Market Intelligence

A glossary of building and asset intelligence terms

Define the buildings, equipment, evidence, buying roles and market quantities behind a useful prospect file, with Anvilfield estimates and a shared review checklist.

Last reviewed: Nov 5, 2026By the Anvilfield Team20 min read

Market IntelligenceAsset IntelligenceBuilding DataGlossaryData Quality

Direct answer

Building and asset intelligence is the practice of locating physical buildings and equipment, documenting what is known about them, and connecting that evidence to a commercial decision. Its core terms distinguish property units, equipment units, evidence quality, purchasing roles and market quantities. Use this glossary when writing a research brief, accepting a data delivery or explaining an estimate, so everyone counts the same objects and understands what the evidence supports.

On this page

Which methods do these terms describe?How do buildings, parcels, sites and accounts differ?What do area, asset class and equipment counts measure?What do location and evidence terms tell you?How do precision, recall and confidence differ?What do TAM, SAM, SOM and annual demand mean?What does Anvilfield estimate when the counting unit changes?Which rules keep a building intelligence file usable?Worked example: a regional HVAC distributor with four branches How should leaders use the glossary when buying and maintaining data?Building and asset intelligence release checklist Methodology

Need to run this in the field?

Open the related tool for the calculation, checklist, or packet.

Key takeaways

Buildings, equipment assemblies and buying accounts need separate linked identifiers and totals.
Anvilfield estimates 433 packaged-cooling buildings and 1,299 units in a 1,000-building reference territory.
Precision checks returned detections; recall checks coverage against an independently established reference population.
Carry the definition, evidence date, source and counting rule into every market estimate and sales handoff.

Which methods do these terms describe?

Building-first prospecting means starting with the building and what is visible on it, then finding the decision maker. It gives contact research a physical reason to pursue a property. Preserve that reason in the account record, even when the people responsible for the property change.

Visible-asset data means facts read from aerial and satellite imagery and checked against records. A cabinet outline, tank location or paved boundary belongs in this category. An inferred brand or operating condition belongs in a separate field with the evidence supporting that inference.

Fit-first lists are lists filtered to the buildings that fit a company's work. State the roof, equipment, size, territory and access requirements before selecting records. Keep physical fit separate from buying readiness so a sales team can tell why a building entered the file.

Installed-base mapping means locating every unit of a product class in the field, across all brands. Define the unit being counted and track coverage gaps. A manufacturer shipment history contributes evidence, while an installed-base register also investigates competing equipment and unresolved identities.

Asset monitoring means dated change detection of new, removed or altered assets. Preserve earlier observations when adding a later view. A change record identifies what became visible between observations; a commissioning record establishes a different kind of event.

Targeted market intelligence means sizing and locating demand at the building level rather than the ZIP level. It connects a defined physical population to the product, channel, service capacity and purchasing roles needed to reach that population.

Concept guides

Building-first prospecting; Visible-asset data; Fit-first lists; Installed-base mapping; Asset monitoring; Targeted market intelligence

How do buildings, parcels, sites and accounts differ?

Write the counting unit beside every total. A campus register, an equipment inventory and a customer file describe different relationships. Link their identifiers so a reviewer can move from a purchasing organization to a building and then to the equipment that supports the proposed work.

The Census Bureau publishes County Business Patterns as establishment statistics. Use those data to understand industry activity, then investigate the actual buildings independently. Matching an establishment address to a parcel does not resolve which structure it occupies or who controls a shared roof.

Census Bureau: About County Business Patterns, accessed 2026Census Bureau: About County Business Patterns, accessed 2026

Owner of record vs operator vs decision maker: who actually buysOwner of record vs operator vs decision maker: who actually buys

Building: A physical structure represented as its own inventory object. Use a stable building identifier and a defined footprint convention. Keep connected wings and separate structures consistent with the counting rules for the project.
Parcel: A land-record unit with an identifier in a particular jurisdiction. Store that jurisdiction with the identifier. Link buildings to parcels explicitly when a site spans boundaries or contains several structures.
Site: The operating location used for a service or commercial decision. A site groups the buildings, yards, lots and other assets managed together. Its boundary should match the intended account or work scope.
Establishment: A business operating location used in economic statistics. Establishment, building and company totals use different units. Retain the source definition when combining economic context with a physical property inventory.
Owner of record: The person or organization named as owner in the property record being used. Retain the record date and jurisdiction, then research the role that actually authorizes the proposed purchase.
Operator: The organization running the activity at a site. Record whether it controls equipment, maintenance, access or purchasing. The operating business and the property owner belong in separate linked fields.
Decision maker: The role with authority over a particular buying decision. Record the decision being researched, such as maintenance renewal or capital replacement, along with the organization and evidence supporting the role.
Buying account: The customer relationship used to manage a purchasing organization or buying center. Several buildings can share an account. Preserve the building links when consolidating contacts, quotes and service agreements.

