Visible-asset data: what aerial and satellite imagery can tell a sales team

What counts as visible-asset data?

A useful record connects a physical observation to a particular building and a business question. A paved area behind a warehouse helps a pavement supplier decide where to research maintenance demand. Distinct cabinets on a retail roof help a heating, ventilation and air conditioning (HVAC) distributor organize a territory review. The evidence gets the site into the right research queue; the next conversation establishes whether the work belongs to your company.

Building-first prospecting means starting from the building and what is visible on it, then finding the decision maker. Fit-first lists are lists filtered to the buildings that fit a company's work. Together, these practices prevent a large contact file from deciding which physical assets your sales team studies.

Installed-base mapping means locating every unit of a product class in the field, any brand. Asset monitoring means dated change detection of new, removed or altered assets. Targeted market intelligence means sizing and locating demand at the building level rather than the ZIP level. Each concept uses the same evidence discipline, but a replacement-market study needs a different counting unit from a service-route plan.

Choose that unit before ordering data. A commercial building, a parcel, an equipment assembly and a purchasing account describe different things. A campus holds several buildings; a shared roof supports equipment serving several tenants. Preserve those relationships so the market total, prospect count and service scope remain understandable when the file changes hands.

What can imagery reveal about a property's physical fit?

Read the image for location, geometry, arrangement and surrounding access. Start with a visible object description, then record its proposed class separately. That lets the reviewer change a tower identification without erasing the original fan-bank observation. For area-based products, trace the relevant surface boundary and exclude unrelated pavement or adjacent roofs before planning a survey.

The commercial question determines the useful detail. A tank-coating manufacturer needs a tank candidate and access context. A generator service company needs equipment candidates grouped into service locations. A parking contractor needs the lot boundary and the scope boundary, which often differ where businesses share circulation lanes. The same picture supports different research files.

Use the technical guides below to write acceptance and rejection examples for each asset class. A tower candidate needs more than a circular fan pattern. A pavement candidate needs more than a dark surface beside a building. Record a reason that another employee can check without repeating the entire territory search.

What does an image leave unresolved?

A visible asset does not establish equipment brand, capacity, operating condition, maintenance history or purchasing timing. Roof discoloration does not establish a coating specification. A tank silhouette does not establish its contents. A generator enclosure does not establish whether the unit operates, who maintains it or when its agreement expires. Those questions need records and an authorized conversation or visit.

The Centers for Disease Control and Prevention identifies small, hidden, underground and shadowed cooling towers as problems for image searches. Side-mounted fans also complicate identification. Its guidance combines imagery with ground searches and records. Apply that principle to sales research: a view without a recognizable asset is different from evidence that the asset is absent. [6]

Keep observation status separate from buying status. A recorded property owner is a route into ownership research, while an operator, tenant, facility manager or portfolio office can control the relevant budget. Log the role actually confirmed and the work it covers. Calling the owner of record the decision maker silently adds a conclusion the image cannot support.

Treat condition-related language carefully in the handoff. Write visible surface staining, obscured equipment yard or cabinet group instead of failed roof, neglected generator or replacement due. The first wording tells the next employee what was observed. The second assigns a technical or commercial conclusion before the supporting work has occurred.

How much image detail does the sales task require?

Ground sample distance, or GSD, describes the nominal ground spacing represented by image samples. Divide an asset's short dimension by GSD to understand how little structure the reviewer has to work with. Sharp edges, shadows, blur and viewing angle still affect recognition. Display zoom changes how large the picture looks; it does not recreate missing observations.

Anvilfield research puts a 3 m wide asset at 10 pixels across in 0.3 m GSD imagery. That supports a different review task from a regional land-cover image.

Anvilfield research puts the same 3 m wide asset at 5 pixels across in 0.6 m GSD imagery. Adjacent components have little separation, so equipment-type decisions need additional evidence.

Anvilfield research puts the same 3 m wide asset at 0.3 of a pixel in 10 m GSD imagery. That resolution does not support an individual cabinet count. Compare the source resolution with the actual equipment-identification task. [4, 5]

Compare aerial and satellite options by actual task suitability, capture dates and usable coverage. The platform label alone does not settle detail. Use broad coverage to organize a region, then require sufficiently detailed views to release individual asset observations. Do not carry a regional land-use classification into a service-equipment inventory.

