Current methodology version: 2.0
Effective date: July 10, 2026
Status: Current
Humanoid Analytics tracks one of the fastest-moving and most heavily promoted technology markets in the world. Our purpose is simple: to separate observable commercial progress from demonstrations, announcements, projections, and hype.
A robot shown in a controlled video is not the same as a robot working for a customer. A partnership announcement is not the same as a paid pilot. A shipment claim is not the same as an operational deployment. A funding round can strengthen a company without proving that its robots work reliably or economically.
Our evidence standards make these differences visible. Every material classification is based on the strength of the underlying sources, the type of activity observed, the degree of customer and commercial verification, and the recency of the evidence.
This page is the governing methodology for Humanoid Analytics trackers, company profiles, and commercial-readiness analysis. When a tracker summary conflicts with this page, the current version of this methodology controls.
Methodology at a Glance
Humanoid Analytics evaluates market claims through five separate lenses:
- Claim status: Is the claim confirmed, partially confirmed, unverified, delayed, or contradicted?
- Evidence Score: How strong is the best public evidence of real-world deployment and commercial activity?
- Confidence level: How certain are we that the underlying facts and classification are accurate?
- Lifecycle status: Is the activity planned, active, completed, expanded, inactive, or stale?
- Source quality: Is the evidence based on customer confirmation and official records, or mainly on company statements and promotional material?
These measures answer different questions. A company can have a highly credible laboratory demonstration with a low Evidence Score. It can also have a major deployment claim with low confidence because the customer, payment status, scale, or operating results have not been independently confirmed.
What We Track
Humanoid Analytics covers:
- Humanoid robotics companies and major corporate programs
- Robots, enabling technologies, and product platforms
- Customer pilots, paid pilots, and operational deployments
- Repeat deployments, fleet expansion, and commercial-scale use
- Funding, valuations, acquisitions, IPOs, and strategic investments
- Manufacturing facilities, production targets, and shipment claims
- Partnerships, supply chains, and distribution agreements
- Technology demonstrations and performance claims
- Pricing, labor substitution, uptime, service cost, and operating economics
- Safety, autonomy, teleoperation, reliability, and human supervision
We focus on evidence that helps answer practical questions:
- Which robots are moving beyond controlled demonstrations?
- Which companies have identifiable customer activity?
- Which deployments are paid, operational, repeated, or expanding?
- Which claims are supported by customers or independent evidence?
- Which companies are manufacturing products rather than announcing targets?
- Which applications appear capable of delivering measurable economic value?
- Which important claims remain unverified, delayed, or contradicted?
Claim Status
Every material claim may receive one of five statuses.
| Claim status | Definition |
|---|---|
| Confirmed | The central claim is supported by strong, directly checkable evidence such as customer confirmation, a financial or regulatory record, procurement documentation, or multiple reliable independent sources. |
| Partially confirmed | Important elements are supported, but the scale, payment status, timing, operating status, autonomy, customer identity, or another material detail remains unverified. |
| Unverified | The claim is publicly available but depends mainly on a company statement, promotional content, anonymous sourcing, or reporting that cannot be independently checked. Unverified does not mean false. |
| Delayed | A previously announced launch, production target, pilot, deployment, or other milestone has not occurred within the stated period, or later evidence indicates postponement. |
| Contradicted | Stronger available evidence conflicts with the central claim, timeline, quantity, operating status, or commercial characterization. |
Claim status is assigned at the claim level, not automatically to an entire company. A company may have a confirmed funding round, a partially confirmed customer pilot, and an unverified production target at the same time.
