Prediction validation
Predicted vs measured RF coverage: a validation workflow
A prediction estimates what the network should do from its inputs; a field measurement records what one device exposed at one place and time. Use both to find the next engineering check—not to expect identical maps.
What each result can tell you
| Prediction | Field measurement |
|---|---|
| A repeatable estimate from site, antenna, power, model and environmental assumptions. | An observation from a route, device, network state and the radio values Android makes available. |
| Useful for choosing sector directions, expected edges and validation points before a visit. | Useful for confirming the route pattern, finding local exceptions and recording the actual serving-cell context. |
The Coverage Planner creates the planning baseline. Network Drive Test records available measurements and route context on Android. They do not automatically synchronize or compare their data.
Why a predicted and measured result differ
First check the input record: site coordinates, active carrier, EIRP, antenna height, azimuth, beamwidth, total tilt and the propagation assumptions. A wrong installation value can create a large, systematic mismatch.
Then separate environmental and live-network effects. Terrain and diffraction can shape a broad shadow; clutter, local obstructions and indoor penetration can change a point result; load, interference and mobility can affect quality even where predicted power is healthy. Finally, Android values are device-dependent: the modem, OEM implementation, Android release, permissions and the registered or unregistered cell information available all affect what can be recorded.
A practical validation route
- Write the baseline down. Keep the calculation date, model, threshold and sector parameters with the map.
- Choose meaningful points. Include a near-site point, boresight, sector edges, a predicted shadow and the expected cell edge.
- Repeat the route where possible. Keep device, band context, time window and test method comparable.
- Compare patterns before isolated values. A consistent offset and a single local obstruction call for different investigations.
- Escalate with evidence. Preserve the map assumptions and the route/device record instead of declaring a model or site fault from one sample.
Parameters worth verifying first
For a broad directional error, verify site position, sector azimuth and tilt. For an unexpected edge, verify frequency, EIRP, antenna height and terrain profile. For good power but poor service, record available RSRP, RSRQ or SINR with the serving-cell and route context, then investigate quality, load and network configuration through the appropriate operator workflow.
Read how terrain, clutter and diffraction differ in a prediction and Android serving and neighbour-cell measurement limitations before treating a mismatch as conclusive.