The robot arm is largely a commodity. What determines whether a deployment works is the design around it, and that is the integrator's work. This makes the integrator choice more consequential than the brand choice, and it is usually made with far less information.
Evidence that means something
Three things worth more than any presentation.
A working installation you can visit. Ask to see a similar system in production. A capable integrator will arrange it.
When visiting, ask the site owner rather than the integrator: how long has it been running, how often does it stop, who maintains it, and what would you change if you did it again.
The people who will do the work. Ask to meet the engineer who will programme and commission, not only the person selling. Ask how many similar systems they have delivered.
A completed documentation set. Ask to see a sample handover package from a previous project, with client details removed. Documentation quality reveals more about working practices than any reference.
These three take about a week to gather and they eliminate most of the risk in the selection.
How they approach the problem
The way an integrator responds to your requirement is diagnostic.
Good signs. They ask for real sample items including the poor ones. They want to visit and measure before quoting. They ask about volumes by variant, product mix and changeover frequency. They raise questions you had not considered.
Concerning signs. A quotation after a phone call. No questions about item variation. Every question answered with an unqualified yes. Committing to a rate without simulating it.
The most revealing question. Where do projects like this usually go wrong. An experienced integrator names three or four specific things — typically part presentation, item variation, or recovery logic. One who has not encountered problems will answer vaguely or claim there are none.
Why this matters. An integrator who has failed before and can describe it is safer than one who has never had a problem, because the first has learned where the risks are.
Reading a quotation
Automation quotations are hard to compare because scope differs.
Read the exclusions before the inclusions. A markedly cheaper quotation usually excludes more, and the excluded items are still needed.
Check the commonly omitted items. Recovery logic. Training. Documentation. On-site commissioning time. Complete safety systems. Interfacing to existing machines. Content preparation for service robots.
Check the ratio of equipment to engineering. If the engineering portion looks small relative to the hardware, either it is under-costed or the work will be superficial.
Check the conditions attached to any rate commitment. Every throughput figure assumes something about input quality. Read that assumption, because it is where disputes originate.
Check the schedule. An unusually short one generally means testing has been compressed, and that time reappears as extended commissioning at your site.
What belongs in the contract
Items that prevent most later disputes.
Measurable acceptance criteria. Rate over a defined period with a defined product mix, quality by a defined method, and availability across a continuous run measured in days.
Test parts. Who supplies them, how many, which variants, and whether difficult ones are included.
Factory testing before shipment. Fixing something there costs a fraction of fixing it after installation.
A specific documentation list. Itemised, including as-built drawings.
Training detail. How many people, in which roles, for how long, and whether fault recovery is included.
Support terms. Response time, remote access arrangements, and cost after warranty.
Recommended spares with part numbers, prices and lead times.
Payment tied to milestones, with the final portion tied to site acceptance rather than to delivery.
The last is the strongest protective term, because it keeps the integrator's interest aligned with the system working in your building.
Working together through the project
Selection is half of it. Conduct during the project is the other half.
Appoint one point of contact. With authority to decide, present at technical meetings. Projects split across departments with nobody leading develop gaps at every boundary.
Send your people to the build. During factory testing. They learn more there than in any subsequent training session.
Supply real items, including the bad ones. Concealing variation to make the project look easier only defers the problem to commissioning, at your site.
Control changes with discipline. Every request states its cost and schedule impact. This does not prevent changes; it prevents unconsidered ones.
Prepare the site on time. Power, air, network, floor, lighting. This is your responsibility and it is where schedules most often slip.
Record every agreement. What is agreed verbally during commissioning will be remembered differently by everyone three months later.
Frequently asked questions
What evidence about an integrator is most reliable?
A working installation you can visit, where you ask the site owner rather than the integrator how often it stops, who maintains it and what they would change. Plus meeting the engineer who will actually do the work.
Which question is most revealing?
Where projects like this usually go wrong. An experienced integrator names three or four specific things; one who has never encountered problems answers vaguely — and the first is the safer choice.
Which part of a quotation should be read first?
The exclusions. A markedly cheaper quotation usually excludes more, and the excluded items — recovery logic, training, documentation, commissioning time — are still needed and reappear as variations.
Which contract term protects most effectively?
Tying the final payment to site acceptance rather than to delivery. It keeps the integrator's interest aligned with the system actually working in your building rather than with the equipment arriving.
More in Cost and selection and AI robots for real work.
