okay so this has been driving me nuts for like three weeks. flying a mavic 3 enterprise, running missions through dronedeploy, outputs look totally reasonable visually ? the ortho looks sharp, the DSM looks clean, no weird artifacts or anything. but when i pull it into QGIS and start measuring things i know the real dimensions of, im sometimes off by like 2-3% which doesnt sound like a lot until youre measuring a 200 meter retention pond and suddenly youre off by 4-5 meters.

i dont have a base station setup, im using the onboard GPS which i know isnt RTK. but i thought the absolute accuracy was supposed to be like sub-meter horizontal with good overlap and not terrible conditions? im flying at 100m, 75/75 overlap, clear day, no wind issues.

is this just a no-GCP reality check situation or is there something else going on i should be looking at first before i go buy a emlid or something

yeah this is pretty much the no-GCP reality check situation unfortunately. that 2-3% error is actually pretty typical for consumer GPS without ground control, the absolute position accuracy just isnt there even if the relative accuracy within the model looks good. the model can be internally consistent but shifted or scaled slightly compared to real world coordinates.

before you buy anything though ? do you have access to any known points on site? like a corner of a building with survey coordinates, a manhole with a known elevation, anything like that. even 3 decent GCPs measured with a good GNSS receiver would probably get you down to 2-3cm horizontal which is a different universe from what youre seeing now. the emlid reach is solid for the price if you end up going that route, just make sure youre using a correction source and not just standalone

had the exact same issue last year, thought my workflow was broken for like a month. turns out the coordinate system in QGIS wasnt matching what dd was exporting, was mixing up a projected CRS with geographic and the measurements were garbage because of it. might be worth double checking that before you assume its a GPS accuracy thing, especially if the error seems consistent in one direction