# How to evaluate a robotics company
> Evaluate robotics claims through use cases, deployment evidence, performance measurement and commercial questions, using official research.
Canonical HTML: [https://orbitrum.io/academy/sector-analysis/robotics-company-due-diligence/](https://orbitrum.io/academy/sector-analysis/robotics-company-due-diligence/)
Prepared for Orbitrum Academy
Published: 2026-10-08
Source-based educational synthesis prepared with AI assistance. Proposed worksheets and fictional examples are labelled. No named expert review is claimed.
Educational material, not personalized investment, legal, tax or safety advice. Research access on Orbitrum depends on the expert's offer.
Evaluating a robotics company starts with a specific task, operating environment and customer. Examine what the system has demonstrated, how performance was measured, what human support it requires and what the commercial evidence establishes. A compelling demonstration is a research input, not a complete assessment of deployment readiness or business value.
## Start with the application, not the appearance of the robot
The public contents of IFR's World Robotics 2025 report distinguish professional, medical and consumer robots, with professional applications including transport, cleaning, inspection and agriculture. This supports organizing research by use case rather than treating every robot as interchangeable. Only the public contents are used here, not the report's paid market data. [IFR, 2025](#source-ifr-2025)
Describe the proposed use in one sentence: the task, the environment, the operator and the result the customer needs. "Moves containers between these workstations" is a more testable starting point than "automates industry."
The research worksheet below is an Academy adaptation, not an IFR investment model or a robot safety certification.
## Build a deployment evidence table
NIST's 2024 mobile-manipulator research describes controlled measurements across localization methods, base speeds and positions relative to an apparatus. It illustrates why the conditions of a performance result matter. Results from that experiment do not rank commercial robots or establish how an unrelated system will perform. [NIST AMS 100-57](#source-nist-robots-2024)
Use that measurement-oriented approach to ask for the following information about the system being researched:
| Question | Evidence to request or locate |
|---|---|
| What is the task? | A defined start state, end state and success condition |
| Where was it tested? | Environment, obstacles, materials and constraints |
| Which configuration ran? | Hardware, software and relevant version information |
| How much help was needed? | Operator actions, remote interventions and resets |
| What failed? | Failed attempts and excluded runs, not only the best demonstration |
| How was the result measured? | Method, sample, duration and an appropriate baseline |
| What changed during testing? | Updates or configuration changes that affect comparability |
An unavailable item should remain unavailable in the research note. Do not infer unattended operation from a video that simply does not show the operator.
## Distinguish the maturity question from the business question
GAO's technology-readiness guide examines how maturity has been demonstrated and whether technology is ready for integration into a larger system. It was developed for acquisition programs, not to rate startup investments. [GAO-20-48G](#source-gao-2020)
For this learning exercise, ask two separate questions: "What evidence supports the system working in the intended setting?" and "What evidence supports a customer choosing and paying for it?" A successful answer to either question is not automatically a successful answer to the other.
Continue with the [demo-versus-deployment guide](/academy/technical-analysis/demo-vs-deployable-product/) for the technology evidence and the [traction guide](/academy/commercial-analysis/startup-traction-evidence/) for the customer evidence.
## A worked reading exercise
The following numbers are fictional and are provided only to practice reading a test result. They are not an industry benchmark or a result for any company.
A demonstration report says that a robot completed 96 of 100 tasks. During 18 of the completed tasks, a remote operator intervened. Every intervention-assisted task is counted within the 96 completed tasks.
The report can therefore support a **96% completion rate under the test definition**. It cannot support a **96% unassisted completion rate**. Under the stated assumptions, 78 of 100 tasks finished without an intervention.
Neither figure alone answers whether the robot is suitable for a particular customer. The exercise leaves task difficulty, run duration, safety, operator time and customer requirements unspecified. Those are questions to resolve, not reasons to assume that the system is either good or bad.
Write the conclusion so that the measurement and the unknowns survive together: "Most tasks completed in this test, but some needed remote support; performance against the proposed deployment requirements remains to be established."
## Examine the commercial arrangement without inventing a universal threshold
The EIC Fund treats commercial and technology due diligence as connected but distinguishable work. In a robotics review, make the commercial questions explicit instead of embedding them in a technology score. [EIC Fund, section 2.8](#source-eic-2023)
Suggested research questions include: is the arrangement a purchase, lease, service or pilot; who handles installation and support; which commitments are documented; and what evidence distinguishes a demonstration partner from a paying repeat customer?
This guide sets no required gross margin, payback period, customer count or funding amount. A threshold chosen without reference to the actual product and customer would create false precision.
## Treat safety and AI claims as separate evidence requests
Where AI is part of the system, NIST's AI Risk Management Framework provides a voluntary approach to considering context, measurement, governance and risk management. It does not certify a robot or make a deployment legally compliant. [NIST AI RMF 1.0](#source-nist-ai-2023)
For the research memo, ask what safety assessment and operational constraints apply to the intended setting, who prepared them and what system version they cover. The relevant requirements depend on the product and jurisdiction. This Academy does not provide machinery-safety instructions or a substitute for qualified assessment.