What do area, asset class and equipment counts measure?

A useful quantity has a boundary, unit and measurement basis. Keep the original measurement as well as any quantity calculated from it. The EIA gross-floor-area field includes finished and unfinished building space; converting it into roof purchasing volume requires a separate physical model.

Use the technical guide for the work to decide which quantities matter. A roof restoration scope needs system condition and coating suitability. A rooftop equipment inventory needs verified assemblies and nameplates. A parking lot scope needs a measured work boundary. A tank or cooling tower register needs an equipment-specific counting convention.

Commercial roof maintenance: inspection schedules and recordsCommercial roof maintenance: inspection schedules and records

Commercial roof asset management and capital planningCommercial roof asset management and capital planning

Low-slope roof coating restoration system field guideLow-slope roof coating restoration system field guide

Rooftop unit installation and startup field guideRooftop unit installation and startup field guide

Parking lot ADA accessibility layout field guideParking lot ADA accessibility layout field guide

Fuel storage tank UST and AST systems field guideFuel storage tank UST and AST systems field guide

Cooling tower types and how they work field guideCooling tower types and how they work field guide

Gross floor area: The sum of included floor space across a building. Preserve the source inclusion rules. Basements, parking levels and unfinished areas affect the result; the total is a whole-building measurement.
Building footprint: The plan-view ground area occupied by a building under a stated outline rule. Distinguish the footprint polygon from a point location, parcel outline and gross floor area before using it in a takeoff.
Roof area: The roof surface quantity defined for a particular measurement or scope. Record whether it is projected area or actual sloped surface area and how openings, overhangs and separate roof sections were treated.
Asset class: A defined category of physical objects sharing a useful function or counting rule. Identify the inclusions and look-alikes before searching. Preserve an unresolved category for objects the evidence does not yet classify.
Equipment assembly: The complete machine or system counted as a unit for the inventory. Record assemblies and their components at separate levels so a serviceable component does not become an extra installed machine.
Installed base: The population of installed assets within a defined product class, geography and observation period. Count the physical stock, retain retired status in its history, and separate installed quantity from annual replacement flow.

What do location and evidence terms tell you?

A position answers where to look; an evidence record explains why a classification was made. Keep geometry, source and review fields separate. The Census Geocoder documentation describes its coordinates as interpolated from address ranges, so an address match still needs a building association.

Document the observation that changed the record. A newly uploaded photograph and a newly captured photograph have different ages. A permit entry and a completed installation establish different statuses. Record each as its own evidence item and preserve the dates needed to reconcile them.

Census Geocoder technical documentation, accessed 2026Census Geocoder technical documentation, accessed 2026

Drone and UAV roof and facade inspectionDrone and UAV roof and facade inspection

BIM and VDC coordination and clash detectionBIM and VDC coordination and clash detection

Digital twin for facility operationsDigital twin for facility operations

Geocoding: Converting an address or other location description into coordinates. Store the match method and quality. An address-range coordinate provides a search location, while a verified building association identifies the structure being researched.
Geometry: The spatial representation of a record, such as a point, line or polygon. Store the coordinate reference system with spatial files and preserve the distinction between an asset location and the boundary used to measure it.
Capture date: The date the image or field observation was collected. Keep it distinct from the date the file was delivered, imported or reviewed. Use the capture date when assessing the age of visual evidence.
Provenance: The traceable source history behind a claim. Include the source reference, observation date, extraction method, reviewer and revisions. A later researcher should be able to find the evidence and understand the original decision.
Observed, inferred and unknown: Evidence states that separate directly supported attributes, conclusions drawn from other information and unresolved attributes. Preserve these states in the data model so an inference does not silently become an observed fact.
Ground truth: An independently checked reference used to evaluate a particular label or measurement. Specify its date, coverage and inspection method. A field record supports the attributes it actually checked, with its own unresolved items retained.
Change detection: A comparison that identifies a difference between dated observations of the same asset or location. Record both observations, the type of change and review status. Investigate whether a viewing difference explains the apparent change.

