How to assess a robot trial before you buy

A useful robot trial should end with a decision you can explain. What work did the robot complete? How much help did it need? What would the full arrangement cost? What remains unproven?

Set those questions before the demonstration. Otherwise, an impressive moment can become the main evidence for a purchase that has to work every day.

Start with one specific job

Describe the task in ordinary language. For example: move a defined container between two agreed stations during a stated operating window. Include the load, route, frequency, people involved and what counts as a correctly completed job.

Record the current process too. How long does it take? Where are the delays? What does staff involvement actually involve? Compare the robot with a sensible improvement to the existing process as well as with doing nothing.

For education or research, name the outcome you want to enable. That may be a particular lesson, experiment or software capability. Labour savings are not a useful substitute for those objectives.

Agree success before the trial begins

Write down a short set of measures and acceptable results with the supplier and customer. Leave no uncertainty about who records them or which trial period counts. Useful measures include:

• Correctly completed tasks divided by attempted tasks.
• Completed work over the whole elapsed trial period.
• Number of human interventions and the staff time they require.
• Time spent charging, waiting, recovering or unavailable.
• Missed work, damaged items and incidents requiring review.
• Setup, training and support needed to repeat the result.

Specify whether the robot is autonomous, remotely operated or receiving another form of assistance. Separate those periods in the record. Agree technical controls and stopping conditions with the people responsible for the site and equipment before testing.

Look for evidence beyond a successful clip

Figure's account of its BMW deployment is useful because it discusses cycle time, placement accuracy and interventions, alongside work completed and hardware lessons. The article includes programme targets as well as reported results. Those are different kinds of evidence and should stay separate. It is a manufacturer's account of its own deployment, not proof that another robot will reproduce the outcome at your site. Read Figure's BMW deployment report.

Apply that distinction to your own trial. Keep a dated record of what was observed, what the supplier said and what still depends on future development. Ask for the exact hardware, software and accessories used. A later configuration change may mean a result needs checking again.

Count all attempts and all help

Here is a fictional illustration. A robot correctly completes 190 of 200 attempted journeys. That is a 95% task success rate. Staff make ten interventions totalling 50 minutes. The success rate alone does not tell you whether the arrangement meets your needs; the intervention time and the consequences of the ten unsuccessful attempts also matter.

Make sure the denominator is clear. “95% success” might otherwise refer to a selected sequence, one movement within a longer task or a period that excluded difficult runs. Preserve the underlying log so the result can be understood later.

Repeat the agreed task under representative, authorised conditions. One good run is encouraging. It does not establish repeatability across another shift, load or site.

Turn the result into a costed choice

Add the costs of equipment, setup, software, training, support and staff involvement. Include the work required to keep the robot available. Keep one-off costs separate from recurring costs and use a stated comparison period.

Time released is valuable, but it is not automatically cash saved. Explain whether it avoids an actual cost, adds useful capacity or lets people do other work. Do not count the same benefit twice. If the financial case only works with optimistic output and no interruptions, ask for better evidence before committing.

Finish with a clear recommendation

Use one of four outcomes: proceed within the tested scope; proceed once named conditions are met; gather more evidence; or stop. Set an owner and due date for every condition. Important unresolved safety, technical or support questions cannot be cancelled out by a high overall score.

The final record should fit into a short decision summary backed by the logs, quotations and evidence. It should explain why the choice makes sense, what its limits are and what would trigger a review.

Robodromo can help structure the brief, supplier comparison and commercial assessment under an agreed scope. Technical validation and site delivery require a named competent provider. Tell us about your proposed project, or start with our UK robot buying checklist.

Editorial illustration: an AI-generated fictional trial. It does not show a real customer project or approved operating layout. Manufacturer source checked 8 October 2026.