Source Quality
Sources are evaluated according to their independence, specificity, direct knowledge, and verifiability.
| Source tier | Typical sources | Weight |
|---|---|---|
| Tier 1: Direct or official evidence | Regulatory filings, financial statements, public procurement records, court records, direct customer confirmation, official records, or operating data attributable to the customer site | Very strong |
| Tier 2: Strong independent evidence | Independent reporting with named sources, site access, photographs or video from the operating environment, identifiable documents, or detailed research from a credible institution | Strong |
| Tier 3: Detailed first-party disclosure | Company announcements, investor materials, product documentation, executive statements, or demonstration material containing specific dates, quantities, customer names, and operating details | Moderate |
| Tier 4: Promotional or incomplete evidence | Edited demonstrations, event presentations, vague partnership announcements, unsourced media summaries, social posts, or claims without sufficient operating context | Weak |
| Tier 5: Unverifiable information | Anonymous claims, copied lists without traceable sourcing, AI-generated summaries presented as evidence, rumors, or marketing language without checkable facts | Not sufficient on its own |
Source tier does not determine truth by itself. A company announcement may be accurate, while independent reporting may contain errors. The tier indicates how much verification the source provides and how heavily it should influence a classification.
Evidence Score
The Evidence Score is a 0 to 10 assessment of a documented deployment or commercial event associated with a company or robot. It measures the strength of public evidence, not technical sophistication, brand visibility, funding strength, company quality, or long-term potential.
Each qualifying event is scored separately. A company’s Best Evidence Score is the highest supported event score in its current record. It is not an average of all company activity and should always be read alongside the underlying event, evidence tier, lifecycle status, and confidence level.
The score is calculated across five dimensions.
1. Source Quality, 0 to 3 Points
| Points | Requirement |
|---|---|
| 3 | Direct customer confirmation, official record, regulatory filing, or strong independent evidence from the operating environment |
| 2 | Detailed, attributable company disclosure or commercial agreement with identifiable parties and specific facts |
| 1 | Company-only statement, demonstration, event presentation, or secondary reporting with limited verification |
| 0 | Untraceable, anonymous, or unverifiable information |
2. Real-World Operation, 0 to 3 Points
| Points | Requirement |
|---|---|
| 3 | Repeated or continuing operation in a real customer environment beyond a limited trial |
| 2 | Active or completed customer pilot in a real operating environment |
| 1 | Planned pilot, internal testing, controlled demonstration, preorder, or shipment without operating proof |
| 0 | No evidence that the robot has operated outside development or promotional settings |
3. Commercial Commitment, 0 to 2 Points
| Points | Requirement |
|---|---|
| 2 | Paid pilot, purchase, repeat order, expansion, disclosed revenue, or binding commercial deployment |
| 1 | Named pilot, customer agreement, preorder, or shipment linked to an identifiable customer, with payment status unclear |
| 0 | No identifiable customer or commercial commitment |
4. Measurable Operating Proof, 0 to 1 Point
| Points | Requirement |
|---|---|
| 1 | Specific operating metric such as active units, hours worked, tasks completed, throughput, uptime, error rate, productivity, labor impact, or deployment duration, with sufficient context |
| 0 | No meaningful operating metric |
5. Continuity and Recency, 0 to 1 Point
| Points | Requirement |
|---|---|
| 1 | Evidence of repeat use, expansion, follow-on activity, or a separate later verification that the activity continued within the applicable review period |
| 0 | A single announcement or event with no subsequent confirmation, or evidence that is stale for the current-status claim |
Evidence Tiers
| Score | Evidence tier | Interpretation |
|---|---|---|
| 9 to 10 | Operating proof with metric | Strong evidence of real customer operation, commercial commitment, and measurable activity |
| 7 to 8 | Named customer deployment evidence | Credible pilot or deployment evidence involving an identifiable customer, but some questions about scale, economics, or repetition may remain |
| 5 to 6 | Commercial signal, limited operating proof | A meaningful customer, shipment, preorder, or deployment signal exists, but operating or payment evidence is incomplete |
| 3 to 4 | Demo or product signal only | The product is visible, orderable, demonstrated, or associated with early activity, but real customer operation is not sufficiently established |
| 0 to 2 | Insufficient commercial evidence | Prototype, early concept, vague claim, or no qualifying deployment event |
Score Ceilings and Safeguards
Component points alone do not allow weakly sourced claims to reach high evidence tiers. The following safeguards apply after the component score is calculated:
- Evidence based only on Tier 5 sources cannot score above 1.