How do precision, recall and confidence differ?

Ask what was counted and what was checked before accepting an accuracy statement. Reviewing returned objects tells you about false detections. Finding missed equipment also requires an independently established population containing equipment the detector did not return.

Anvilfield's quality worksheet gives 90% precision and 75% recall for 100 returned detections checked against 120 reference assets. The file contains 90 supported matches, 10 false positives and 30 false negatives. Keep both denominators in the report so a clean-looking list does not hide missing assets.

Use those ratios to choose the next improvement. False positives consume review and outreach capacity. False negatives leave gaps in territory coverage. Review the ambiguous equipment types, shaded areas and complex sites as separate groups so a strong overall result does not hide a weak operating segment.

Google: Classification, accuracy, recall, precision, accessed 2026Google: Classification, accuracy, recall, precision, accessed 2026

True positive: A returned detection that agrees with the reference label under the written matching rule. Specify the asset class, spatial matching convention and reference date used to establish agreement.
False positive: A returned detection that fails the reference check. Record the failure reason, such as the wrong equipment class, a duplicate assembly or an object outside the requested boundary.
False negative: A reference asset that the search failed to return. Investigating misses requires reference coverage beyond the returned detections. Preserve the asset and its evidence so the missed class can be reviewed.
Precision: The share of returned positive detections that are supported positives. Divide true positives by all returned positives. State whether the denominator is objects, buildings or another counting unit.
Recall: The share of reference positives successfully found. Divide true positives by all reference positives, including misses. Report the reference coverage and matching rule beside the result.
Confidence score: A model or reviewer score associated with a classification. Document what the score represents and how it was evaluated. Measure precision and recall on checked records before turning a score into a purchasing rule.

What do TAM, SAM, SOM and annual demand mean?

Write market terms as a chain from physical stock to serviceable work and then to attainable sales. State whether the value represents buildings, equipment, annual units or revenue. A large installed population supports a research program; purchasing timing requires its own evidence.

Keep access, channel and product suitability visible in the serviceable market. A distributor needs supported equipment families and reachable contractors. A service company needs a defined service scope, travel capacity and a responsible buyer. Use the maintenance agreement guide to connect those filters to the actual obligations being sold.

HVAC maintenance agreements: scope, exclusions, and renewalsHVAC maintenance agreements: scope, exclusions, and renewals

Total addressable market (TAM): The full demand population or spending pool for a defined offer and period. State the geography, product category, unit and time basis. An installed-unit total and an annual revenue total require different labels.
Serviceable available market (SAM): The part of the addressable market supported by the company's product fit, territory, channel and operating capabilities. Retain the exclusion rules so another reviewer can reproduce the filter.
Serviceable obtainable market (SOM): The part of the serviceable market targeted for capture during a stated period. Connect it to selling capacity, channel access, competition and evidence from the sales process.
Annual replacement demand: The flow of equipment replaced during a year. Estimate it from stock, installation history and a stated retirement model, then refine timing with dated equipment histories.
Qualified opportunity: A sales record with supported physical fit, buying responsibility and a defined commercial next action. Keep the qualification evidence separate from the original building or equipment observation.

What does Anvilfield estimate when the counting unit changes?

Anvilfield estimates 5,918,000 commercial buildings in the 2018 US reference stock, ranging from 5,462,000 to 6,375,000, and 96.5 billion square feet of gross floor area, ranging from 90.8 to 102.3 billion. Anvilfield estimates mean gross floor area at 16,300 square feet per building, ranging from 15,500 to 17,100.

Anvilfield estimates 745,000 commercial buildings above 25,000 square feet in that stock, ranging from 692,000 to 799,000. Their estimated share is 12.6% of buildings, ranging from 11.6% to 13.6%.

Anvilfield estimates 2,565,000 buildings used packaged central cooling in the 2018 stock, ranging from 2,350,000 to 2,779,000. The estimated building share is 43.3%, ranging from 40.2% to 46.5%.

Anvilfield estimates 433 equipped buildings in a 1,000-building reference territory, ranging from 402 to 465, and 1,299 packaged units, ranging from 804 to 1,860. Anvilfield estimates 325 buying accounts for those buildings, ranging from 302 to 349.

