Wi-Fi site surveys can be expensive; that’s a fact.
Here’s the thing: You have to know exactly what you want and what that means for your business case, especially in the WLAN space. As a field becomes more complex and features grow, the cost grows as a consequence. This in turn has led to a reluctance to spend money for capital expenses across almost every sector of IT. Companies just aren’t throwing money at technology “on spec” like they used to. But let me tell you a secret. We wireless engineers don’t want you spending any more of your precious capex than you do. Yes, this even applies to consultants -- the good ones, at least. At this point we run into the best argument for wireless surveys, besides ensuring basic functionality: your return on investment.
But first, let’s address one of the most common objections to the Wi-Fi site survey: “Don’t bother with the survey; just install more access points!” Let me explain why installing more APs is just not a solution.
First, and most obviously, it will cost you more money.
Second, you need to consider CCI/CCC (co-channel interference/co-channel contention, which I’ll shorten to CCC henceforth). More access points means more channel reuse is needed, which causes CCC and uses more precious airtime, resulting in slower communication. Radios can only “talk” in one direction at a time, to one client device at a time. Every time that a radio “talks,” it has triumphed in a contention algorithm. Any other device that can “hear” the one talking can’t transmit anything until it wins the algorithmic competition for a chance to do so. In properly designed networks, more latency-sensitive devices get the chance to send before less latency-sensitive devices do.
The problem is this: The number of APs you have will scale as high as you can wire and power them (and provide licenses for -- I’m looking at you Cisco/Meraki), but the available channels do not likewise scale. In 2.4 GHz (yes, a lot of clients still lean hard on the lower band) you effectively have just three channels to use in the U.S. You could use channels 2 through 5 and 7 through 10, but doing so makes you even more prone to interference, as any channels outside of 1, 6 and 11 physically overlap the frequency space of others. (The wonderful people at MetaGeek have a great explanation for why channels 1, 6 and 11 are the only non-overlapping choices in 2.4GHz.)
As you add clients, more devices must contend for the transmit opportunity, and it takes longer to transmit data. This effect is magnified if your clients are older and using legacy data rates (transmitting slower, for all purposes). The problem unfortunately gets worse when you have multiple networks close to one another that aren’t part of the same system, such as in an office building with multiple tenants. These can actually end up positioned so they can’t “hear” each other properly and try to “talk” at the same time. This is interference as we understand it classically, which keeps the AP (or client) from properly decoding the message, and results in retries of the transmission, consuming even more precious airtime.
There are multiple other reasons to not just add more APs, but for brevity, let’s move on.
How do these RF concepts affect Wi-Fi site survey ROI? Building on our earlier point, one way to cut down on CCC is to make sure the coverage cells don’t overlap, with techniques like reducing transmit power or using directional APs. This reduces the area in which RF is intelligible and means you can reuse those channels a bit more aggressively. This also means that by making the cell sizes smaller, there’s more capacity available per access point. Capacity is the most common goal in modernity; coverage is the minimum. Here’s where the survey comes in. Wi-Fi nework planning tools like Ekahau Site Survey are invaluable in this domain. Ekahau (and similar tools) allow you to literally map out a number of functions – such as latency, channel overlap and configurable signal strength -- in different surveys. If you simply add APs, you can end up with dead areas or areas with reduced capacity.
RF can exhibit unpredictable behavior, but it’s measurable, and predictive surveys therefore certainly have a limitation. An active or passive survey however, can be as precise as you are willing to be. Is that AP in the hallway of your hotel covering all six rooms the way you intended? (Spoiler alert: No.) Let’s say that you need a certain response time and want to find out exactly where the response time is inadequate. An active survey – which primarily gathers live network-specific data such as latency or throughput -- will do that. If you’re actively remediating an environment and want more control, you can do an APoS (access point on a stick) survey, where you place an AP where you project the need to be, and re-conduct the survey with it in place.
If you’re designing for adequate coverage (because that’s a prerequisite for adequate capacity) you can do a passive survey (which looks at RSSI — received signal strength indicator — and channel structure exclusively) and find out how each area will support your desired data rates. You can also simulate your ideal client in this situation with RSSI offsets.
The point is, simply throwing up APs and expecting machine learning or AI to handle things is a waste of your money. I consider both to be anemic at best, especially without the right seed data, which still requires knowing what in the wide world of RF is going on in your environment to make it work. Do it right the first time. Data is your key. You wouldn’t go into a business transaction knowing nothing about your partners!
Finally, let’s talk about the labor involved in forgoing Wi-Fi site surveys. If you have the philosophy that you can always just add more APs, not only are you going to spend manhours (and likely consulting hours) installing the APs, you are also going to be pouring your even more valuable and likely already committed opex budget into getting the right stuff installed … again. It reinforces the idea that your IT department is a cost center instead of an investment. But that’s another article.