There is a conversation happening right now about wireless design efficiency. Nobody in that conversation is asking whether the source document is accurate. That is the problem.

The current version of that conversation centers on AI-assisted floor plan processing and automated wall placement. The argument is that AI saves time so engineers can focus on RF design. If the floor plan you feed into that workflow is a poorly scaled PDF, the AI places wall objects from inaccurate geometry faster. Speed does not fix source document accuracy.

Your predictive model is only as accurate as the floor plan underneath it.

I have been in engagements where the post-installation APoS survey came back with coverage gaps that had nothing to do with the RF design. AP placements were right. Channel plan was right. Attenuation values were right. But they were right relative to a floor plan that did not accurately represent the building. The PDF had a scaling error. Every calculation built on top of it was off. Garbage in, garbage out (CWNP, “CWDP Exam Objectives”).

PDF vs. CAD: What You Are Actually Working With

When a floor plan is exported to PDF format, the drawing is converted to a fixed raster grid of pixels. At standard viewing resolution the layout appears accurate and the walls look defined. Zoom in to the precision required for AP placement and the wall edges become indistinct. A wall that should represent six-inch drywall now spans a pixel range that could correspond to four inches or ten inches, depending on where the cursor lands. The survey platform ingests that wall definition and applies an attenuation value against it. The model is running calculations against geometry that was never accurately measured (Adobe Systems, “PDF Reference, Sixth Edition”).

Scaling makes it worse. When you import a PDF, you set the scale manually by selecting two reference points and entering a known distance. If that PDF was printed, scanned, or re-exported, the dimensional accuracy is already gone before you touch it. A one percent error across a 400-foot facility puts AP placements off by four feet. In a high-density deployment that is a design deficiency. Not a rounding error (CWNP, “CWDP Exam Objectives”).

A DWG or DXF file stores building geometry as mathematically defined vector objects: lines, polygons, and dimensional data with precise coordinate values. Every wall carries its actual measured thickness as a property of the object, not as an estimate derived from pixel width. The scale is inherent to the file’s coordinate system and requires no manual calibration from the engineer. When Ekahau AI Pro, Aruba RF Planner, or Hamina imports a DWG file, the platform reads those vector definitions directly, deriving wall thickness and room dimensions from the geometry rather than from pixel pattern analysis. The file can be zoomed to any magnification without degrading, because there are no pixels to degrade (Autodesk, “About DWG Files”; Ekahau, “Importing Floor Plans”).

Manual Wall Dragging Is a Symptom

Manual wall dragging in survey software is not a technique. It is a diagnostic. It means the source document did not provide the AI with the geometric accuracy required to place walls automatically.

Modern wireless survey platforms are engineered to interpolate building structure from accurate source documents without manual intervention. When a DWG file is imported into Ekahau AI Pro, Aruba RF Planner, or Hamina, the platform derives wall objects, material types, and spatial relationships directly from the vector geometry, eliminating the need for manual placement. Manual wall editing exists as a contingency for bad source data. It is a documented workaround, not a design methodology (Aruba Networks, “Aruba Central User Guide”; Hamina, “Getting Started: Floor Plans”).

What to Ask For Before the Project Starts

Ask for the original CAD files before scope is finalized. DWG or DXF, exported from the drawing set. If the customer can only provide a PDF, document that limitation in writing and qualify the predictive accuracy of the deliverable accordingly. Make clear that the design is built on estimated geometry and what that means for the gap between predicted and measured coverage at post-install (CWNP, “CWDP Exam Objectives”).

The tools are getting better. AI wall detection will continue to improve. But the most capable platform available today cannot fix a floor plan that was never accurate. Fix the source document first. Everything in the design depends on it.