Anvilfield estimates 65 annual replacement units in that territory, ranging from 32 to 124, and a $6,495 equipped-building list acquisition budget, ranging from $4,020 to $9,300.

Anvilfield estimate	Low	Central	High
US commercial buildings, 2018	5,462,000	5,918,000	6,375,000
US gross floor area, 2018	90.8 billion sq ft	96.5 billion sq ft	102.3 billion sq ft
Mean gross floor area per building, 2018 (sq ft)	15,500	16,300	17,100
US buildings over 25,000 sq ft, 2018	692,000	745,000	799,000
Share of buildings over 25,000 sq ft, 2018	11.6%	12.6%	13.6%
US packaged-cooling buildings, 2018	2,350,000	2,565,000	2,779,000
Packaged-cooling building share, 2018	40.2%	43.3%	46.5%
Equipped buildings in a 1,000-building territory	402	433	465
Packaged units in that territory	804	1,299	1,860
Buying accounts for those equipped buildings	302	325	349
Annual replacement units in that territory	32	65	124
Equipped-building list acquisition budget	$4,020	$6,495	$9,300

How to cite this estimate: Anvilfield, A glossary of building and asset intelligence terms, 2026, https://anvilfield.com/field-guides/market-intelligence/a-glossary-of-building-and-asset-intelligence-terms/

Which rules keep a building intelligence file usable?

Use these as Anvilfield file-design benchmarks. They govern identifiers, evidence and reporting; choose acceptance thresholds separately for the work. A repair dispatch, a national market study and a dealer territory plan need different levels of confirmation.

Keep one persistent building ID per structure. Give each counted equipment assembly its own unit ID. Preserve two dated observations for every monitored physical change. Separate the three identity roles of owner, operator and decision maker. Keep the three evidence states of observed, inferred and unknown visible in the export.

For the quality worksheet, compare 100 returned detections with an independently complete reference inventory. Evaluate two populations: what the system returned and what the reference says exists. Give every reported rate one explicit denominator and a stated period. These rules make the result reviewable when the sales team asks why an account was selected.

Field rule	Anvilfield benchmark	Why it changes the decision
Building identity	1 persistent ID per structure	Prevents tenant or contact changes from duplicating the building
Equipment identity	1 unit ID per counted assembly	Keeps components from inflating equipment stock
Monitored change	2 dated observations	Preserves the interval in which the change was observed
Property and purchasing roles	3 separate role fields	Supports ownership research and buyer qualification
Evidence status	3 states: observed, inferred, unknown	Keeps research tasks visible
Quality worksheet	100 returned detections plus full reference coverage	Supports separate precision and recall calculations
Quality denominator	2 populations: returned and reference	Explains false detections and missed assets separately
Published rate	1 explicit denominator plus period	Prevents inventory size from becoming annual opportunity

Worked example: a regional HVAC distributor with four branches

A regional heating, ventilation and air conditioning (HVAC) distributor with four branches builds a packaged-equipment territory file using Anvilfield's reference case. Sales managers need equipment support opportunities; branch managers need a shared account structure. The register starts with buildings and links units, buying accounts and contractor relationships.

Anvilfield's central territory case contains 433 equipped buildings within a 1,000-building register. Check equipment families, access, channel relationships and the responsible buyer before assigning a service opportunity.

Anvilfield puts that territory at 1,299 units and 325 buying accounts. The branch team keeps those totals in separate columns: list acquisition follows buildings, parts support follows units, and account ownership follows the purchasing relationship.

Anvilfield puts annual replacement volume at 65 units in the central case. The category manager uses that flow to plan replacement research and channel conversations. Nameplates, service histories and owner capital plans refine the timing evidence for particular units.

The equipped-building list budget is $6,495. The operations manager assigns review work across branches by service coverage, while retaining a common identifier register. Shared accounts receive coordinated ownership so the same portfolio does not appear as unrelated opportunities in competing branch reports.

The useful handoff is a building record with linked equipment, evidence, account and follow-up fields. The distributor can revise the asset count without losing the account history, and revise the buyer without rewriting the equipment inventory. Each change has a reason that the next branch reviewer can understand.

How should leaders use the glossary when buying and maintaining data?

Make the glossary part of the research brief. Define which objects qualify, what a delivered row represents and which fields require review. Ask a supplier to explain those choices against the actual purchasing decision. A shared vocabulary makes a pilot easier to accept and a later territory expansion easier to compare.

