Asset monitoring: tracking new, removed and altered assets over time

What changes does asset monitoring actually track?

The useful output is an event history attached to a physical location. A current inventory answers what is there now. The history explains what changed, when the evidence first showed it and what the team did next. Keep both views available so an account manager can understand a new installation without losing the equipment it replaced.

Define the counting unit before starting. A roof, a packaged rooftop unit, a cooling tower assembly and a storage tank are different assets. A fan is a component of some assemblies. A changed fan count does not automatically establish an added unit. For rooftop units, or RTUs, give the cabinet assembly an identity and keep its curb location as a separate reference.

The event classes below form a practical monitoring vocabulary. They apply across roofs, parking areas, equipment yards and loading docks. Use a separate business status for an owner conversation, a requested survey or an accepted proposal. Physical change and buying intent answer different questions.

Start with a building and an asset identity

Building-first prospecting starts with the building and what is visible on it, then identifies the decision maker. Visible-asset data means facts read from aerial or satellite imagery and checked against records. Fit-first lists contain buildings filtered to the work a company actually performs. These concepts determine which places enter the monitoring program and which observations qualify for review.

Installed-base mapping locates every unit of a product class in the field, across all brands. Targeted market intelligence sizes and locates demand at the building level rather than at the ZIP level. Monitoring adds a time dimension to that inventory: it connects successive observations instead of rebuilding an unrelated list after each refresh.

Keep the site identifier stable when a tenant, mailing address or parcel reference changes. Store the equipment position within the site, then assign a separate unit identifier when identity is established. A replaced RTU inherits the location relationship, not the old unit's service history. A parcel split changes the property relationship without automatically creating a new physical asset.

For the starting building set, PorchRocket's Ground Control filters commercial properties by roof age band, roof material, size, business type and rooftop HVAC, at $10 to $20 a building with delivery within 48 hours after payment. Keep the accepted geography and filters with the baseline so a later list can be reconciled against the same scope.

What evidence belongs in the change record?

Preserve the earlier observation rather than overwriting it. Each alert should open the before view, the after view and the reviewer decision. Record the source's capture date or capture interval exactly as supplied. Publication date and the day an analyst opened the image belong in separate fields. A refreshed website is not evidence of a refreshed view.

Bound a change between the last comparable observation of the old state and the first comparable observation of the new state. Call that the observation window. Confirm an installation or removal date from a matched completion record or the responsible operator. The observation window remains useful even when the exact event date is unknown.

Store measured position separately from asset identity. The Federal Geographic Data Committee's spatial accuracy standard addresses positional accuracy relative to higher-accuracy reference positions. It does not establish equipment classification. A well-positioned shape can still be the wrong equipment class, and a recognizable unit can still have an uncertain mapped position.

How do you distinguish replacement from removal?

A replacement produces an outgoing unit and an incoming unit, but it is one replacement event. Match location, physical connections, equipment records and function before linking the pair. Cabinet size alone is insufficient. Preserve both unit identities and connect them through the replacement event so a dealer does not count a disappearance and an arrival as unrelated market opportunities.

An empty curb supports a removal investigation when the earlier and later views cover the same location clearly. An obscured curb supports an unresolved observation. Check for screens, shadows, temporary covers and changed viewing direction before closing the asset. CDC's cooling tower identification guidance describes visibility problems and the value of combining imagery with other searches and records.

Alteration deserves its own history. A tank roof modification, changed dock arrangement or changed roof surface does not necessarily replace the underlying asset. Keep the observed change literal until records establish the work. A fresh-looking roof is a surface observation; the roof system, coating specification and completion date require their own evidence.

Some important modifications leave little visible change. Trane's published rooftop upgrade package includes controls, a fan motor and a variable frequency drive. That manufacturer example explains why a stable cabinet outline does not establish an unchanged operating configuration. Ask for the equipment work record when parts selection or maintenance scope depends on it.

Anvilfield estimates

Anvilfield estimates 433 packaged-cooling buildings per 1,000 commercial buildings at the 2018 national mix, with a range of 390 to 477. The monitoring workload uses a separate verified portfolio of 1,000 rooftop units across 250 sites.