## Write a robotics assessment people can inspect
Use a short sequence: task and environment; evidence actually reviewed; measured result; remaining deployment gap; commercial question; next evidence needed. Keep the source attached to each important observation.
For readers, the useful question is not "Does the expert like robotics?" It is "Can I follow the reasoning from the evidence to this conclusion?" For authors, a bounded conclusion can be more informative than an impressive adjective.
## Continue on Orbitrum
Use [Sectors](/sectors/) to locate relevant companies, then examine the public profiles and available research. The catalog follows its application taxonomy; a Robotics label is not assumed to exist, and not every industrial company should be treated as a robotics company.
## Questions about robotics research
### Does a humanoid form prove a more capable or investable product?
Not on its own. This guide asks which task and environment the evidence covers. Product appearance does not logically answer deployment, customer or investment questions.
### Does a customer logo prove a paid deployment?
Not by itself. Locate the underlying statement or documentation and distinguish a demonstration, pilot, commercial agreement and ongoing use. The exact relationship matters.
### Should all robots use the same performance metric?
This guide does not recommend one universal metric. Define success for the task being assessed, describe the measurement conditions and preserve the limits of the result.
## Explore the research on Orbitrum
Browse public project profiles and the research available from their authors. Public previews are free; full research access depends on the expert's offer.
- [Explore projects](https://orbitrum.io/)
- [Discover experts](https://orbitrum.io/?mode=experts)
- [Browse sectors](https://orbitrum.io/sectors/)
## Sources and reading notes
### World Robotics 2025 - Service Robots: public contents
Publisher/authors: International Federation of Robotics / VDMA Services
Date: 2025
Type: Industry-association report; public contents only
Locator: Public contents: professional, medical and consumer robots; application groups
Official source: https://ifr.org/img/worldrobotics/Contents_WR_2025_Service_Robots.pdf
Supports: Application-oriented scope, including logistics, cleaning and medical robotics.
Scope and limitations: Only the public contents were consulted. Do not claim to have reviewed the paid full report, its case studies or its market estimates.
Sources checked: 2026-10-08
### Continuous Mobile Manipulator Performance Measurement Data
Publisher/authors: Aboul-Enein et al.; NIST Advanced Manufacturing Series 100-57
Date: January 18, 2024
Type: Official experimental research report
Locator: Abstract and linked report
Official source: https://www.nist.gov/publications/continuous-mobile-manipulator-performance-measurement-data
DOI: 10.6028/NIST.AMS.100-57
Supports: A controlled experiment varying localization method, mobile-base speed and apparatus side; moving-platform uncertainty and context-dependent measurements.
Scope and limitations: One industrial mobile-manipulator experimental setting. Not a commercial ranking of robotics companies or a benchmark that validates all robot categories.
Sources checked: 2026-10-08
### Technology Readiness Assessment Guide: Best Practices for Evaluating the Readiness of Technology for Use in Acquisition Programs and Projects
Publisher/authors: US Government Accountability Office
Date: 2020; reissued February 11, 2020
Type: Official technical assessment guide
Locator: Overview and full guide, GAO-20-48G
Official source: https://www.gao.gov/products/gao-20-48g
Supports: Evidence-based assessment of technology maturity and readiness for integration; contextualized demonstrations.
Scope and limitations: Government acquisition context. Technology maturity is not company valuation, customer adoption, profitability, or an Orbitrum score. Retained as a foundational reference rather than labeled recent research.
Sources checked: 2026-10-08
### EIC Fund Investment Guidelines, Horizon Europe Compartment
Publisher/authors: European Innovation Council
Date: December 2023
Type: Official investment-process guidance
Locator: Section 2.8, especially 2.8.5-2.8.7; printed p.18 (PDF page 18)
Official source: https://eic.ec.europa.eu/document/download/f2d3a62e-76c7-45ef-a6a8-39bb67a5774a_en
Supports: Financial/commercial due-diligence dimensions and the distinction from additional technology due diligence.
Scope and limitations: The EIC Fund mandate is not a universal investor checklist. No evidence that following this Academy guarantees investment performance. The downloaded cover identifies December 2023; an old query-string filename misleadingly says March 2022.
Sources checked: 2026-10-08
### Artificial Intelligence Risk Management Framework (AI RMF 1.0)
Publisher/authors: National Institute of Standards and Technology
Date: 2023
Type: Official voluntary risk-management framework
Locator: AI RMF 1.0 and its Govern, Map, Measure, Manage functions
Official source: https://www.nist.gov/itl/ai-risk-management-framework
DOI: 10.6028/NIST.AI.100-1
Supports: Contextualized AI risk assessment, measurement, governance and management.
Scope and limitations: Not an investment scoring formula, product certification, or a claim of legal compliance.
Sources checked: 2026-10-08
## Continue learning
- [How to distinguish a demo from a deployable product](https://orbitrum.io/academy/technical-analysis/demo-vs-deployable-product/)
- [How to examine a startup's traction evidence](https://orbitrum.io/academy/commercial-analysis/startup-traction-evidence/)