Leading manufacturers, distributors and service companies build the research brief around measurable building features. PorchRocket's Ground Control supplies standard commercial building lists filtered by features such as roof age band, roof material, size and rooftop HVAC, delivered within 48 hours. Write the physical fit criteria before the buyer-research team adds contacts.

Keep the source record when importing data into a customer relationship management (CRM) system. Preserve external building and asset IDs in their own fields. Match organizations and contacts through linked records, and retain the reason a building qualifies when accounts merge or an employee changes roles.

For equipment outside a standard building filter, PorchRocket's Asset Search finds and monitors custom classes of assets visible from above and returns the owner of record on each find. Define the search boundary, counting convention and monitoring scope together so the delivery can be reconciled to the same glossary.

Use review outcomes to improve the next request. A wrong object class calls for a clearer asset definition. A correctly found object with poor product fit calls for a tighter selection rule. A supported physical match without a confirmed buyer calls for role research. Preserve those reasons rather than deleting difficult records from the audit history.

Maintain a revision history for definitions as well as data. When the business changes what counts as an assembly or serviceable site, record the effective version and recalculate affected views. Compare suppliers or territories using the same definitions so changes in counting practice remain visible.

Building and asset intelligence release checklist

Use this checklist before a prospect file reaches sales or an estimate reaches a planning meeting. Assign the unresolved work to a person or role and keep the original evidence available. Approval means the file is ready for its stated decision and the remaining research has a clear owner.

Methodology

Anvilfield research uses the 2018 Commercial Buildings Energy Consumption Survey (CBECS) public-use file as its reference stock. The calculation reads 6,436 records, applies FINALWT and retains the public-use SQFT values. Packaged-cooling presence is PKGCL=1; the large-building filter is SQFT>25,000. Blank equipment indicators are not affirmative. The national denominator remains all buildings in the CBECS universe.

National low and high values are approximate 95% sampling bounds: the central estimate plus or minus 1.96 standard errors. The EIA jackknife variance is the sum of squared differences between each of 151 replicate estimates and the full estimate, with factor 1. Ratios are recalculated for every replicate. Counts round to the nearest thousand, mean area to the nearest hundred and shares to a percentage decimal. The public-use floor-area result differs from the unmasked Table B1 total because masking changes published microdata values. The public-use count above 25,000 square feet also differs from summing the rounded published size-band cells; retain the data basis when comparing articles.

Anvilfield territory settings: 1,000 buildings follow the national 2018 packaged-cooling mix. The calculation rounds equipped buildings before downstream multiplication. Low, central and high cases use 2, 3 and 4 units per equipped building and 25, 20 and 15 years per replacement cycle. Account consolidation is 0.75 distinct buying accounts per equipped building in all cases. These are Anvilfield planning settings. Stable stock and staggered retirements define the annual flow; the reference stock receives no growth adjustment to 2026.

For acquisition methodology, PorchRocket's Ground Control publishes commercial building prices of $10 to $20 per building; Anvilfield uses the $15 midpoint for the central list budget. Multiplication follows equipped buildings, with verification, outreach and custom searches scoped separately. The quote is for a building-list input to the model.

The quality worksheet sets 100 returned detections, 90 supported matches and 120 independently established reference assets. Subtraction gives 10 false positives and 30 false negatives; division gives 90% precision and 75% recall. The benchmark settings specify identifiers, dates, roles, evidence states and comparison populations. The worked distributor has four branches; branch count changes assignment of work, with territory totals retained.

The following source register identifies each public input and its year. S1 supplies the weighted quantities; S2 and S3 control their interpretation and uncertainty; S4 cross-checks the survey totals. S5 and S6 support the location and business definitions, S7 supports the quality ratios, and S8 and S9 document the live product facts. All were opened with curl and saved with HTTP status and content hashes on November 5, 2026.