- Evidence based only on Tier 4 sources cannot score above 2.
- A company-only claim without customer, official, or credible independent corroboration cannot score above 6.
- An event without a named customer, identifiable operating site, or equivalent verifiable context cannot score above 4.
- Planned future activity, including announced pilots and deployment agreements that have not started, cannot score above 4.
- Scores of 7 or 8 require an identifiable customer and evidence that the pilot or deployment has started.
- Scores of 9 or 10 require strong customer or official evidence, real-world operation, commercial commitment, and a meaningful operating metric.
These ceilings prevent a detailed promotional claim from receiving the same treatment as customer-confirmed operating evidence.
Funding Does Not Increase the Evidence Score
Funding, valuation, investor quality, and manufacturing ambition may affect a company’s resources and ability to execute. They do not increase the Evidence Score unless the associated deployment evidence independently meets the criteria above.
A company can therefore be very well funded and still have a low deployment score. A less visible company can have a higher score if it has stronger evidence of useful customer operation.
Deployment Evidence Levels
The Evidence Score is complemented by a plain-language deployment level.
| Deployment level | Definition |
|---|---|
| Lab demonstration | Capability shown in a controlled development environment with no customer-use evidence |
| Public demonstration | Capability shown at an event, in a promotional video, or through a staged public presentation |
| Internal testing | Testing by the developer or an affiliated organization without independent customer-deployment evidence |
| Planned customer pilot | A named customer or partner has announced a future trial, but operating activity has not yet been shown |
| Active customer pilot | A robot is being evaluated in a real customer environment, with payment or production use unclear |
| Paid pilot | Payment, commercial terms, procurement, or revenue involvement has been confirmed or reliably reported |
| Operational deployment | The robot is performing useful work in a real operating environment beyond a staged demonstration or limited evaluation |
| Repeat deployment | The customer has extended, renewed, repeated, or expanded the initial deployment |
| Commercial-scale deployment | Deployment spans multiple sites, customers, fleets, or workflows at a scale that indicates meaningful recurring commercial activity |
Movement between levels is based on new evidence. A future agreement does not become an active pilot solely because its scheduled date has arrived. Operating evidence must be observed or confirmed.
Confidence Level
Confidence measures the reliability of the classification, not the commercial maturity of the company.
| Confidence | Standard |
|---|---|
| High | At least one Tier 1 source, or multiple consistent Tier 2 sources, supports the classification. Material details are specific and no major conflict remains unresolved. |
| Medium | The classification is supported by a detailed Tier 2 or Tier 3 source with partial corroboration, but one or more important details remain uncertain. |
| Low | Evidence is company-only, promotional, vague, stale, difficult to verify, or materially disputed. |
A high-confidence prototype is still a prototype. A low-confidence deployment claim may still be important, but it should not be treated as proven commercial activity.
Lifecycle and Recency
Historical events do not disappear when they become old, but they must not be presented as evidence of current activity without recent confirmation.
Each event may be marked as:
- Planned: Announced but not yet shown to have started
- Active: Currently operating or being executed
- Completed: The stated pilot, deployment, or transaction concluded
- Expanded: Follow-on activity, additional units, sites, tasks, or customers are confirmed
- Inactive: Activity ended, was discontinued, or is no longer operating
- Stale: The event remains historically valid, but current status has not been verified within the applicable review period
Review intervals depend on the claim type:
| Claim type | Target review interval |
|---|---|
| Active deployment, manufacturing, or shipment claim | 90 days |
| Pilot, preorder, commercial agreement, or production plan | 120 days |
| Prototype or public demonstration | 180 days |
| Funding, acquisition, or completed historical transaction | Reviewed when new information emerges |
If an active-status claim cannot be refreshed within 12 months, it is marked stale. If no supporting activity is found for 24 months, it may be removed from current rollups while remaining in the historical record.
