An inspection round finished. Did the team get usable evidence?

Start with the scheduled work

A completed inspection route answers a navigation question. It does not, by itself, tell the receiving team how many of its required readings are available, usable or ready for review. Those are different outcomes. An asset may have been visited while its observation was obstructed; a capture may exist without the context needed to interpret it; an acceptable result may still have no receiving owner.

Before comparing results, agree the denominator: one identified requirement for an asset and reading, within one operating window. A pump temperature reading during the day shift and the same reading during the night shift are separate requirements. Three attempts to obtain the day-shift reading are not three completed requirements. This article proposes a practical accounting method, not a statistical reliability test or a certified inspection procedure.

Report three counts, using the same denominator

Capture coverage is the number of scheduled requirements with a recorded capture, timestamp and evidence reference, divided by all scheduled requirements. Usable-evidence coverage counts the subset that meets an explicitly recorded quality rule. Handoff completeness counts the further subset with a named receiving reviewer and a recorded completed handoff. Keep all three denominators tied to the same scheduled scope.

Consider an illustrative round with 100 scheduled requirements: 92 have captures, 81 meet the agreed quality rules, and 75 have completed handoffs. Report 92%, 81% and 75% respectively. These invented numbers are not RobotAtom deployment results. Reporting 75 out of 81 as the only headline would conceal the original missing work. A zero-requirement schedule is not measured, not a perfect score.

A file is evidence only in context

Boston Dynamics' Spot 5.1.9 data-acquisition documentation describes downloaded capture files and associated metadata grouped by mission or teleoperation session. Position CSVs are optional, and the documentation notes that entries can be empty where the relevant data was not captured. That supports checking the content of an export rather than treating the existence of an archive as proof of completeness. It is a product-specific example, not a RobotAtom integration claim.

ANYbotics describes Data Navigator as providing asset-centred access to inspection data and historical comparisons. This is a manufacturer description, not independent evidence of maintenance savings. Its announcement currently displays a 25 February 2025 page date and a 27 February dateline. Both sources were checked again on 9 September 2026. Neither manufacturer defines the accounting method proposed here.

Preserve what went missing

If a temperature reading is missing, record it as missing. Do not replace it with zero: zero can be a meaningful measurement, while a blank indicates the measurement is unavailable. Record the reason where known, such as an obstructed viewpoint, incomplete transfer or a capture that failed the agreed quality rule. If the cause is uncertain, retain that uncertainty instead of assigning a convenient explanation.

Keep retries beside the original requirement and identify which capture was retained. Do not remove a failed requirement from the denominator after the run to improve the percentage. Scope changes need an explicit explanation and a comparable reporting basis. Use consistent requirement identifiers and operating-window labels; different spellings of the same checkpoint can otherwise appear to be separate work.

Agree what usable means before marking a pass

The responsible site and provider specialists must define the quality rule for each reading. Relevant details might include viewpoint, units, asset identity, operating condition and the reference needed for comparison. A complete checklist cannot establish that a sensor was appropriate or a result was correct. Record the rule and the assessment evidence so another authorised reviewer can understand the decision.

Likewise, assigning a reviewer is not proof that evidence reached that person. Record a handoff reference, such as a receiving queue item or acknowledgement, and keep the next action separate from the quality assessment. An image can satisfy its capture rule while showing a condition that still needs urgent specialist attention. A high completeness percentage is not a statement that the asset is healthy.

Keep the observation when the interpretation changes — added 11 October 2026

A receiving team needs to know which part of a result came from the capture and which part came from a later judgement. A file reference and a conclusion in the same cell make that distinction difficult to recover. Retain the original evidence, record what it shows in a separate observation field, and attribute the interpretation to its author or responsible role. This is an evidence-handling recommendation, not a diagnostic method.

Boston Dynamics' current Spot 5.2.0 Data Acquisition Output documentation distinguishes captured files from metadata and allows metadata to refer to a specific image or to an acquisition action. Its example includes image-file identifiers. The Data Acquisition Overview separately describes action identifiers, capture channels and downloadable groups. Both undated documentation pages were checked on 11 October 2026. These product-specific structures illustrate traceable association; they do not establish that an interpretation is correct or that RobotAtom connects to Spot.

For a hypothetical example, a retained image shows a gauge pointer near a marked division. The observation records the visible pointer position, the units if legible, and the capture reference. A later statement that the equipment is operating normally is an interpretation. That statement needs a stated comparison or method and the relevant operating context. If the scale cannot be read or the task condition is unknown, record that limitation; do not invent a value or turn an unreadable image into a pass.

Record the interpretation's basis, author and timestamp separately from the capture timestamp. The basis might refer to an agreed threshold version, a comparable earlier capture or a qualified review. Listing a reference is not proof that it was appropriate or applied correctly. The responsible specialist decides what the evidence supports and whether additional checks or escalation are needed. A form must not make that decision on their behalf.

Keep unresolved questions and their next owner alongside the conclusion. A reviewer may receive an intact capture while still lacking the load condition needed to assess it. That can be a completed transfer with an unfinished interpretation. Conversely, a well-written interpretation does not prove the underlying file was delivered. Preserve both states so a handoff percentage cannot silently become an asset-health claim.

When testing a handover, ask the receiver to locate the original capture, identify the recorded observation and explain who supplied the interpretation. Include one deliberately unresolved example. A useful acceptance test checks whether the receiver can recover the distinction, rather than whether a summary screen displays a reassuring status. Agree the test and escalation responsibilities with the actual site and provider teams.

The results worksheet now exports these separate fields and flags missing support fields when an interpretation is entered. An empty interpretation remains not recorded. Its checks only assess whether specified entries exist; they do not verify timestamps, inspect files, diagnose equipment or change the capture, usable-evidence and handoff counts. Keep the original evidence in your approved system and download the local record before leaving the page.

Use the worksheet to expose unfinished work

The RobotAtom inspection coverage planner now includes a separate results worksheet beneath the planning brief. Enter every scheduled requirement, including misses, and record the retained capture, quality rule, reviewer and handoff reference. The worksheet withholds percentages when identities are incomplete, duplicate requirement-window rows exist or asserted statuses lack references. It records your inputs; it does not open files, verify timestamps or independently inspect evidence.

Download the results before leaving the page. Entries stay in the current page's memory and are not submitted to RobotAtom. Adopters can use the record to explain unresolved work; providers can use the same scope to distinguish capture, assessment and receiving-team responsibilities. Discuss the actual operating requirements with your delivery team. The worksheet is not permission to operate in a hazardous environment and does not replace qualified engineering, inspection or safety assessment.