The Pre-Deployment Wireless Stick Test: Why Walking Your Site Before WiFi Deployment Saves $100K and Prevents Failures

How a simple walk-around with a WiFi analyser uncovers real obstacles, interference sources, and coverage gaps that drawings miss, what the stick test actually measures versus what it can’t predict, and why skipping this $2K pre-deployment validation causes $100K+ in redesigns and retrofits after live deployment

👉 Not a member? Free link Here

A generic WiFi deployment is only based on architectural drawings and predictive modelling software. The model predicts -70 dBm coverage everywhere with 10 APs. Looks perfect on screen. Deployment proceeds. Result: dead zones in the shipping area, weak signal in the mezzanine office, unexpected interference from a colocated company upstairs. Redesign required: add 5 more APs ($50K), move 8 existing APs ($30K), investigate interference source ($20K). Total cost: $100K. Root cause: predictive model didn’t account for metal racking that wasn’t in the CAD drawing, structural steel beams that weren’t modelled, and a colocated WiFi network that wasn’t documented. A two-day walk of the site with a WiFi analyser and doing some validation end-to-end tests would have revealed all three issues before deployment.

The stick test is the simplest, cheapest, most valuable step in wireless deployment. It’s a portable RF survey where engineers physically walk a deployment area with WiFi analysers, measuring signal strength, identifying obstacles, detecting interference, testing some temporary AP installations and comparing reality to predictive models. Takes 2–7 days depending on site size. Costs are miserable compared to what it can prevent in post-deployment surprises. Yet most deployments skip it, believing models are accurate or that “we can fix it after go-live.” They can’t. Retrofits cost 10–20x more than pre-deployment fixes.

This article explains what the stick test is, why models can miss critical factors, the importance of field-validating preliminary design, what you actually find during a stick test, how to execute one properly, and why skipping it is the most expensive cost-cutting mistake in wireless deployment.

Predictive Models

Predictive RF models (like Ekahau, Cisco PRTG, Fortinet FortiRF) use sophisticated algorithms to predict signal strength in three dimensions.

Input: building floor plans, material properties, AP specifications, terrain. Output: coverage heat maps showing predicted signal at every location.

What Models Can’t Account For:

Incomplete Drawings: Building CAD rarely shows every wall, metal stud, pipe, beam, and equipment cabinet. The warehouse model shows the main structure but not internal metal racking. The office shows walls but not partition modifications made after the original construction. The model predicts coverage assuming an empty building; the actual building is cluttered.

Unpredictable Multipath Effects: RF signals bounce off surfaces, arriving via multiple paths, interfering constructively or destructively. The model assumes an “average” multipath environment. Actual sites have areas with extreme multipath (signal strengthened by 5+ dB) and areas with destructive multipath (signal weakened 10+ dB) based on surface arrangement. Impossible to predict without measuring.

Material Properties: Standard model assumes “drywall” has uniform attenuation (loss). In reality: drywall with insulation ≠ drywall without insulation ≠ drywall with metal studs. The model assumes 10 dB loss per drywall layer; the actual might be 8 dB or 15 dB depending on materials. Some special materials used in mines, ports and tunnels may be adapted in the modelling tool.

Colocated systems unknown: Model shows YOUR WiFi network. Adjacent companies’ WiFi, Bluetooth, microwave ovens, and neighbouring radios all create interference not in your model. You don’t know who’s deploying what in adjacent spaces.

Typical result: Model predicts -70 dBm coverage everywhere. Actual measurement: -65 dBm in open areas (better than predicted), -85 dBm in metal racking zones (worse), -50 dBm near colocated interference (way worse), interference > -62 dBm ED threshold in specific hallways.

Figure 1: Predicted vs Actual Coverage — Why Models Fail

What the Stick Test Actually Measures

1. Signal Strength (RSSI) at Every Location — Validate your design!

What you measure: Install a temporary AP at the predicted defined location. Walk the site with a WiFi analyser simulating the client (laptop with a WiFi scanner app, or dedicated device). Record signal strength at every location, noting what obstacles are nearby.

Why it matters: Compares actual signal to model prediction. -65 dBm in open area vs model’s -70 dBm (better than expected). -85 dBm in metal racking (much worse). Identifies under-coverage areas needing additional APs or repositioning.

For repositioning, change the position and test again. Adapt the model based on the findings.

What you discover: Dead zones model missed, areas better than predicted, where APs should actually be placed (not where the model said).

2. Obstacles & Attenuation

What you measure: Note every significant obstacle: metal racking, concrete walls, steel beams, water pipes, electrical conduits, large machinery, inventory stacks.

Why it matters: The model assumes generic “warehouse” attenuation. Reality: metal racking is 20–30 dB loss, concrete-with-rebar 25–40 dB, water pipes 5–10 dB depending on size/arrangement. Knowing what’s actually there lets you adjust the design.

