Transparent assumptions

How the RouterReach model works

Predictive Wi‑Fi planning is useful only when its assumptions and limitations are visible. This page documents the model used by the browser tool.

1. Distance model

The planner uses a log-distance path-loss model with a band-specific reference level and indoor path-loss exponent. This creates a realistic planning curve where signal falls as distance increases. It is not calibrated to a specific router or client antenna.

BandPlanning reference at 1 mIndoor exponentFloor penalty per level
2.4 GHz-38 dBm2.69 dB
5 GHz-43 dBm2.814 dB
6 GHz-46 dBm3.018 dB

These are deliberately general planning assumptions. Actual radios, antennas and buildings vary.

2. Barrier attenuation

When the direct path crosses a wall you draw, the model subtracts a material-specific planning loss. Cisco’s Wireless RF Reference Guide gives typical examples of about 3 dB for drywall, 10 dB for brick and 12 dB for concrete, while stressing that metal, moisture, thickness and conductivity change the result. WatchGuard publishes similarly broad deployment estimates and shows that loss often increases at 5 GHz.

Planner material2.4 GHz5 GHz6 GHz*
Drywall / plasterboard3 dB4 dB5 dB
Plain glass3 dB7 dB8 dB
Brick / masonry9 dB15 dB18 dB
Concrete / dense masonry12 dB20 dB25 dB
Metal / foil-backed barrier18 dB28 dB32 dB

*6 GHz values are conservative planning extensions rather than a claim that every material has a fixed measured 6 GHz loss.

3. Use-profile thresholds

The planner grades target-profile coverage using approximate received-signal thresholds: basic browsing -76 dBm, video -70 dBm, work/video calls -67 dBm and gaming/low-latency -65 dBm. These are planning thresholds, not guaranteed service levels. Throughput and latency also depend on interference, channel width, contention, client capability and internet connection quality.

4. Wireless mesh backhaul

For a wireless mesh node, the model separately estimates the node’s connection back to the main router. A weak backhaul applies a penalty to the node’s effective coverage contribution. Wired backhaul removes that penalty. This is why the optimiser avoids suggesting a wireless node deep inside an existing dead zone.

5. What the model does not know

The browser cannot know antenna patterns, router transmit-power settings, neighbouring network interference, furniture, mirrors, pipework, precise wall thickness, moisture, rebar, low-emissivity coatings or the radio capability of every client device. It also does not model multipath reflections in three dimensions. Those omissions are material, so RouterReach is presented as a scenario-comparison tool.

Use an RF survey when the outcome is critical. Business, safety-critical, high-density or unusually complex wireless deployments should be measured with professional survey tools and qualified expertise.

Primary technical references

Last methodology review: 30 August 2026.

Optional measurement calibration

RouterReach can accept RSSI readings that you enter manually at known points on the plan. For each point, the tool calculates the difference between the measured value and the uncalibrated model at that location. It then applies the mean difference as a bounded global correction of no more than ±15 dB. This is intentionally simple: it does not turn the browser into a professional RF survey tool, but it can reduce systematic over- or under-estimation for scenario comparison.

Priority zones and optimisation

Priority zones do not change radio propagation. They change the optimiser's objective function. Samples inside a marked priority zone receive four times the weight of ordinary floor area when candidate mesh positions are scored. This makes the suggested node more likely to solve the room the user actually cares about instead of maximising an undifferentiated whole-floor percentage.

Uploaded floorplans

Uploaded floorplan images are processed locally in the browser and used only as a visual tracing background. They are not uploaded by RouterReach. Scale calibration changes the plan dimensions from a known line distance; accuracy therefore depends on selecting two points whose real separation is known.