You reduce site visits by capturing the site properly once, then letting everyone with a question look at that capture instead of driving out. In practice that means a 360, drone or scan capture that covers more than the question you had that day, organized by location and date, opened in a browser the whole team can use, with a way to measure, leave notes and share findings without a second trip.
What does a site visit actually cost?
Most firms count a site visit as the hours on site. The real cost is wider.
Travel time. A two-hour inspection on a site ninety minutes away is a five-hour day for one person, usually one of your more experienced people. Regional sites turn a half-day check into an overnight. Multiply by everyone who “just needs to have a look” over the life of a job.
Safety exposure. Every trip puts someone on the road and then on an active site. Driving is the part people forget: in the US Bureau of Labor Statistics’ Census of Fatal Occupational Injuries for 2024, transportation incidents were the most frequent fatal work event, at 38.2 percent of the 5,070 fatal work injuries recorded that year (BLS, CFOI 2024 release). Then the site itself: plant, heights, confined spaces, live traffic. A trip you avoid is exposure you avoid.
Access windows. Rail corridors need track possessions, operating plants need shutdowns, occupied buildings need tenant notice, mine sites need inductions. A question that comes up on Tuesday might not get a site answer until the next window, and the decision waits with it.
The question that waits. An engineer wants to confirm a clearance, a quantity surveyor wants to check an area, a client wants to see a defect before approving a variation. If each needs a trip, they pile up until someone bundles them into a visit, and the project moves at the pace of the next one.
Remote inspection does not remove the first trip. It removes most of the second, third and fourth.
What should you capture so nobody has to go back?
The first trip is where you win or lose. Whoever captures the site is capturing for every question that will be asked over the next six months, as well as the one they were sent to answer. Three capture types cover most of that work.
| Capture type | Best for | Capture it so that | What it will not give you |
|---|---|---|---|
| 360 photos | Interiors, plant rooms, roofs, progress walks, condition of finishes | Every room and every face is in at least one 360; shots are geotagged or walked on a GPX track; consecutive shots overlap so the walkthrough connects | Dimensions. A 360 has no scale, so it answers “what does it look like”, not “how wide is it” |
| Drone orthomosaic | Site layout, earthworks progress, stockpile extents, roofs, large external areas, corridors | The same flight plan each time; ground control if anyone will measure from it; enough overlap for a clean mosaic; the DSM exported alongside | Vertical faces, undersides, interiors, anything under a canopy |
| Scan (terrestrial, mobile or XGRIDS) | Structures, clearances, as-built checks, deformation, anything that needs a measured dimension | Scan positions cover every surface someone might measure; registered to a known coordinate system, or at least true scale; 360s at each setup if the scanner supports them | Color detail in dark or reflective areas, and anything behind a lining |
A few rules apply across all three.
Capture more than the question. Another ten 360s or another flight line costs minutes. Discovering in July that nobody photographed the north elevation in March costs a trip.
Name by location, not by photo number. IMG_4471 means nothing in six months. Room, level, chainage, gridline, asset number. Geotagged captures name their own location, which is the main reason to insist on GPS or a GPX track for walked 360s.
Keep the clock honest. Camera clocks drift, and a 360 placed along a GPX track by capture time lands in the wrong place if the camera is eleven minutes out. Sync before you start.
Repeat the same way. Same flight plan, same walk route, same scan positions. Then a difference between captures is a change on site, not a change in method.
What can you do remotely, and what still needs someone on site?
The people who stop trusting remote inspection are the ones who were told it could do everything.
Works well from the office
- Visual condition: cracking, corrosion, staining, water ingress marks, missing elements, vegetation
- Progress against program: what is built, what is not, what has changed since the last capture
- Dimensions and areas from a scan or a controlled orthomosaic: clearances, setbacks, floor areas, lengths for quantities
- Pre-tender and pre-start familiarization, so the first physical visit is spent on things that need hands
- Client review and sign-off against the evidence
- Dilapidation reference (pre-construction condition survey, in US terms): checking a claimed damage location against the earlier capture
Needs someone on site
- Anything you have to touch or operate: moisture readings, hammer tests, valve checks, bolt tension
- Anything hidden: behind linings, below ground, inside ducts, under insulation
- Anything you have to smell, hear or feel: gas, bearing noise, vibration, damp
- Instrument readings that were not captured: thermal, levels on control marks, settlement monitoring
- Statutory and contractual inspections where a qualified person must attend
- Anything on a site that has changed since the last capture. At that point the capture only shows how it used to be
Two cautions. A dimension is only as good as its frame: a distance off a controlled drone survey or a registered scan is a measurement, a distance off a 360 photo or a Gaussian splat is a rough size, and any tool worth using will say which you are looking at. And a capture tells you what is there, not why. That still takes a conversation with the site supervisor.