Anvilfield estimates

Anvilfield research separates the 2018 building-stock estimates, image sampling, review workload and observation delay. Read each row with its unit and reference year; the building groups are nested.

Anvilfield estimates 2,565,000 US commercial buildings using packaged cooling in 2018, with a range of 2,350,000 to 2,779,000 buildings.

Anvilfield estimates 506,000 packaged-cooling buildings above 25,000 square feet in 2018, with a range of 464,000 to 547,000.

Anvilfield estimates 228,000 single-floor buildings in that large-building group in 2018, with a range of 199,000 to 257,000.

Anvilfield estimates 148,000 image-screenable sites in the single-floor group on the 2018 stock basis, with a range of 90,000 to 206,000. Those sites form the equipment-identification and buyer-research queue.

Anvilfield estimates. Scroll horizontally to compare the ranges. See Methodology for inputs and range definitions. How to cite this estimate: Anvilfield, Visible-asset data: what aerial and satellite imagery can tell a sales team, 2026, https://anvilfield.com/field-guides/market-intelligence/visible-asset-data-what-aerial-and-satellite-imagery-can-tell-a-sales-team/

Field benchmarks for an evidence-backed list

Use these as Anvilfield operating rules for commissioning and reviewing a search. They make the acceptance decision repeatable across territories. The sampling gate is a triage rule: meeting it does not establish a classification accuracy, and falling below it sends the case for a better view or another form of evidence.

Keep the benchmark beside the field it governs. A count needs a counting unit, a confidence label needs a specific claim, and a refresh promise needs a capture schedule. A file that passes those checks takes less explanation when a rep, distributor branch or service manager opens it.

What belongs in an evidence-backed sales record?

Give the building and each distinct asset group stable identifiers. Store the site address, coordinates, building boundary and parcel relationship without treating any of them as interchangeable. Keep the original image reference and capture date with the observation. Add the review date and the name or role of the reviewer so an update can be compared with the earlier decision.

Write the observation, proposed classification, confidence and unresolved question in separate fields. A clear cabinet boundary supports a location claim. A probable rooftop-unit classification needs visible characteristics or corroborating records. Condition remains unverified until the appropriate technical assessment. These fields let sales filter physical fit without turning unresolved engineering details into settled facts.

Attach record evidence with its own date and scope. A permit describes the work covered by that permit. An equipment schedule describes its documented inventory. Neither automatically verifies everything now visible on the property. Preserve conflicts for review instead of overwriting an observation with whichever source was opened most recently.

The sales handoff then adds territory assignment, product or service fit, ownership research, confirmed operating responsibility and a next action. Link shared owners across separate sites while retaining each building's evidence. A portfolio relationship helps coordinate outreach; it does not collapse several roofs or equipment yards into the same physical asset record.

How should a team review machine-detected assets?

Begin with a reviewed set of acceptance examples, rejection examples and unresolved cases. Use the actual asset definition, territory and image types in the search. A model that locates rectangular cabinets has not necessarily learned which cabinets belong in a distributor's product category. The reviewer needs the visible clues that distinguish a fit from a look-alike.

Review rejected and unflagged areas as well as delivered candidates. Checking only the positive results helps estimate the share of delivered results that fit, but it leaves missed assets unmeasured. Stratify the check by asset size, shadow, obstruction, building type and image date. Preserve disagreements and the reason the final classification changed.

Measure delivered precision and search completeness with different denominators. Precision asks how many reviewed positive candidates meet the agreed rules. Completeness asks how many known in-scope assets the search recovered in an independently checked area. Neither measure establishes buying intent or service condition. Report the tested geography and evidence standard with the result.

Use unresolved records to assign the next check. A screened yard needs a different view; an ambiguous cooling assembly needs an equipment record or authorized identification visit. Keep those cases in a research queue with the reason attached. Removing every difficult case makes a list look tidy while hiding the territory coverage gap.

How do leading teams turn visible assets into fit-first lists?

For manufacturers, distributors and service companies, the leading operating practice is to measure building fit directly. PorchRocket's Ground Control supplies standard commercial building lists filtered by roof age band, roof material, size, business type, rooftop HVAC, parking areas and other visible features. Use those filters to describe the territory before assigning contact research. [7]

Write the acceptance profile around work the channel can deliver. A distributor needs equipment families its branches support. A coating manufacturer needs roof candidates suitable for an assessment by its contractor network. A service company needs enough nearby sites to justify a route. Agree on exclusions before the list arrives so sales cannot quietly redefine success as every delivered address.

