Most of what I know about inspection records I learned from the buyer's side of the table. For a good part of my career I did due diligence on wind and solar projects and portfolios changing hands. When you are about to put money into a plant, you ask for the O&M history. What comes back is usually a stack of monthly checklists, every line ticked, and a folder of photos named IMG_4471.
Two cases from those data rooms stay with me. A wind farm where the seller disclosed a crack in a turbine foundation. The repair had been done, they said, but there was not a single photo of it and no report from the crew that did the work. And a solar plant where an inverter had caught fire. It had been replaced, they said, and there was no record of the replacement anywhere in the file. In both cases it took weeks of phone calls, and an independent engineer sent to the site, to establish what had been done. The work itself may well have been fine. The record could not prove it.
That is the difference between a completed checklist and a record. The checklist says the round happened. The record says what was found, where, when, on which piece of equipment, under what conditions, and by whom. On a utility-scale plant the gap between the two is where warranty claims stall, where availability reports get questioned, and where a buyer prices in a risk the seller never documented away.
KNOW WHO WILL TEST YOUR INSPECTION REPORT
The record gets tested in three situations, and each one asks something different of it.
Routine preventive maintenance rounds. The O&M contract and the PPA set the interval and the availability reporting. The record has to show that the round covered the assets it was meant to, on the date it was due, and what condition they were in. If the asset manager cannot reconstruct that from what you sent, the round did not happen as far as the contract is concerned.
Warranty and defect claims. When a module shows cell cracking, an inverter trips repeatedly, or a tracker drive fails early, the OEM asks for serial number, position in the array, date first observed, photographs and conditions at the time. What you captured on the day decides whether the claim moves or dies in an email thread.
Handover and re-inspection. Repairs, repowering, a change of O&M provider, a change of owner. Whoever comes next inherits the record or inherits nothing. I have sat on the acquiring side of enough of those handovers to know that a plant with a thin record is a plant that takes weeks of phone calls, and a site visit by an independent engineer, before anyone will put a number on it.
IEC 62446-1 is the reference framework for PV inspection, testing and documentation, and you know it already. This article is about the record itself, which the standard assumes you know how to produce.
GIVE EVERY FINDING SIX THINGS
Would this survive a warranty dispute two years from now, argued by someone who has never been to the site?
Which asset. Not "inverter 12" typed from memory at the end of the row, but the asset confirmed on site. Scan the barcode or QR label on the inverter, the combiner, the tracker controller or the module before you record the finding. The scan ties the finding to that exact unit and proves you were standing in front of it. Two inverters inspected the same afternoon and written up that evening are easy to swap in a report. Typed IDs drift. Scanned IDs do not.
Where. GPS coordinates captured at the moment the photo is taken, not added back at the truck. On a 200 MW site "block 14, somewhere in the south half" is not a location.
When. A timestamp on every note and every photo. The date the observation was made, not the date the report was written.
Conditions. A string current reading means nothing without the irradiance and module temperature it was taken at. A low reading at 300 W/m² under cloud is normal. The same reading at 900 W/m² is a finding. Record the conditions with the data, not in a separate log that never gets matched up.
What. The observation in plain language, the severity, and the reading if there is one. "Discoloration on fuse holder 3, no visible arcing, string 4 current 12% below neighbours" is a finding. "Combiner issue" is not.
Who. The technician who found it, and later the person who closed it. Not for blame. So that someone can ask the right person what they saw when the photo is ambiguous.
SHOOT PHOTOS THAT STAND ON THEIR OWN
Photos get the least thought and carry the most weight. These are the rules I wish every package I ever reviewed had followed.
Wide, then close. One frame that shows where you are: the row, the inverter pad, the block sign. Then the frame that shows the finding. The wide shot lets a reviewer in an office place the close shot on the site.
Get the label in the frame, and scan it too. The scan gets the ID right; the photo makes it proof.
Same angles for the same asset, every time. Modules: face and backsheet. Connectors: both halves of the pair, in focus. Combiners: door open, interior and label in one frame. Inverters: the display with any fault code, then the cabinet interior. Trackers: the drive, the stow position, the row alignment against its neighbours. Vegetation: the plant against the lower edge of the module. When every technician shoots the same asset the same way, month-over-month comparison becomes possible.
Mark the finding on the photo. A circle, an arrow, a measurement. The reviewer should not have to hunt for what you saw.
Pair every thermal frame with a visible-light frame from the same position. A hotspot without the visible image is a blob.
And the two habits that ruin more records than anything else: photos left in the phone gallery with no location, and photos sent over WhatsApp, where the metadata is stripped and the context is gone the moment the chat scrolls.
TURN FINDINGS INTO CORRECTIVE ACTIONS
A finding without an owner is a note. A corrective action has a severity, an assignee, a due date and closure evidence.
Severity first. Three levels are enough. Safety (arc damage, exposed conductors, a tracker that will not stow) is escalated the same day, not folded into the monthly report. Production (underperforming strings, failed bypass diodes, soiling beyond threshold) is scheduled. Cosmetic is tracked for trend.
Assign it to a person, not a company.
Close it with evidence. The "after" photo, taken at the same asset, ideally after scanning the same label so before and after are tied to the same unit. An action closed with "fixed" and no photo is still open as far as the owner is concerned.
For string-level findings, put the string reference on the action rather than tracking every string as its own asset. On a large site that would mean thousands of records for very little benefit.
SEND THE REPORT FROM THE SITE
The weakest document in the chain is the report written on Friday night from memory, with the photos pulled in from three phones.
Details get averaged. Findings get merged. The one combiner that mattered becomes "some combiners showed signs of heat".
The record is strongest when it is assembled from the field data itself, on the day, before the crew leaves. What the asset manager wants is not long: a map of where the findings are, the photos with their coordinates and asset IDs, the open actions with owners and dates, and a summary they can forward without rewriting it.
If you can send that from the site, the inspection is done when you drive out the gate. If you cannot, it is done when somebody finds the time, and on a busy O&M schedule that time is never Friday.
HOW I BUILT THIS INTO FIELDSIGNAL
I built FieldSignal after twenty years of reading field records that could not answer basic questions. The six things above are its spine.
Scan the asset before the finding. The barcode and QR scanner reads the label on the inverter, combiner or tracker controller into the inspection checklist, or records it with the visit. The code appears in the report.
Location and time come with every capture. Notes, photos and voice notes carry GPS coordinates and a timestamp from the moment they are taken. Nothing to add back at the truck.
Current conditions can be recorded with the visit in one tap: temperature, humidity, wind and precipitation. Irradiance and module temperature come from your own instruments; put those readings in the note.
Mark the photo in the field. Circles, arrows and text are in the camera flow, not in a desktop step later.
Start from one of the Solar PV inspection templates, or build your own. Team-owned checklists carry your own procedure.
Findings become corrective actions with a severity, an assignee and a due date. Every action raised on the visit is listed in the report with its status.
The report is generated from the data, on site: a Field Observations report while the visit is under way and a structured Site Visit Report when it ends, as branded PDFs emailed from the app before you leave.
Capture works with no signal. Most of the sites I have worked on have none. Notes, photos, voice, checklists and GPS tracks are captured offline and sync when coverage returns. Generating and emailing the report needs a connection.
FieldSignal is on the App Store and Google Play, with full access for 14 days and no card required. What each plan includes is on the pricing page.
If you inspect solar plants for a living, I would like to know what it gets wrong. That is how it got built.