Key takeaways
- Labour cost alone is a weak case for a robot; consistency, throughput and hazardous work are stronger reasons.
- The fit depends on volume and variation together, not on volume alone.
- A robot cell needs stable power, compressed air, maintenance and a programmer as much as it needs a robot.
- Automate one stable operation first, prove it, then extend the cell rather than starting with a full line.
The case is not labour cost alone
The usual automation proposal starts with a wage comparison. An operator costs a certain amount per shift, a robot costs a certain amount to install, and the difference is presented as a payback. That arithmetic is necessary but not sufficient. It ignores changeover, quality, downtime and the cost of the engineering that keeps the cell running.
A better case starts with the problem. Is the issue throughput, consistency, fatigue, a hazardous task or a skill shortage? Each one points to a different level of automation. A press that needs a part every four seconds is a robot problem. A batch of thirty welded frames with a new design each month is usually a fixture and a welder problem.
In India the labour case is also uneven. In a cluster with a deep skill pool, manual work can be flexible and cheap. In a plant with high turnover or a difficult-to-fill night shift, the same manual cell can be fragile. The decision belongs to the specific plant, not to a national average.
Four decision zones
Plot the work on two axes: volume and variation. High volume with low variation is the classic robot case. High volume with high variation calls for a flexible cell with quick changeover and vision or fixture correction. Low volume with low variation can suit hard automation or a dedicated machine. Low volume with high variation is usually manual, with poka-yoke and better tooling.
Most disappointing robot projects sit in the fourth zone. The volume was promised, the mix changed, and the cell spends its time waiting for a new programme or a new fixture. If the sales forecast is uncertain, the automation should be built around the process rather than the product.
- High volume, low variation: robot cell with fixed tooling
- High volume, high variation: flexible cell with quick changeover
- Low volume, low variation: dedicated machine or hard automation
- Low volume, high variation: manual cell with poka-yoke
Where robots usually pay
The strongest cases share three features: a repeated motion, a stable part position and a clear quality gain. Press tending, machine tending, welding, moulding take-out and palletising fit that pattern. The robot does not need to be intelligent. It needs to do the same thing thousands of times without a rest.
Quality is often the larger saving. A robot that places a part in the same position every cycle removes a source of variation that the die or the fixture then amplifies. In welding, the travel speed and the torch angle become repeatable, which reduces rework and post-weld dressing.
| Task | Typical fit | Reason |
|---|---|---|
| Press or machine tending | Strong | Repeated pick and place with a fixed cycle |
| MIG or spot welding | Strong | Repeatable path, quality gain, fume exposure |
| Moulding take-out and packing | Strong | Stable position and high cycle count |
| Palletising and case packing | Strong | Repetitive lift with a clear pattern |
| Deburring and blending | Conditional | Force control and part variation decide the result |
| Final assembly of flexible parts | Weak | Handling and judgement are hard to programme |
| First-off inspection | Weak | A gauge or a person is usually faster and clearer |
Illustrative task map. The fit changes with part size, cycle time and the condition of the incoming part.
Where a robot rarely pays
A robot is a poor fit where the part arrives in an unknown position, the task needs force feedback or the batch is too short to programme. Deburring a casting with variable flash, fitting a connector that needs feel and reworking a painted part are examples. A person can adapt in a second; a cell needs a sensor, a programme and a fixture.
Short batches are the other trap. A programme that takes a day to teach and a fixture that takes a week to build cannot be spread over a run of two hundred parts. The cell may still be justified for quality or safety, but the case has to say so honestly rather than hiding behind a labour saving.
Power quality is a practical limit. A plant with frequent voltage dips needs a stable supply, a controlled stop and a restart procedure before it can run an unattended cell. A robot that stops mid-cycle and needs a manual recovery can be slower than the manual line it replaced.
An illustrative payback
Assume a two-shift cell with 4,000 available hours a year. The robot cell costs £120,000 installed, including the gripper, guarding, integration and training. It saves 1.2 operators per shift at a fully loaded cost of £18,000 a year each. The annual saving is £43,200 before maintenance, power and programming. Simple payback is about 2.8 years.