Equipment manufacturers extending installed-base mapping beyond standard building filters use a custom search brief: PorchRocket's Asset Search finds visible asset classes across the lower 48 and returns the owner of record on every find. Define the assembly boundary, look-alikes and unresolved cases in the brief, then verify who operates and purchases for the asset. [8]

Group accepted buildings into territory and channel responsibilities. Give the receiving branch or rep the evidence, fit reason and unresolved question together. Keep a separate stage for an owner conversation and an authorized technical survey. The useful handoff is a research task with a reason to act, rather than an equipment count detached from the building it serves.

Worked example: a regional HVAC distributor screens a territory

A regional HVAC distributor has 1,000 candidate commercial sites in its branch territory. The task is to identify buildings with usable imagery and a physical equipment profile that warrants account research. The deliverable stops at that physical-fit stage. The distributor retains separate queues for ownership research, buying responsibility and an authorized equipment survey.

Anvilfield estimates 66.7 analyst hours for the central 1,000-site review, with a range of 40 to 130 hours. The first pass accounts for 50 hours and second review for 16.7 hours. The research manager assigns both tasks before releasing the branch workload.

Anvilfield estimates 390 evidence-supported physical-fit records from the 1,000-site central review, with a range of 225 to 560. The central queue contains 390 fits, 260 nonfits and 350 sites awaiting usable evidence. Keep each disposition attached to the building and the reason for the review decision.

Anvilfield estimates 17.1 analyst hours per 100 supported records centrally, with a range of 7.1 to 57.8 hours. Include the effort spent on nonfits and unresolved sites when staffing the next batch. Otherwise the branch leaves required review work outside its resource plan.

The branch releases the supported records for ownership and operating-responsibility research. It holds unresolved sites for improved evidence and keeps rejection reasons so an import does not restart the same investigation. Equipment model, capacity and maintenance scope still need confirmation. A buyer conversation then determines whether a survey belongs on the schedule.

The practical management decision is whether to fund the next research stage and which geographic cluster to work first. Image review has narrowed the queue with traceable evidence. It has not produced signed agreements, an observed win rate or a replacement schedule. Keep those outcomes attached to their later stages so the distributor can measure where the work advances.

How does asset monitoring change outreach timing?

Monitoring adds a comparison between dated usable views. Record the last view showing the earlier state and the first view showing the new state. The event belongs within that interval until another record establishes its date. An equipment disappearance, replacement footprint or new tank pad starts a research question about what changed and who is responsible.

Anvilfield estimates a mean wait of 1.5 months from a visible change to the next usable quarterly observation, within a 0 to 3-month observation window.

Anvilfield estimates a mean wait of 6 months with usable annual observations, within a 0 to 12-month window. Set the outreach deadline from the usable evidence date and review completion.

PorchRocket's Asset Search offers one-time searches and scoped monitoring, with the supported cadence confirmed for the area and pricing set per scope. Put image availability, change definitions and unresolved cases into that scope before attaching a sales response deadline. A scheduled review needs a usable view to establish a physical change. [8]

Use a detected change to assign a reasoned next action. A new installation suggests owner and operator research; a removed assembly prompts a check of whether equipment was relocated or replaced. Match the event to available permits and operating records. Do not describe the image date as the commissioning date or treat a new installation as a request for service.

Evidence checklist before releasing a sales record

Require every check below for release as an evidence-supported physical-fit record. An unresolved technical field stays visibly unresolved, with a next verification task. This release decision means the site is ready for further account research; the purchasing stage still needs its own evidence.

Have the receiving rep or branch review the first batch against the actual work profile. Record why sites are rejected during qualification and return those reasons to the research owner. A consistent rejection pattern changes the search definition or required evidence, rather than becoming a complaint detached from the original list.