Most questions after the first visit are visual, dimensional or about change, and can be answered from the office. The ones that need hands get batched into one planned trip instead of four unplanned ones.
How do you review a site remotely?
Remote review needs a routine. Here is the one we see holding up across building health checks, site inspections, dilapidation reports and quantity surveying.
-
Organize by site, then by capture date. One place per physical site, every capture under it in date order. “What did this look like before” becomes a scroll.
-
Open everything in a browser. Everyone has a browser. Few have CloudCompare or a GIS. If the point cloud only opens on the one licensed workstation, one person does the review and everyone else asks for a trip.
-
Measure where the frame allows. Take distances, areas and heights off the scan or the orthomosaic and keep the ones that matter. Do not let an indicative number end up in a quantity.
-
Pin findings to the spot. “Crack at the base of column C4, see photo 12” is one email away from being lost. A note pinned to that spot on the scan, with a sketch over the captured view showing what you mean, is findable by anyone who opens the site later. A color and an icon per note keep a defect, a question and a sign-off apart at a glance.
-
Group findings into lists. A snag list, an inspection round, and a list of items for the next physical visit. The last one is the point: anything that cannot be answered remotely goes on the trip list, and one trip covers all of it.
-
Share one link and let people reply where the note is. The client opens the link, sees the capture, reads the notes and replies on them. In our experience this is where clients usually approve: on the link, against the evidence, without a meeting.
-
Recapture, and let the notes carry over. On an orthomosaic or georeferenced scan, the open notes should still be there on the next capture, at the same spots, labeled with when they were made. Resolved items close, new ones get pinned, and the site record builds instead of resetting.
There is a longer-term reason too. Site knowledge lives in the heads of the people who have been there, and it leaves when they retire. Tying findings to the location, where they outlast any one report, keeps that knowledge with the site. That is the site record idea.
Where Swyvl fits
Swyvl is built around this routine. Every upload belongs to a site (a physical place on a map), and captures sit under each site in date order. Point clouds, orthomosaics, 3D Tiles, XGRIDS scans, 360 photos, photos, video, DWG, IFC and PDFs open in the browser with nothing to install.
For 360 captures, geotagged shots land on the site map by GPS, and a GPX track places untagged shots along the route by capture time. Connected 360s become a street-view-style walkthrough with a synced minimap, and E57 scans with embedded 360s open each panorama beside the point cloud. One survey firm we work with puts thousands of 360s a month on a map by job number, next to the job’s DWGs, with one link per job.
Measurements (point, distance, area and height) work in every 3D and map viewer, with unit and coordinate system switching, and every result carries a badge: georeferenced, true scale or indicative. They do not work on 360 photos or plain photos (no scale to measure against), and there is no volume or cut-and-fill calculation.
Notes pin to an exact spot, trace an area, mark a volume shape (a traced base plus a height you type; it does not compute cubic volume), or carry a sketch drawn over the captured view. They take a color and an icon, attachments, @mentions and an assignee, and they work on 360 photos, photos and video as well as scans and orthomosaics. Notes follow the place, not the file: a note made on the March capture shows on the June capture at the same spot, as a hollow pin labeled with where and when it was made, as long as the files have a position. Notes group into lists that can span files, export as a PDF, or go out as a walkthrough link that steps through each one.
Share links are one link per job or site, branded, password or invite-only by email, and editable after sending. People invited by email verify with a code and reply to notes in the browser, with no account to create and never a paid seat. Clients can approve a delivery or request changes on the link, and you can see when it has been viewed. An activity log records uploads, shares and views. The free plan (5 GB, 3 sites, 5 share links, 10 saved notes and measurements) is enough to run the routine on one job.
For the same model on linear assets and multi-year records, see how infrastructure operators document inspections. For property work, see site records for inspectors and developers.