Every reality capture pitch leads with the spec sheet. Point density. Range. Accuracy to the millimeter. And in a parking lot demo, all of it holds up. The problem is that almost no construction worth verifying happens in a parking lot.
It happens in dense cores. Downtown towers wrapped in curtain wall. Mechanical floors packed with conduit and duct. Sites where the thing the scanner quietly depends on, a clean satellite signal, is the first thing the building takes away.
01 / The problemThe signal starts bouncing
GNSS positioning assumes the receiver is talking to satellites in a more or less straight line. In a high-rise environment, it isn't. The signal hits steel, concrete, and glass, then arrives at the receiver as an echo of itself, fractions of a microsecond late. The receiver can't tell the difference between the real signal and the reflection. It trusts both.
That is multipath interference, and the error it introduces isn't a rounding issue. In the worst urban canyons it pushes positioning off by tens of meters. A scanner that promises sub-centimeter accuracy on paper can be wrong by the width of a room, and it will report the bad number with total confidence.
The scanner doesn't know it's lying to you. That's the dangerous part.
This is why the spec sheet is the wrong thing to read. The accuracy figure describes ideal conditions. What you actually need to know is what the system does when conditions fall apart, and that answer lives in the hardware supply chain, not the marketing.
02 / Why integration winsWho controls the chip controls the truth
Most scanner vendors assemble. They buy a positioning module from one company, a sensor from another, write some software on top, and ship it. When the signal degrades, they're at the mercy of a component they didn't design and can't tune.
The exception is the small set of manufacturers that own the stack down to the silicon. Tersus GNSS designs its own GNSS chips. That isn't a vanity detail. It means the multipath filtering happens at the hardware level, before the bad data ever enters the calculation, and it means the positioning can be fused with onboard SLAM so the system keeps a coherent fix even when the satellite signal drops out entirely. The result is a device built to hold sub-3cm accuracy in exactly the GPS-denied environments where assembled scanners fall apart.
- In-house chip design — multipath filtering at the silicon level, not bolted on in software
- RTK-SLAM fusion — satellite correction and onboard motion tracking cross-check each other
- Signal dropout tolerance — the fix survives walking behind a column, into a stairwell, under a slab
- Vertical ownership — one company answers for the whole chain, from antenna to point cloud
This is the part that gets lost when reality capture gets sold as a commodity. Two scanners can print the same number on the spec sheet and behave completely differently the moment the jobsite stops cooperating. The difference is who controls the supply chain.
03 / The part nobody sellsGood data with nowhere to go
Here's the harder truth, and it's the one that pulled me into this in the first place. Even when the capture is perfect, the data usually dies on delivery.
A crew scans a floor. Out comes a beautiful point cloud, tens of millions of points, accurate to the centimeter. And then it gets handed to the client as a LAS or E57 file that almost no one downstream can actually use. It sits on a drive. It doesn't tell anyone whether the work matches the drawings. It doesn't confirm scope. It doesn't back a pay application. It's a perfect photograph of a jobsite that answers no question anyone is being paid to answer.
That gap is the entire reason VIA exists. The hardware front end, done right, produces a trustworthy record of what physically exists in the field. VIA is the intelligence layer that sits downstream of it, turning that point cloud into a structured, auditable answer to the only question that matters at billing time: does what was built match what was drawn, and can we prove it.
Capture without intelligence is storage. Intelligence without trustworthy capture is a guess. You need both, and they have to be built to work as one chain. That's why the supply chain behind the scanner matters more than any single number on its spec sheet. The number is a claim. The chain is whether the claim survives contact with the building.
The reality capture in VIA's pilot program runs on Valhalla Instruments, an authorized Tersus GNSS distributor running the MVP S2 SLAM platform. They handle the front end of the chain. VIA handles what happens after.