Standards by Claim Type
Different claims require different evidence.
| Claim area | Stronger evidence | Weaker evidence |
|---|---|---|
| Customer activity | Customer confirmation, procurement record, paid pilot, repeat order | Logo use, memorandum, generic partnership |
| Deployment | Robots operating in a real customer environment with dates and tasks | Lab video, event demonstration, future rollout statement |
| Manufacturing | Operating facility, supplier records, production output, delivered units | Factory rendering, future capacity target, opening announcement |
| Shipments | Units linked to customers, delivery dates, and operating status | Aggregate shipment claim without recipients or use context |
| Funding | Filing, investor confirmation, company announcement with transaction details | Target raise, rumored valuation, unspecified strategic financing |
| Technology | Repeated task performance in realistic conditions with runtime and failure information | Edited demonstration without autonomy, assistance, or failure disclosure |
| Economics | Price, uptime, supervision, service cost, productivity, payback, renewal | Broad labor-replacement or cost-saving claim |
Autonomy and Human Assistance
Humanoid robot performance should not be described as autonomous unless the available evidence supports that characterization.
Where material, Humanoid Analytics distinguishes among:
- Fully autonomous operation
- Supervised autonomy
- Human-in-the-loop assistance
- Remote intervention
- Continuous teleoperation
- Scripted or preprogrammed behavior
- Unknown or undisclosed control mode
Teleoperation and human assistance can be commercially useful. They are not treated as failures. They must, however, be disclosed when they materially affect claims about autonomy, labor substitution, cost, safety, or scalability.
How Conflicting Evidence Is Handled
When sources conflict, Humanoid Analytics does not silently choose the most favorable or most recent claim.
We apply the following rules:
- Direct customer evidence and official records generally outweigh company marketing.
- A later source does not automatically outweigh an earlier source if it is less direct or less specific.
- The disputed elements are separated from facts that remain supported.
- Both positions are documented when the conflict cannot be resolved.
- The claim is marked partially confirmed, delayed, or contradicted when the evidence requires it.
- Confidence is reduced until the material conflict is resolved.
Research and Review Process
Material tracker entries follow a consistent workflow:
- Capture the claim: Record the exact statement, date, organization, and source.
- Decompose the claim: Separate customer identity, quantity, timing, payment, autonomy, operating status, and performance into checkable elements.
- Collect sources: Prioritize direct customer evidence, official records, and reliable independent reporting.
- Classify the sources: Assign source tiers based on independence, specificity, and verifiability.
- Score the event: Apply the published 0 to 10 evidence model.
- Assign status and confidence: Determine what is known, uncertain, delayed, or contradicted.
- Record recency: Publish the event date, last-checked date, and lifecycle status where available.
- Review material classifications: Scores of 7 to 10 and claims of commercial-scale deployment receive an additional source recheck before publication.
- Monitor and correct: Update classifications when stronger or newer evidence becomes available.
Worked Example
The following hypothetical example shows how evidence can progress without overstating commercial maturity.
Stage 1: Public Demonstration
A company publishes an edited video of one robot performing a task in its own laboratory. No customer, duration, payment, or operating metric is disclosed.
- Source quality: 1
- Real-world operation: 0
- Commercial commitment: 0
- Measurable operating proof: 0
- Continuity and recency: 0
- Evidence Score: 1, insufficient commercial evidence
- Demonstration claim status: Confirmed
- Commercial-readiness conclusion: Not established
- Confidence: High that the demonstration occurred
Stage 2: Named Customer Pilot
The developer and a named manufacturer confirm that a pilot is active at a customer facility. A later customer update confirms that the pilot is continuing. The robot performs a defined task, but payment, scale, uptime, and productivity are not disclosed.
- Source quality: 3
- Real-world operation: 2
- Commercial commitment: 1
- Measurable operating proof: 0
- Continuity and recency: 1
- Evidence Score: 7, named customer deployment evidence
- Pilot claim status: Confirmed
- Commercial status: Partially confirmed
- Confidence: High
Stage 3: Repeat Paid Deployment
The customer confirms that ten robots have operated across three sites for six months under a paid agreement. The customer reports task volume, uptime, and expansion into an additional workflow.