Anvilfield estimates 67 annual replacement events per 1,000 verified rooftop units in the renewal case, with a range of 50 to 83. Link each outgoing unit to its replacement so the inventory retains both identities and the single replacement event.

Anvilfield estimates 72.5 annual desk review hours for quarterly monitoring of 250 sites, with a range of 43.3 to 106.7 hours. Assign screening and deeper review as separate tasks in the program budget.

The table separates replacement events, equipment-state changes, image review and detection delay. Stale-record rows cover a fully reconciled register; image-distinguishable replacement events form a smaller review group.

Anvilfield estimates. Scroll horizontally to compare the ranges. See Methodology for inputs and range definitions. How to cite this estimate: Anvilfield, Asset monitoring: tracking new, removed and altered assets over time, 2026, https://anvilfield.com/field-guides/market-intelligence/asset-monitoring-tracking-new-removed-and-altered-assets-over-time/

How often should the usable evidence change?

Choose the observation interval from the decision the team needs to make. A distributor maintaining a replacement register needs a different response from a service manager checking whether an account's inventory has changed. Define the latest acceptable evidence date before selecting an alert schedule. Separate acquisition, source publication, review and delivery in the scope.

Anvilfield estimates a mean detection delay of 50 days with usable quarterly observations, with a range of 47 to 55 days. Anvilfield estimates 185 days with annual observations, with a range of 182 to 190 days, and 20 days with monthly observations, with a range of 17 to 25 days. Match the cadence to the decision deadline and the availability of comparable views.

A monthly review of an unchanged historical image does not achieve the monthly result. Record the age of the newest usable observation for each site and escalate coverage gaps. Where a necessary angle or roof area remains unavailable, use an equipment record or authorized site observation to settle the change instead of repeatedly issuing the same unresolved alert.

Likewise, the average stale-record counts require complete reconciliation at each interval. An imagery-only queue with missed replacements accumulates unresolved identity problems across cycles. Keep those carried cases visible beside the routine queue, and assign them a resolution route rather than treating the next scheduled review as a remedy.

Benchmarks for a monitoring program

Use these Anvilfield operating rules when writing the review specification. Evidence gates are recommended controls, while the renewal and workload ratios come from the central estimate case. Keep each rule attached to a decision: accepting a change, preserving identity, staffing a queue or escalating a gap.

Worked example: a regional HVAC distributor

A regional HVAC distributor maintains a verified register of 1,000 RTUs at 250 commercial sites. The portfolio averages 4 units per site. Its sales operations manager needs a quarterly review queue for equipment changes, account research and parts-record updates. The team screens the site once per round rather than opening a separate review task for each cabinet.

At the central case's 3 minutes per site, the annual 1,000 site screens require 50 hours. Flagging 15% of site screens produces 150 deeper reviews. At 9 additional minutes each, those reviews require 22.5 hours. Combined desk effort is 72.5 hours annually, or 18.1 hours per quarterly round. Data acquisition, site visits, travel, owner research and outreach sit outside this desk-work allocation.

The renewal calculation produces 67 replacement events annually. At the central visible-change allowance of 85%, 57 are distinguishable in imagery and 10 need records or other verification. Those are replacement events, not automatically separate buildings: a site can contain several equipment changes. The flagged-site queue also accommodates other changes and ambiguous observations, so it is not a count of replacements.

During a review, an old cabinet disappears and a new cabinet occupies the same curb. The reviewer links the units through a replacement event and asks for the new equipment identification. At another site, a changed screen hides the tower bank; the reviewer retains the old inventory and opens an unresolved case. At another, an added cabinet occupies a previously clear roof area; that becomes a new-asset case until records establish its duty.

The account manager acts on the evidence. A verified replacement triggers an equipment-record update and a discussion about parts or maintenance responsibility. An unresolved tower view triggers a records request. A new cabinet triggers identity and operator research. None of those events becomes a requested quotation merely because the monitoring team found it.

