Every measurement you take hangs off your control. Set it right and everything built on it inherits that quality. Set it carelessly and you have poured a foundation on sand: the numbers look fine until the day they don't, and by then the error is baked into everything you shot off them. A control point earns its keep by being two things at once, accurate the day you set it and still exactly there the day you come back.
A point that moves is worse than no point
The quiet danger of bad control is that it doesn't announce itself. A monument that settles, gets bumped by a loader, or heaves over a winter still gives you a reading. It just gives you the wrong one, and you build on it in good faith. That is why where and how you set a point matters as much as how carefully you measured it in.
Site it where it survives
Pick ground that will not move and a spot that will not get disturbed. Stable, undisturbed soil, not fresh fill, not a stockpile that gets trucked away next week, not a slab about to carry load. Keep it clear of construction traffic, equipment swing, and the path of the work itself. And match the spot to your method: open sky where you are shooting GNSS, clean intervisibility to the points it needs to see where you are running a total station.
Monument it to match the job
The mark should last exactly as long as you need it and be findable that whole time. A nail and washer or a PK in pavement is fine for short-term work. A rebar with a stamped cap suits a boundary or a long project. Primary or permanent control earns a driven rod or a set concrete monument. Whatever you drive, make it recoverable: stamp it, flag it, and give it a home in your notes.
Never trust a lone point
Redundancy is the whole game. Set control in pairs or a small network so that every point is checked by more than one measurement, never hanging off a single shot. That redundancy is also exactly what a least squares adjustment feeds on: with extra observations it can weight every measurement by its quality, distribute error across the whole network rigorously, and flag the one bad shot statistically. Give it a single measurement to a single point and it has nothing to check. Tie the network into the datum your job runs on (state plane, a site datum, existing monuments) so your work lives in the right coordinate frame instead of a private one.
Write it down like someone else has to find it
A control point nobody can recover next month is a one-day point. Record its coordinates, a plain description, a photo, and recovery notes: how to get there, what the mark looks like, what it ties to nearby. The crew that finds your control in six months and trusts it is the crew you are writing for.
The takeaway from the dirt
Control is a promise to your future self and to every trade downstream: build to this, it has not moved. Set it in stable ground, monument it for the life of the job, give it the redundancy an adjustment needs, and document it so it can be found again. And the chain does not end at the monument. Your control is only as trustworthy as the read you take off it when it counts, so the same care that sets a solid point belongs on the point you shoot at the other end.
Shields Precision Solutions builds field-proven survey tools for crews who care about getting the number right the first time. See the lineup at shieldsprecision.com.