The work order comes in at 8:10: a string fault at combiner box CC-17 on a 38-hectare plant, fifty minutes from the depot. The technician arrives at nine, parks next to the transformer station and opens the layout drawing as a PDF on his phone. It is the design drawing, not the as-built: row numbering changed during construction and nobody redrew it. In front of him are three hundred identical tracker rows, with no streets, no house numbers and spring grass up to his knees.
He tries the block the drawing marks as 2B. The first box he finds has a label bleached unreadable by the sun; the second says CC-21. He calls the site manager, who is at another plant and talks him through it from memory: "the one in the aisle facing the north fence, past the inverter with the dented cabinet." At twenty to ten, after forty minutes on foot between rows, he finds the box. Replacing the fuse takes fifteen minutes.
That afternoon the order shows as closed with two and a half hours of labour. Nobody knows how much was repair and how much was searching, and the O&M manager will only see the problem the day someone adds up how many orders in the portfolio look like this one.
Why everything on a plant is in the same place
A ground-mounted solar farm is the extreme case of an asset without an address. Components repeat by the thousand — trackers, combiner boxes, strings, inverters, trenches — they look identical and they are spread over ground covered on foot. The reference a building gives you, floor and room, does not exist here: the only coordinate system the technician understands is the one in the head of whoever built or has maintained the plant from day one.
Three further factors are underestimated by the office. First, the gap between design and as-built: renumbering, relocated boxes, trench runs that do not follow the drawing. Second, the terrain: as Héctor de Lama of UNEF put it, on ground-mounted plants "covering a large extent of land can be laborious" and "you have to deal with ground vegetation", which in spring hides labels and pits. Third, cable theft. In July 2026 the Guardia Civil recovered two tonnes of copper stolen from solar farms in Seville, according to El Debate, after operators reported "repeated cable thefts". Every replacement after a theft changes what is in the field, and if the inventory is not updated, the next search starts from false data.
What searching costs, with example figures
Search time is rarely measured because it disappears inside the order. It is worth running the numbers with your own data. As an example: in a portfolio of twelve plants with 600 corrective jobs a year, if each carries twenty minutes of searching, that is 200 technician hours a year spent finding things. Add the effect on repair time: while the technician searches, the string is still down, and the MTTR reported to management includes that walk without separating it out.
The pressure on that cost has grown. According to UNEF's 2026 annual report, the captured price for solar fell from €42.28/MWh in 2024 to €36.39/MWh in 2025 and €29.68/MWh in the first half of 2026, with "physical and economic curtailment that compromises profitability". With every MWh worth less, there is less room to absorb unproductive O&M hours, and management's first question is no longer what the maintenance contract costs but where its hours go.
What a georeferenced inventory really is
A spreadsheet with two coordinate columns is not a georeferenced inventory. It is a list with numbers attached that almost nobody has checked on site. A georeferenced inventory that serves maintenance has four features.
Real geometry by asset type. A combiner box is a point, a cable trench a line, a block or parcel a polygon. With the right geometry you can ask what lies inside an area or within a hundred metres of a fault, which a list cannot answer.
A unique code, readable in the field and in the system. The same identifier on the physical label, on the map and on the work order. If the label is lost, the map still knows what each item is.
Hierarchy. Plant, transformer station, inverter, box, string. When the order opens on the box, the technician sees what it belongs to and what it contains, and the office spots incomplete assemblies before the technician does.
A visual reference of the terrain. On a plant without streets, the orthophoto is what lets you recognise the aisle, the fence or the internal track. Without it, the map is a cloud of dots on a white background.
Bring the map to the technician
The inventory only saves time if it is in the pocket of whoever is searching, and on many plants that means working without coverage. A map that loads from the internet goes blank in exactly the aisle where it is needed. The field map has to travel pre-downloaded, orthophoto included, and the order has to open from the asset, not from a list where the technician types a code that may not match.
The other half is correction. The person best placed to notice that a box is not where the map says is the technician standing in front of it. With no way to propose the change, the error waits for the next visit. With unrestricted editing, the inventory degrades. The sensible approach is a proposal from the app, approved by someone responsible.
What Maptainer provides is exactly that combination: a CMMS with GIS in which the inventory sits on real geometry, is imported from the as-built in GeoPackage, Shapefile, DXF or CSV, is organised in a composition hierarchy and can be searched by code; and a field app that carries the asset tiles and the national PNOA orthophoto pre-downloaded, so the map renders the same offline, with geometry corrections arriving as requests pending approval.
What to measure before deciding
Before changing tools, find out how much the problem weighs in your own portfolio. Three simple measurements give a clear picture within a month: the time between arriving on site and starting work, on a sample of corrective orders; how many orders need a phone call to someone else to locate the asset; and what share of the inventory has coordinates verified in the field rather than inherited from the design.
With that data, the questions for any vendor are specific: does it import the as-built in the formats the engineering firm delivered, does the map stay fluid with tens of thousands of unfiltered elements, does it work in airplane mode with the orthophoto, and who approves changes coming from the field. Answers checked on a plant similar to yours say more than any feature list.