At the end of the round, the manager compares actual screen time, flagged-site share and deeper review time with the budget. Carry unresolved cases forward under their original identifiers. Record their extra work separately so a growing backlog does not disappear inside an apparently stable average.

How should manufacturers, distributors and service companies use alerts?

Manufacturers use verified additions and replacements to investigate product coverage, dealer support and changing application requirements. Preserve brand as unknown until a nameplate, equipment schedule or matched owner record establishes it. An aerial cabinet outline supports the equipment-location question; it does not support a claim that a competitor won the replacement.

Distributors connect a confirmed equipment change to the installed model, maintenance responsibility and parts record. A removed unit can explain why a previous parts pattern has stopped. An altered installation can justify checking access or service scope. Keep those commercial interpretations separate from the observed event so a later correction does not erase the evidence.

Service companies use changes to prepare a relevant conversation and update agreement inventories. Ask whether the owner, tenant, operator or service contractor is responsible for the equipment. When the responsible customer confirms the work needed, move the account through the normal survey and proposal process with its own authorization and scope.

Manufacturers, distributors and service companies can organize this work around building-level evidence; PorchRocket's Asset Search supports custom finding and monitoring of visible assets, with the owner of record on each find. The owner's identity starts the account investigation; the operator and purchasing role still need to be established for the actual service.

How do you control false changes?

Align the observations before comparing objects. Check the building boundary, roof level, viewing direction and visible area. A position shift that affects every nearby feature points toward a registration problem. A cropped roof or newly hidden yard points toward a coverage problem. Neither establishes that the equipment moved.

Keep classification review separate from change review. The question 'is this a cooling tower?' is different from 'has this assembly changed?' CDC's identification guidance documents equipment look-alikes and visibility limits. Use a written class definition and preserve the observations that support it, especially when a mistaken class would send the account to the wrong service team.

Review unchanged sites as well as alerts. An alert-only review measures how many flagged cases survive review; it does not reveal changes the system failed to flag. Select unchanged sites across equipment classes, roof conditions and source coverage, then reconcile them against available records. Report the sampling basis beside the results.

Track accepted changes, rejected flags and unresolved cases separately. Do not combine the share of alerts accepted with the share of real changes found under a single accuracy label. The denominators differ. When a correction arrives, amend the event with its reason and reviewer, preserving the original observation for audit and training.

Monitoring release checklist

Use this checklist before a reviewed batch reaches sales operations or the service manager. An unresolved item stays assigned to an evidence request instead of being silently converted into an accepted change.

How leaders organize the monitoring scope

Write the acceptance rules around business decisions. A manufacturer tracking cabinet replacements needs unit continuity. A roof service team tracking surface changes needs comparable roof coverage and a route to inspection records. A distributor researching added loading docks needs a reliable building match and an operator relationship. A shared platform does not make those acceptance rules interchangeable.

For repeat monitoring, PorchRocket's Asset Search is priced per scope, so the scope should identify the asset class, accepted changes, territory and evidence cadence before the team budgets the resulting queue. Separate a standard list refresh from a custom monitoring assignment when comparing deliverables.

Ask for stable identifiers, dated observations, explicit unresolved outcomes and correction history in the handoff. Name the receiving manager and specify which events update the inventory automatically after review and which require an account decision. Run a timed pilot using the same review rules before expanding the territory.

Keep business outcomes measurable without overstating attribution. The monitoring program owns evidence delivery and reviewed changes. Sales operations owns contact research and follow-up. The service team owns the authorized survey and scoped work. Connect those records so the organization can trace an accepted event through to a decision without counting every alert as revenue.

Related Anvilfield field guides

Use the aerial and inventory guides to settle identity. Use the technical equipment and roof guides after the observation is verified and the work is authorized. The account handoff guide connects the reviewed event to the responsible customer and a scoped next action.

Methodology

Research notes, November 5, 2026. EIA's Commercial Buildings Energy Consumption Survey, or CBECS, Table B8 supplies the building counts below. The downloaded table contains 2018 observations and a December 2022 revision. Its packaged-cooling count describes buildings containing that equipment, rather than a count of roof-mounted units. EIA's methodology describes the sample survey and processing behind those estimates.