- Source quality: 3
- Real-world operation: 3
- Commercial commitment: 2
- Measurable operating proof: 1
- Continuity and recency: 1
- Evidence Score: 10, operating proof with metric
- Claim status: Confirmed
- Confidence: High
Corrections
Humanoid Analytics updates articles, trackers, and classifications when new or stronger evidence becomes available.
Material factual corrections include:
- The date of the correction
- The previous statement or classification
- The corrected information
- The reason for the change
- The evidence supporting the correction
Silent changes are limited to spelling, formatting, broken links, and other edits that do not alter the substance of a claim. Changes to funding amounts, customer identity, deployment status, Evidence Score, confidence, or commercial classification are documented.
Material corrections are published in the following format:
Correction, [date]: This entry previously stated [previous information]. It has been updated to [corrected information] based on [source or new evidence]. The Evidence Score changed from [old score] to [new score].
Company Responses and Right of Correction
Companies, customers, investors, researchers, and readers may submit supporting evidence or request a correction. Requests should identify the specific statement, provide a source, and explain the proposed change.
Evidence and correction requests may be submitted through the Humanoid Analytics contact page.
Humanoid Analytics reviews submitted evidence but does not guarantee a requested classification. Companies do not receive advance approval over independent analysis, Evidence Scores, or editorial conclusions.
Independence and Disclosures
Humanoid Analytics does not offer paid rankings or allow commercial relationships to determine evidence classifications.
Any material financial, advisory, sponsorship, research, data, or partnership relationship involving a covered organization is disclosed alongside the relevant content.
Material relationships are disclosed in the following format:
Disclosure: Humanoid Analytics has [relationship] with [organization]. The organization did not determine the evidence classification or editorial conclusion.
Sponsored content, if published, is clearly labeled and kept separate from tracker classifications and independent analysis.
Use of Artificial Intelligence
Artificial intelligence may assist with source discovery, transcription, translation, document comparison, data extraction, and duplicate detection. AI output is not treated as evidence and is not cited as a source.
No material claim status, Evidence Score, confidence level, or correction is published without human review of the underlying sources. When translation affects a material classification, the original-language source is retained where possible.
Scope and Limitations
Humanoid Analytics is based primarily on publicly available information. Some companies, customers, and governments do not disclose commercial terms, operating metrics, deployment quantities, or failures. The absence of public evidence does not prove that an activity has not occurred. It means the activity cannot be assigned a stronger public evidence classification at that time.
Evidence Scores are comparative research judgments produced under this published methodology. They are not technical safety certifications, audits, credit ratings, investment recommendations, or guarantees of future performance.
Scores may change when stronger evidence becomes available, a deployment expands or ends, a claim becomes stale, or a material error is corrected. Historical changes remain visible through dated tracker records or correction notes.
What We Do Not Treat as Commercial Proof
The following may be important signals, but none proves commercial readiness on its own:
- Demo videos
- Trade-show appearances
- Founder or executive interviews
- Funding rounds and high valuations
- Partnership announcements and memorandums
- Customer logos without confirmation
- Future production targets
- Factory renderings or announced capacity
- Preorders without delivery evidence
- Shipment claims without customer or operating context
- Generic statements about AI, autonomy, embodied intelligence, or labor replacement
- Technical benchmarks that do not reflect real operating conditions
Funding can reduce financing risk. Demonstrations can show technical progress. Partnerships can create future opportunities. Preorders can indicate demand. These signals matter, but they must be described accurately and must not be presented as proof of real-world commercial adoption.
Our Editorial Standard
Humanoid Analytics is evidence-first.
We distinguish technical achievement from commercial proof. We distinguish a planned pilot from an active one, an active pilot from a paid deployment, and a shipment from useful work. We preserve uncertainty when the public record does not support a stronger conclusion.
The humanoid robotics market will not be proven by videos, valuations, or announcements alone. It will be proven by robots performing useful work reliably, safely, and economically in real operating environments.