Run the same cell on one shift and the saving falls to £21,600 a year. Payback moves to about 5.6 years, which may still be acceptable for a ten year programme but not for a three year one. Add £6,000 a year for maintenance, spares and programming and the one-shift case becomes harder again.
These figures are illustrative. The point is the shape of the model, not the numbers. Include the cost of the engineer who programmes the cell, the spares holding, the changeover time and the scrap rate during ramp-up. A payback that ignores those lines is a sales calculation.
Automate the operation, not the whole line
One stable cell that runs reliably is a better first project than a full line that depends on every station working on day one.
Infrastructure decides the result
A robot cell is a small factory of its own. It needs a stable power supply, clean compressed air, a floor that does not transmit vibration and enough space for the safety zone. It needs a maintenance routine, a spare parts list and someone who can recover the cell after a fault at two in the morning.
The incoming part matters as much as the robot. If the previous process delivers a part with variable flash, a burr or a bent leg, the gripper will jam. Automation moves the quality problem upstream. That is an advantage, but only if the upstream process is capable of meeting the new requirement.
Skills are the long pole. A plant can buy a robot in months, but a programmer who can teach a new part and diagnose a fault takes longer to build. Plan the training with the cell, not after it. Keep the manuals, the backups and the programme archive under change control.
- Supply quality and voltage dips over a full shift
- Compressed air quality and pressure stability
- Floor condition, layout and access for maintenance
- Programmer, electrician and mechanical fitter coverage
- Spare gripper, sensor and drive availability
- Programme backup and change control
Stage the change
The safest route is incremental. Start with a manual cell that is stable and measured. Add poka-yoke and a fixture so the part arrives in a known position. Then automate the single most repetitive step and run it in parallel with the manual line until it proves its cycle and its quality.
Only then extend the cell. Each extension should have its own acceptance test and its own payback check. A programme that adds stations without a gate tends to discover the integration problems all at once, at the worst moment.
Safety belongs at the start. Carry out a risk assessment for the cell, guard the reach, interlock the door and train everyone who works near it. Where a cobot is used, the risk assessment still decides the separation distance and the permitted force. A collaborative robot is not automatically safe in every application.
Frequently asked questions
Are robots worth it in India given lower labour costs?
Sometimes. The case is stronger where the work is repetitive, hazardous or quality critical, or where the plant struggles to staff a shift. It is weaker where volumes are low, the mix changes often and the part needs manual judgement. Build the model on the specific plant, not on a national wage average.
What is a typical payback period for a robot cell?
There is no typical figure. A two-shift cell with a stable part and a clear labour saving can repay in two to four years in many illustrative models. A one-shift cell with frequent changeover can take much longer. The result depends on utilisation, integration cost and the savings you can actually bank.
Should I start with a cobot or a traditional robot?
Start with the task and the risk assessment. A cobot suits a light task that shares space with people and needs frequent redeployment. A traditional robot suits a fast, heavy or high force task inside a guarded cell. The label on the robot does not remove the need for a risk assessment.
What should I automate first?
Choose one stable operation with a high repeat count, a clear quality gain and an incoming part that is already consistent. Prove the cell in parallel with the manual line before you extend it. Avoid starting with the most difficult assembly step.
Sources and further reading
- ISO 10218-1:2011, Robots and robotic devices, safety requirements for industrial robots
- ISO/TS 15066:2016, Collaborative robots, safety requirements
- ISO 13849-1:2023, Safety of machinery, safety-related parts of control systems
Qumek Sourcing Desk
Sourcing team
The sourcing team that runs supplier selection, cost models, audits and programme launches for buyers in Europe, the UK and North America.
Testing an automation case for an Indian plant?
Send the operation, the cycle time and the volume band. The sourcing desk will help frame the payback and the stage gate before a robot is ordered.
Discuss an automation case