What you discover: Which obstacles are RF-blocking (metal) vs RF-transparent (plastic), what material combinations create worst-case attenuation, where APs must avoid being placed (behind metal) and where they must be placed (clear lines of sight).

3. Interference Sources

What you measure: Note any non-WiFi RF sources: Bluetooth devices, colocated WiFi networks (visible as other SSIDs on analyser), cordless phones, medical equipment, manufacturing machinery.

Why it matters: Colocated company’s WiFi on adjacent channel creates baseline -60 to -70 dBm interference.

What you discover: Whether colocated interference will be manageable (spectral mask helps) or severe (same channel, needs redesign). When interference occurs (time-based = microwave, sustained = neighbour network). Whether you need RF filters, different channels, or distance increase.

4. Multipath & Dead Zones

What you measure: Walk slowly through areas, noting signal fluctuations. Some spots signal 10 dB stronger or weaker than neighbours (multipath constructive/destructive interference). Some spots signal completely absent (dead zone behind large metal obstacle).

Why it matters: The model shows an average signal. Reality may have dead zones. Dead zones need additional AP coverage or repositioning. Multipath areas need to know they’re inconsistent (users moving slightly get different signals).

What you discover: Exactly where dead zones are, how large they are, and whether AP repositioning or additional AP solves them. Whether multipath is severe (needs different AP placement) or acceptable (normal RF behaviour).

👉 Deploying WiFi network based on predictive models without walking the site first? My 178-page Industrial Wi-Fi Guide includes the stick test execution frameworks with model-to-reality comparison templates wireless engineers use to validate designs BEFORE deployment — catching 80% of design problems that models miss while they’re still cheap to fix. [Get the guide and validate before you deploy]

What the Stick Test CAN’T Measure (And Why It Doesn’t Matter)

Future interference: Can’t predict what neighbours will deploy next. But can identify WHERE interference might come from (shared walls, common frequency bands) and design defensively (channel separation, enough SNR margin to tolerate reasonable future interference).

Performance under high user load: Stick test measures coverage, not capacity. Doesn’t tell you if APs can handle simultaneous 50 users or only 20. But if coverage is good, AP capacity is usually adequate (coverage-limited sites are better than capacity-limited sites).

Client-specific performance: Different devices roam, hand off, and manage power differently. The stick test shows network infrastructure quality. Client behaviour is separate. Good infrastructure + good clients = good network. Good infrastructure + bad clients = mediocre network. Bad infrastructure = bad network guaranteed.

How to Execute a Proper Stick Test

Timeline: 2–7 days depending on building size (warehouse 3–5 days, office 1–2 days)

Equipment needed:

• APs, tripod, RJ45 cables, RF cables: All needed for a complete 1 AP installation.

• WiFi analyser app (WiFi Scanner on Android, WiFi Explorer on Mac, inSSIDer on Windows)

• Laptop or tablet with WiFi

• Spreadsheet or mapping app to record locations and measurements

• (Optional) Spectrum analyser for interference assessment

Methodology:

1. Walk every area systematically — start in one corner, walk in a grid pattern, record the signal at 10–20-meter intervals, note obstacles near measurement points.

2. Mark problem areas — Anywhere signal <-75 dBm; note what obstacle caused it (metal, concrete, metal racking). Record the exact location for later redesign discussion.

3. Identify interference sources — See other WiFi SSIDs? Note timing, location, and approximate power level. This becomes input to your design (channel selection, RF filter decision, etc).

4. Compare to model predictions — Print model heat map, walk the predicted coverage areas, verify predictions match reality. Note where the model was wrong (usually 10–20 dB off in problem areas).

5. Document findings in report — “Predicted -70 dBm, actual -85 dBm in metal racking area (AP03 vicinity). Recommend repositioning AP or adding AP between metal racks.” This becomes your design change list.

6. Iterate design — Based on findings, adjust AP count/placement/power, select channels avoiding detected interference, identify RF filters if needed. Run a new predictive model WITH real data (obstacles, interference) to validate adjustments.

7. Validate redesign — Optional: Walk site again with new design (AP relocations, additions) verified on map. Confirm new placement will address the identified problems.

👉 Planning wireless deployment and unsure if predictive model matches reality, or struggling to justify $2K pre-deployment stick test to management? I work 1:1 with network engineers to execute proper site validation (walk grid patterns, identify obstacles, detect colocated interference), compare actual measurements to model predictions, calculate exactly how much pre-deployment validation will save in post-deployment redesigns — transforming model-based guesses into data-driven designs that work first time. [Check coaching availability]

Conclusion:

The stick test costs time, but it catches 80% of the design problems that models miss. It lets you fix them during design (cheap) instead of after deployment (expensive). It’s the highest ROI investment in wireless deployment: spend 1–2% of project budget on validation, save 10–20% on redesigns and retrofits.

Best practice: Always run predictive model (cheap, fast, establishes baseline), then always validate with stick test (catches reality). Never skip the stick test, assuming models are accurate. That assumption costs you $50K-$200K. For $5K of insurance, don’t take that bet.