What a predictive design is actually built from

A predictive design is a simulation. Somebody imports a floor plan, scales it, traces every wall, assigns an attenuation value to each wall type, places access points of a chosen model at a chosen height, and lets the tool calculate signal across the floor. Out comes a set of heat maps, an access point count and a placement drawing.

It is the right first step on almost every project: it produces the access point count for procurement and the cable drops for the cabling specification. All of it rests on the attenuation values assigned to the walls, and those are estimates. Cisco's RF reference material puts drywall at roughly 3 dB, brick at 10 dB and concrete containing metal at 12 dB, and notes the real figure "depends on metal, moisture, thickness, and conductivity."

The floor plan Scaled and traced. An out-of-date plan puts every later calculation out by the same margin.
Wall and obstruction types Each carrying an assumed loss in dB, and the largest single source of error in the model.
Access point model and antenna pattern A substitution at order time invalidates the placement.
A stated design target Cell edge level, signal-to-noise ratio and a reference client device. A heat map cannot be read without it.

Where the model and the building disagree

None of these is exotic. They show up on ordinary mid-market sites, and each moves measured signal away from predicted signal by more than the margin most models carry.

Construction that is not what the plan says Metal stud framing, foil-backed insulation, plaster over wire lath, lead lining around imaging rooms, coated glass. The plan describes the building as designed.
Ceiling height HPE Aruba Networking's warehouse guidance notes that path loss increases with mounting height, and that signal-to-noise ratio directly beneath the access point typically decreases with it. Above roughly fifteen meters the guidance moves to directional antennas.
Racking and stock The same guidance is blunt: "Shelving acts as a significant barrier to RF propagation." It models dry goods at 3 dB per meter and liquids at 10 dB per meter, so an aisle of bottled product costs more than three times an aisle of empty cartons.
Seasonal occupancy and inventory A distribution center surveyed in March is a different environment in November, and a lecture hall measured in August is empty. Bodies absorb energy at these frequencies, so a report should state what it assumes about full conditions.
Glass, metal and reflection Curtain wall, elevator shafts, mesh partitions and ductwork reflect rather than pass signal, and reflection leaves no clean shadow to model.
Spectrum you do not control Adjacent tenants and non-Wi-Fi equipment in the 2.4 GHz band. Predictive tools assume a clean noise floor.

Capacity and coverage are different problems

Most wireless refreshes are still scoped as coverage projects. Someone asks for signal everywhere, the design delivers signal everywhere, and the network fails on the first day the room is full. Coverage asks where the signal reaches. Capacity asks how many devices can transmit in the same space at the same moment, which depends on available spectrum and cell size.

Cisco's high density design guidance describes the failure mode precisely: "Adding more APs can reduce the number of users per cell and may appear to give more coverage when the space is empty. But once it fills up, the effect will be that of one large super cell covering the room with limited bandwidth and sporadic connections for all." It also gives a baseline: a typical office may cover 2,500 to 5,000 square feet per access point at -67 dBm with twenty to thirty users per cell. A three-hundred-seat lecture hall, a clinic where every clinician carries two devices, and a pick-and-pack aisle with a hundred handhelds are not that space. There the count comes from the device population, and placement and power are worked backwards from it.

Bands, channels and the spectrum you actually have

Channel planning is where a design either creates capacity or spends it. In the United States the 2.4 GHz band offers three non-overlapping 20 MHz channels, which is why 2.4 GHz radios are often disabled on some access points in dense deployments. The 5 GHz band offers as many as twenty channels, and a large share of those are DFS channels that must yield to radar.

Cisco's documentation states that on detecting radar the access point stops transmitting data frames, broadcasts a channel switch announcement and disassociates clients, and that moving to another DFS channel "imposes (at least) a one-minute outage" because the new channel must be scanned for sixty seconds first. Near airports, coastlines and weather radar that is a live risk, and only on-site measurement establishes whether a building sees it.

The 6 GHz band changes the arithmetic. The FCC made 1,200 megahertz available across 5.925 to 7.125 GHz, with indoor low-power operation permitted across the whole band. Those indoor radios run at lower permitted power than their 5 GHz counterparts, so 6 GHz cells are smaller and a 6 GHz layer laid over a placement drawn for 5 GHz will have holes in it. Dense space usually calls for narrower channels as well, per the same high density guidance: "If forced to reuse 5 GHz channels, more consistent results will be delivered using strictly 20 MHz channels."

Roaming is where handhelds and voice break

A laptop that pauses for two seconds while it changes access points is an annoyance nobody reports. A call that drops in a stairwell, or a scanner that loses its session between aisles, generates a ticket every time. Roaming often forces a design denser than a coverage map suggests. Cisco's VoWLAN guidance is still what most designs get measured against: a cell edge of -67 dBm, a 25 dB signal-to-noise ratio, twenty percent cell overlap at 2.4 GHz and fifteen to twenty percent at 5 GHz, and channel utilization under fifty percent.

