Field-Side QC vs Office-Side QC: Why Catching Errors at the Truck Beats Catching Them in CAD

The traditional model

In most surveying shops, QC works like this. Crew shoots in the field, gets back to the office or to the truck-mounted laptop, downloads the data collector. A senior tech or PLS opens the file, runs the standard checks (Trimble Business Center, TopconLink, MicroSurvey, take your pick), notices that the backsight residuals on Setup 3 are out of spec, and either calls the crew to ask if they remember anything weird about that setup – or schedules a re-visit.

The structural problem isn’t that the office-side checks are wrong. They’re correct and they catch real errors. The problem is that they fire too late. By the time the QC catches the bust, the crew has packed up, driven 30 miles back to the office, and started on the next job. Now fixing the bust requires re-mobilizing the crew, the truck, the equipment – a fully-loaded $670 to $1,200 cost per the re-visit math from last week.

Office-side QC has another structural property: it’s bottlenecked on senior staff. The PLS or senior tech who knows what “the backsight residual on Setup 3 is 0.04 ft when spec is 0.02 ft” actually means is also the most expensive person in the firm. Their hours don’t scale with the firm’s job count.

Field-side QC, when it works, doesn’t replace any of this. It just catches the obvious-in-retrospect busts before they become re-visits.

The field-side menu

The set of QC checks that earn their keep at the truck (or the data collector, or a paired phone) shares a few properties: they fire automatically without crew intervention, they have low false-positive rates, and they’re catchable while the crew can still re-occupy the setup.

1. Rod-height consistency. The single most common bust in survey data is the rod height not getting reset between setups. The crew sets the rod to 5.0 ft for the back-sight from a known elevation, takes the foresight, then doesn’t reset to the actual prism height before shooting the next set of foresights. Result: 30 ft of incorrect elevation across the next dozen points.

A field-side QC check for this: track the rod-height field on every shot. When it changes, flag the change to the crew (a chime or vibration, not a popup that requires acknowledgment – the crew shouldn’t have to context-switch). Most data collectors track rod height; very few make a noise about an inconsistency.

2. Backsight precision against spec. The backsight residual is the difference between the computed backsight position and the known control. If it’s larger than the project spec (typically 0.02 to 0.05 ft for typical work), the setup is bad and every shot from that setup inherits the error.

This is a check that’s often done after-the-fact in office software. It works fine there. But if it’s done at setup time – before the first foresight is taken – the crew can re-occupy the backsight, dust off the prism, level the instrument, and try again. Costs 30 seconds. Catches an entire setup’s worth of bad data.

3. Coordinate-out-of-bounds. The crew is supposed to be working within a project area. If a shot lands 2,000 ft outside the project’s bounding box, that’s almost always a fat-finger error – the crew typed the wrong control-point ID into the data collector and now everything is offset by the difference between the control points. Catchable as soon as a shot is taken.

The check is one line: if the new shot’s distance from the project centroid exceeds 2x the project’s bounding diagonal, flag. Tunable per project; some projects span large areas.

4. Code mismatches. Crew shot 47 fence-line codes (EP FENCE) and 3 manhole codes (MH STORM); two of the manholes were really fence corners but the crew muscle-memoried into the wrong code. The check is statistical: if a code’s spatial pattern doesn’t match the rest of the codes in its class (manholes are typically isolated points; fence codes are typically chained along a line), flag.

This is genuinely a field-side win. By the time the crew is back at the office, they don’t remember which point was a manhole and which was a fence corner. While they’re at the truck, the answer is “oh yeah, point 47 was at the corner of the fence – let me re-code that one.”

5. Z-spike. A point that lands 30 ft above or below its spatial neighbors is almost always a bust – accidentally hitting a rooftop instead of grade, accidentally hitting an antenna, rod-height confusion. The check is a moving local average: for each new shot, compare its elevation to the median elevation of the 5 nearest neighbors within 50 ft. If the delta exceeds a project-specific threshold (typically 5 ft for residential, more for hilly), flag.

We rebuilt this check last month in PointScout Mobile – the previous version had a hardcoded latitude assumption baked into the distance math, which we discussed in the v1.0.2 release notes. The fix was a cosine-corrected tangent-plane projection in the QC engine, which made the elevation-neighbor search work correctly anywhere in the world rather than only near 38° N.

What stays office-side

Three categories of QC that do not work at the truck and shouldn’t be moved:

Cross-job consistency. “This boundary doesn’t match the adjacent boundary we did in 2019.” That’s a comparison query against the firm’s archive, which is hundreds of gigabytes of CAD files. The data collector can’t load that. The phone can’t load that. Office-side review with the right tooling is the only place this can happen.

Deliverable completeness. “Did we capture every easement corner described in the title commitment?” This is a comparison between the field data and a legal document. The field tech doesn’t have the title commitment open. The PLS does, in the office. Always office-side.

Final PLS sign-off. The PLS’s stamp goes on the final survey after they’ve reviewed the deliverable end-to-end. This is a legal requirement, not a QC step we’d consider moving. Always office-side, always by a licensed surveyor, always after the field-side QC has already cleaned up the obvious.

The integration

The right shape: real-time field-side QC catches the predictable busts before the truck leaves the site, where the fix is a minute. The office-side review catches the subtle and does the legal sign-off, where the senior staff time is well-spent on judgment calls and not on chasing rod-height busts.

The economic argument is straightforward. Re-visits at $670 to $1,200 each, an industry-average rate of 8 to 12 percent on jobs, and a meaningful fraction of those caused by busts that field-side QC would have caught. Even cutting half of those re-visits puts $7,000 to $15,000 a year on a 200-job firm’s bottom line.

The deployment problem is harder than the technical problem. Field-side QC has to run on something the field tech actually carries – ideally not “yet another device on the truck.” A phone in their pocket is the right form factor; a paired Android tablet on the dashboard is the second-best option. A laptop in the truck is a no-go (won’t survive the work environment, takes too long to boot when the crew needs to check something fast).

This is the form-factor reasoning behind PointScout Mobile being a free Android app: it’s the most deployable thing we could ship. The customer downloads it, hands the field tech a $20 holster mount, and the QC stack is in service the same afternoon. No procurement cycle, no per-seat license, no IT review.

What to do this week

1. Audit your current QC stack. Make a list of every QC check that runs in the office today. For each, ask: could this be moved to the field side without losing accuracy? Most can. A few – cross-job consistency, deliverable review – can’t.

2. Pick one check to move first. Backsight precision is the easiest win because it has zero false positives, the spec is documented per-project, and the cost of a missed setup is high. Z-spike is second-easiest because the bust is dramatic when it happens.

3. Test for two weeks before deploying broadly. Run the field-side check in parallel with the office-side equivalent. If you see the same flags, the field-side check works. If you see different flags, dig into why.

4. Track the re-visit rate before and after. This is the only number that matters. If field-side QC is real, the re-visit rate should drop by at least 30 percent within the first quarter.