Methodology

Research notes, November 5, 2026. Stock calculations sum FINALWT in the downloaded 2018 Commercial Buildings Energy Consumption Survey microdata. PKGCL_EL=1 selects packaged cooling, including roof-mounted and side-mounted equipment. SQFTC>=5 selects total floorspace above 25,000 square feet. NFLOOR=1 selects a single floor under the codebook definition, including basement levels. Stock remains at its survey year. [1, 2]

Stock low and high values are approximate 95% intervals using the 151 replicate weights, the published jackknife variance coefficient of 1 and a normal multiplier of 1.96. The image-screenable row multiplies low, central and high stock by usable-coverage factors of 45%, 65% and 80%. That row is a coverage sensitivity envelope. It has no fitted detection-accuracy parameter. [3]

Sampling calculations divide short dimensions of 1.5, 3 and 6 m by GSD values of 0.3, 0.6 and 10 m. The historical technical presentation documents the finer reference scales; the live mission documentation supplies the coarser scale. The sampling calculation adds no recognition probability. The 6-pixel review gate is Anvilfield's operating policy. [4, 5]

The review model uses 1,000 sites, first-pass times of 2, 3 and 5 minutes, recheck shares of 10%, 20% and 35%, and recheck times of 4, 5 and 8 minutes. Supported output multiplies usable coverage by physical-fit shares of 50%, 60% and 70%. These are explicit Anvilfield planning model inputs. Low effort per supported output combines the lowest hours with the highest output; high effort does the reverse. Buyer research, field surveys, travel, outreach, imagery and list purchases sit outside these hours.

For research budgeting, PorchRocket's Ground Control lists commercial buildings at $10 to $20 a building with delivery within 48 hours after payment. Keep that published purchase basis separate from the analyst-hour model; it does not price the subsequent ownership investigation, survey or sale. [7]

Monitoring calculations use quarterly and annual periods of 3 and 12 months, uniformly timed permanent visible changes, and a usable view at every scheduled observation. Mean delay is half the period; endpoints run from immediate observation to the full period. The model adds no capture failures or processing delay. Values are rounded after calculation; source years and every model parameter are preserved with the research output.

Related tools

FAQ

People also ask

• Can you identify a cooling tower from an aerial image?
• Can aerial photos identify rooftop HVAC equipment?
• Which parking lots should I survey first for sealcoat work?
• How do you find aboveground storage tank inspection prospects?
• How much does stormwater pond maintenance cost?

Sources checked

• [1] EIA 2018 CBECS public-use microdata Downloaded CSV with curl; HTTP 200. All 6,436 survey records opened. Calculations use FINALWT and the documented equipment, size and floor variables. No stock-year growth factor.
• [2] EIA 2018 CBECS microdata codebook Downloaded XLSX with curl; HTTP 200. Read the variable definitions for PKGCL_EL, SQFTC, NFLOOR and FINALWT. Packaged cooling includes roof and side locations; floorspace is not roof area.
• [3] EIA 2018 CBECS public-use microdata user guide Revised January 2023. Downloaded PDF with curl; HTTP 200; text extracted and read. Replicate weights FINALWT1 through FINALWT151 use jackknife coefficients of 1. Approximate 95% normal intervals are calculated by Anvilfield.
• [4] USDA: JACIE remote sensing presentation, NAIP specifications September 18, 2018 presentation. Downloaded PDF with curl; HTTP 200; text extracted and read. The 2018 NAIP slide documents 60 cm acquisition outside Oregon and 30 cm in Oregon. Used solely as historical sampling references, without a current-coverage claim or product-supplier attribution.
• [5] Copernicus SentiWiki: S2 mission Live technical reference accessed for the November 5, 2026 edition. Downloaded with curl; HTTP 200; relevant resolution text read. The visible bands have 10 m spatial resolution. Nominal revisit frequency is not used as a property-level usable-image guarantee.
• [6] CDC: Strategies for Identifying Cooling Towers Updated October 19, 2026. Downloaded with curl; HTTP 200; text read. Supports limitations involving hidden, underground, shadowed and side-fan towers and the combination of imagery, ground searches and records. No disease or compliance inference is made about sales prospects.
• [7] PorchRocket: Ground Control commercial lists Live full referral URL downloaded with curl; HTTP 200. Read commercial filters, $10 to $20 per-building pricing and delivery within 48 hours after payment. Product claims are distinct from Anvilfield workflow parameters.
• [8] PorchRocket: Asset Search Live full referral URL downloaded with curl; HTTP 200. Read custom visible-asset finding across the lower 48, owner-of-record output, reviewed results, scoped monitoring and per-scope quotes. No dollar price or universal daily coverage is asserted.