Three standards help. In Cisco's description, 802.11k lets a client request a neighbor report listing good roam candidates so it does not have to scan the whole band, 802.11v lets the infrastructure suggest a better access point to a client holding on too long, and 802.11r lets the client complete the key exchange with the target before it roams. Clients without support for them fall back to a full reauthentication on every roam, and older handheld scanners and industrial and medical devices are exactly that population. The design is decided by the oldest device that has to work, so the device inventory is settled before the channel and power plan.

What a validation survey catches after the install

A validation survey is a walk of the finished installation with a measurement tool, comparing what was designed against what the building now does. It is the cheapest defect-finding exercise on a wireless project and the one most often cut when a schedule slips. Each of the following is a routine finding on an installation everybody believed was complete.

Access points that are not where the drawing put them Moved a few meters to reach a joist or an existing drop, which changes which aisle has coverage.
Mounting that defeats the antenna Units above a ceiling tile, inside a metal plenum, or wall-mounted where the model assumed a ceiling.
Real attenuation against modeled attenuation Measured signal below prediction across a zone means the wall assumption was wrong, usually for the whole floor.
Interference and neighbor networks Only a spectrum measurement on site finds these.
Switch power capacity Access points that negotiate down to a lower power class lose radios or reduce transmit power. The symptom looks like RF and the cause is the closet.
The places nobody surveyed Stairwells, elevators, loading docks, cold storage and parking structures, where emergency calls get made.

The deliverables to insist on

Ask for these in the statement of work before the purchase order. This is the set we produce on a typical mid-size wireless project, and we agree during scoping which of it a site needs. One deserves naming: the survey project file itself, not only the exported PDF, because that is what lets your team re-run the model later.

The usual sequence is a predictive design before procurement, then a validation survey after installation. Wireless is one layer of the [network design and connectivity work](/networking-connectivity/) we run, and closet power and cabling get decided alongside it, which is why a survey usually sits inside a wider [implementation and migration](/implementation-migration/) scope. Where voice or [collaboration platforms](/collaboration-unified-communications/) ride on the result, those applications set the targets.

An assumptions register The attenuation value used per wall type, the design target, the mounting heights, and the occupancy conditions assumed.
A scaled placement drawing Locations with model, antenna, mount type, height, orientation and cable drop, in a form a cabling partner can install against.
Heat maps per band Signal, signal-to-noise ratio and channel overlap, separately for each band, at the stated threshold.
A channel and power plan Channel width, which DFS channels are in or out, and what happens if radar is detected.
A validation report with a defect list Measured results against the design, each variance accepted in writing or assigned to somebody.
As-built records Final positions, names, addresses, switch ports and the labeling scheme.

Common questions

We already have a predictive design from the manufacturer. Do we still need a survey?

The two answer different questions. The predictive design tells you how many access points to buy and where to run cable. The validation survey tells you whether the building behaves the way the model assumed. On a straightforward office fit-out the gap is often small. In a warehouse, an older building, or anywhere with unusual construction, it is frequently large enough to require additional access points or a revised channel plan.

Can a survey be done before the building is finished?

Not a measurement survey in any meaningful sense, because the walls, the furniture and the stock that shape the signal are not there yet. What can be done at that stage is a predictive design, which is the correct deliverable before construction. The measurement work happens once the space is built out, and it should be scheduled and budgeted at the same time as the design so it does not become the item that gets cut.

How long does a wireless site survey take?

It depends on floor area, how many distinct environments are involved, and how much of the building is accessible during working hours. A single-floor office is a different exercise from a distribution center where every aisle has to be walked. We give a duration once we have seen the floor plans and the device inventory, and anyone quoting one before that is guessing.

Our current wireless works. Why does the new design want more access points?

Usually because the requirement changed rather than the building. Device counts per person keep rising, voice and video have tighter roaming requirements than web traffic, and a design built for coverage produces cells too large for the density now using them. If the existing coverage is genuinely adequate for the new requirement, we will say so. There are refreshes where the honest answer is that the access points are fine and the switching underneath them is the problem.

Do you need to survey every floor and every site?

Not always. Where floors are genuinely identical in construction and use, surveying a representative sample is reasonable, provided the sample is chosen for the hard cases rather than the easy ones. Multi-site estates with a repeatable store or branch format are usually surveyed by format rather than by address. Any site with different construction, a different ceiling or a different device population needs its own look.

Sources

Have the wireless designed before it is bought

Trybus Solutions surveys the building, designs for capacity as well as coverage, and validates the result after the install so you know what you actually have.

Request a wireless assessment Explore Networking and Connectivity

Or call 423-633-1817 · [email protected]

Wireless Site Surveys and What Predictive Design Misses