S1. EIA 2018 CBECS public-use microdataS1. EIA 2018 CBECS public-use microdata

S2. EIA 2018 CBECS variable codebookS2. EIA 2018 CBECS variable codebook

S3. EIA CBECS public-use microdata user's guideS3. EIA CBECS public-use microdata user's guide

S4. EIA 2018 CBECS Table B1S4. EIA 2018 CBECS Table B1

S5. Census Bureau: About County Business PatternsS5. Census Bureau: About County Business Patterns

S6. Census Bureau: Census Geocoder documentationS6. Census Bureau: Census Geocoder documentation

S7. Google: Classification, accuracy, recall, precisionS7. Google: Classification, accuracy, recall, precision

S8. PorchRocket Ground ControlS8. PorchRocket Ground Control

S9. PorchRocket Asset SearchS9. PorchRocket Asset Search

Input	Source year	Evidence used
S1. EIA public-use microdata	2018; revised 2022	6,436 records; FINALWT, SQFT, PKGCL and replicate weights
S2. EIA codebook	2018; revised 2022	Variable definitions, equipment responses and area inclusions
S3. EIA microdata guide	2023	151 replicate weights, jackknife factor 1 and public-use masking
S4. EIA Table B1	2018; revised 2022	Published national building and floor-area totals for cross-check
S5. Census County Business Patterns	Undated; accessed 2026	Employer-establishment statistics and business-register context
S6. Census Geocoder documentation	Live documentation; accessed 2026	Address-range interpolation and coordinate interpretation
S7. Google classification documentation	Live documentation; accessed 2026	Precision and recall denominators
S8. Ground Control live offer	2026	Commercial building filters, $10 to $20 building prices and delivery
S9. Asset Search live offer	2026	Custom visible-asset searches, monitoring and owner of record

Next step: Define a commercial building list

FAQ

Is a parcel the same thing as a building?

A parcel is a land-record unit; a building is a physical structure. Keep their identifiers separate and store the relationship. Review complex sites explicitly so changes to a parcel boundary or an operating account do not silently alter the building count.

Does packaged-cooling presence tell me the rooftop unit count?

The EIA presence indicator identifies buildings reporting packaged central cooling. Equipment quantity and rooftop location require additional evidence. Use the building result as a search frame, then establish the assemblies and their attributes from appropriate records and review.

Can gross floor area be used as roof area?

Gross floor area follows included space across floors; roof area follows a roof measurement boundary. Retain both fields and their measurement conventions. A roof takeoff needs its own geometry and treatment of slope, openings and separate roof sections.

Is the owner of record always the buyer?

The owner field identifies the party in the property record. Research the organization and role responsible for the specific work. Keep ownership, operations, access and purchasing responsibility linked so the contact team can qualify the right relationship.

Does a high precision result mean the whole territory was found?

Precision measures the supported share of returned detections. Recall measures the share of reference assets found. Assess coverage against an independently checked reference population and report both denominators. Reviewing only the delivery leaves missed assets unresolved.

What is the difference between confidence and verified fit?

Confidence is a score attached to a classification. Verified fit is a decision supported by the evidence required for the offer. Retain the score, review result and selection rule separately so the commercial team knows what still needs confirmation.

Does an installed-base total measure annual sales?

Installed base describes physical stock. Annual replacement demand describes a flow over a stated period. Connect stock to a replacement model, then investigate age, condition, purchasing responsibility and timing before using it in an opportunity forecast.

What should a glossary contribute to a supplier brief?

It should define the counting unit, spatial boundary, evidence states, date fields, role fields and acceptance method. Use those definitions in the pilot review and final export. Version the glossary when the offer or counting practice changes.

Sources checked

S1. EIA 2018 CBECS public-use microdata Source year: 2018; revised December 2022; opened with curl November 5, 2026; HTTP 200.
S2. EIA 2018 CBECS variable codebook Source year: 2018; revised December 2022; opened with curl November 5, 2026; HTTP 200.
S3. EIA CBECS public-use microdata user's guide Source year: 2023; opened with curl November 5, 2026; HTTP 200.
S4. EIA 2018 CBECS Table B1 Source year: 2018; revised December 2022; opened with curl November 5, 2026; HTTP 200.
S5. Census Bureau: About County Business Patterns Source year: undated; accessed 2026; opened with curl November 5, 2026; HTTP 200.
S6. Census Bureau: Census Geocoder documentation Source year: live documentation; accessed 2026; opened with curl November 5, 2026; HTTP 200.
S7. Google: Classification, accuracy, recall, precision Source year: live documentation; accessed 2026; opened with curl November 5, 2026; HTTP 200.
S8. PorchRocket Ground Control Source year: 2026 live offer; opened with curl November 5, 2026; HTTP 200.
S9. PorchRocket Asset Search Source year: 2026 live offer; opened with curl November 5, 2026; HTTP 200.