figures below are illustrative assumptions chosen to show the method. They are not quotes, prices, or guarantees, and your numbers will differ.

Illustrative plant: 4 people per shift move materials over 2 shifts (8 roles). Fully loaded cost is an illustrative $40,000 per person per year. After automation, 3 of 4 roles per shift go; 1 person per shift stays for exceptions.

• Illustrative upfront (robots, software, integration, site work, training): $360,000.
• Illustrative labor saving: 6 roles × $40,000 = $240,000 per year.
• Illustrative overtime and turnover: $15,000 per year.
• Illustrative damage and error: $5,000 per year.
• Illustrative gross savings: $260,000 per year (throughput left out of the base case).
• Illustrative annual service and software: $24,000 → net savings $236,000 per year.
• Illustrative payback: $360,000 ÷ $236,000 × 12 ≈ 18 months.
• Illustrative 5-year ROI: ($236,000 × 5 − $360,000) ÷ $360,000 ≈ 228% (about $820,000 five-year net benefit).
• Sensitivity — same investment, 3 shifts: if the fleet can cover the extra hours and charging, 9 roles go instead of 6; net savings rise to about $356,000 per year and payback falls to about 12 months (still illustrative).

What moves the result most

Three inputs drive most AMR and AGV payback results. Labor cost scales savings almost linearly: cut the illustrative fully loaded cost in half and payback roughly doubles. Shifts multiply the return on the same fleet: robots that run every shift spread one CAPEX across more saved hours. Utilization decides whether CAPEX is wasted or under-sized: a fleet built for a rare peak adds investment without savings; a fleet that is too small leaves people on the roster.

Do not assume robots are busy 100% of the time. Charging, traffic, and waiting at stations consume clock time. Size to measured task data—ideally with simulation—and state the utilization source on the assumption checklist. Those three levers matter more than fine-tuning secondary damage lines once labor, shifts, and fleet size are honest.

When you stress-test the model, change one lever at a time. Show finance the payback at your current shifts and at a plausible third shift, and at a utilization drawn from your own logs rather than a brochure default. That single page of sensitivities often matters more than a longer narrative.

AMR vs AGV in the cost model

The formulas are the same; the cost stack often differs. AGV projects may carry more guidance infrastructure and higher costs each time a route changes. AMR projects typically put more cost into mapping, software, and fleet management, with lower floor-work cost when layouts change. For a fixed route that will not change for years, an AGV can still be the sound economic choice. For layouts that rebalance often, include future change cost in the total cost of ownership before you compare payback.

This page is about money, not navigation theory. For how AMRs and AGVs differ in guidance, flexibility, and traffic with people, read AMR vs AGV (/insights/amr-vs-agv/), then fold those cost differences into the upfront and lifecycle lines above.

Scenario payback by job type

Payback is not one industry average. It follows the job family you automate. Choose the mechanism first with the pallet handling selection guide (/insights/amr-agv-pallet-handling-selection/), then apply this ROI method to that flow.

Finished-goods pallet put-away and retrieval with autonomous forklifts: in typical finished-goods projects, autonomous forklifts integrated with the WMS reduce manual forklifts by about 50%, run three shifts, and pay back in about 14 months. Treat those as comparable-project values; your proposal uses your labor, shifts, and volumes. See the finished-goods case (/cases/finished-goods/) and the autonomous forklift page (/robots/autonomous-forklift/).

Process-transfer with lifting AMRs under racks or carts: in a typical process-transfer project, lifting AMRs connected to the MES cut handling labor by about 65%, ran 24 hours a day, and went live in about 9 weeks. Those are comparable-project operating results; we do not publish a single payback month figure for this scenario here—ask for a plant-specific estimate. See the lifting AMR page (/robots/lifting-amr/) and the process-transfer case (/cases/process-transfer/).

Browse the full robot line-up (/robots/) once the job family is clear, and use support (/support/) for how deployment and service enter OPEX.

Mixing finished-goods forklift economics with process-transfer lift economics in one blended payback hides the truth. Keep one model per job family, then roll them up only after each stands on its own assumptions.

Common pitfalls

Most weak business cases fail on assumptions, not on the vehicles. Watch for these:

If a supplier model looks dramatically better than your own sheet on the same flows, the gap is usually hidden scope or utilization—not magic productivity. Reconcile line by line against the board checklist before you accept the headline months.

• Counting roles that will not leave payroll or avoid a named hire.
• Using base wages instead of fully loaded cost—or loading overheads that do not change.
• Leaving out integration, site work, AGV guidance, or internal project time.
• Assuming full savings from day one with no ramp months.
• Assuming robots are busy 100% of the time (charging, traffic, station wait).
• Forgetting recurring service and software.
• Double-counting the same benefit as labor and throughput.
• Accepting a vendor calculator’s default utilization without your own peak data.
• Treating theoretical labor hours as cash (labor capture below 100% is normal).

Homepage estimator vs free assessment

A simple estimator that uses headcount per shift, number of shifts, and a regional labor band is useful for a first ballpark. It cannot see your aisle congestion, charge plan, integration scope, or which roles truly move off the handling roster. For a number you can take to a budget meeting, send a layout (CAD, PDF, or phone video) and ask for a proposal that sizes the fleet to your flows and states every major assumption. Treat any vendor model—including interactive calculators—as a starting point until those assumptions are written down and challenged.

On ApexStow, the homepage estimator (/) takes three inputs—people moving materials per shift, shifts per day, and a regional labor band—and returns ballpark payback, annual savings, and a five-year net benefit from industry averages (including overtime and turnover effects in that average model). A free assessment via contact (/contact/) uses your CAD, PDF, or phone video and, within about 48 hours, returns routes, robot count, and a payback estimate based on your plant. There is no obligation to buy.

Attach what you can: people and shifts on the target routes, average and peak moves, layout file or video, and whether tasks come from WMS, MES, PLC, or call buttons. After go-live planning, the deployment guide (/insights/amr-deployment/) covers the usual 6–10 week path so the ramp months in your ROI model stay realistic.

How to read any vendor’s ROI model

Apply the same ruler to every shortlist model—whether from Hikrobot, Quicktron, Geek+, ApexStow, or another bidder. Ask whether integration and site work are inside the upfront total; whether software and service are subtracted; where utilization comes from; how many ramp months are assumed; and whether redeployment is treated as a cut or as an avoided hire. Models that hide those rows are not comparable, no matter how attractive the headline payback looks.

You do not need to disparage any brand to run this checklist. You need one spreadsheet, the assumption list above, and the honesty to mark illustrative numbers as illustrative until your assessment replaces them.