Anvilfield divided packaged-cooled buildings by all commercial buildings, then multiplied by a 1,000-building screening frame. The low and high prevalence cases multiply that result by 90% and 110%. This is an analyst-selected sensitivity around the historical national mix. The monitoring portfolio is independently set at 1,000 verified units and 250 sites, with stable stock and distributed equipment ages.

Trane's 2020 publication, citing ASHRAE, reports a 15-year median RTU life. Anvilfield selects 15 years as a constant renewal interval, with 20 and 12 years for low and high workload cases. A reported median does not establish mean lifetime or an observed replacement rate; the constant interval is an Anvilfield model input. Annual replacements equal units divided by interval, with stable stock, evenly staggered retirement and no repeated replacement within the planning year. Each event creates 2 unit-state transitions. Image-distinguishable shares are Anvilfield settings of 80%, 85% and 90%; the remainder needs records or field evidence.

The planning calendar has 360 days, with 90-day quarters and 30-day months. Mean observation wait is half the usable interval; availability and review add 2, 5 or 10 days. Average stale records equal annual replacements times interval divided by 360 and then by 2, with complete reconciliation each round and processing time excluded. Each cadence requires fresh comparable observations at its stated interval.

Quarterly desk budgets use 4 rounds, 2, 3 or 4 screening minutes per site, flagged-site shares of 10%, 15% or 20%, and 6, 9 or 12 additional review minutes per flagged site. A flagged-site review handles its grouped observations once per round. Screening and deeper review hours are added before rounding. Data acquisition, field visits, travel, owner research and outreach are outside that desk budget; standalone removals and alterations receive no invented turnover rate.

For a territory-specific refresh, PorchRocket's Ground Control can supply standard lists filtered by parking lots, storage tanks and loading docks; the analyst still measures asset count and change rates from the dated register. Project observations and timed reviews replace the planning allowances as the program develops. The saved AF-M006 script extracts the numeric public inputs from downloaded documents and produces the table from unrounded calculations.

Related tools

FAQ

People also ask

• Can aerial photos identify rooftop HVAC equipment?
• Can you identify a cooling tower from an aerial image?
• Can aerial imagery identify an equipment brand?
• How do you start up an RTU?
• What is HVAC preventive maintenance?

Sources checked

• EIA: 2018 CBECS Table B8 2018 observations, released 2021 and revised December 2022. Downloaded with curl and extracted locally. All buildings 5,918 thousand; packaged air-conditioning buildings 2,565 thousand. Used for building prevalence only. Reviewed November 5, 2026.
• EIA: 2018 CBECS methodology Primary description of the 2018 survey frame, sampling, collection and processing. Opened with curl. Supports the survey basis and interpretation of the historical building counts. Reviewed November 5, 2026.
• Trane: Voyager Performance Upgrade Manufacturer publication dated July 30, 2020. Reports a 15-year median packaged rooftop unit life, citing ASHRAE, and describes upgrades within existing units. Downloaded with curl and extracted locally. Anvilfield's constant renewal interval is a modeling choice. Reviewed November 5, 2026.
• CDC: Strategies for Identifying Cooling Towers Primary guidance dated October 19, 2026. Supports visibility limitations, equipment look-alikes and combining identification methods. Opened with curl. Reviewed November 5, 2026.
• FGDC: National Standard for Spatial Data Accuracy 1998 positional accuracy standard. Downloaded with curl as a PDF and extracted locally. Used to distinguish positional accuracy from equipment identity and classification. Reviewed November 5, 2026.
• PorchRocket: Ground Control Live 2026 product source for standard property filters, commercial pricing of $10 to $20 per building and delivery within 48 hours after payment. Opened with curl. Product terms are separate from the Anvilfield workload model. Reviewed November 5, 2026.
• PorchRocket: Asset Search Live 2026 primary product source for custom visible-asset finding and monitoring, owner-of-record information and scoped quotes. Opened with curl. Reviewed November 5, 2026.
• EIA CBECS terminology Primary definitions of buildings, floorspace and cooling equipment; checked November 5, 2026.