Average Number Of Cameras Per Project

At a recent trade show the disucssion turned to number of camera statistics. One industry savy individual claimed that 60% of all security projects use 6 cameras or less. Another engineer mentioned that systems with16 cameras or less account for 85% of the overall camera market.

What are the statistics? Are the above estimates correct?


Let's agree that we are only including businesses here, not residential / home use (which obviously would skew numbers down given they frequently have 1 or 2).

Here's our stats from the integrator Winter 2014 survey:

Average then around ~20, which is consistent with what we surveyed 3 years ago. Numbers from SSI are published periodically and I believe their average is 14 to 16, which is lower likely do to their readership based being more intrusion / smaller video dealers.

So those 2 metrics provided to you are very likely to be wrong.

Also, the number is even higher if one looks at the mean rather than the median, as installs with 1000s of cameras would skew the number notably higher.

Let's agree that we are only including businesses here...

Why?

If the discussion is how large is the average system that is installed, specified, maintained, or operated by the average IPVM subscriber, that's probably a valid assumption, but if you're discussing all surveillance video out there, >6 cameras is probably right, and >16 cameras is almost certainly correct.

Do you have any data besides your personal experience?

I've cited IPVM and SSI survey statistics, both that are far higher than your estimate.

Secondly, any meaningful number should segment between home and commercial use, otherwise you are lumping two groups with significantly different usage patterns.

Secondly, any meaningful number should segment between home and commercial use, otherwise you are lumping two groups with significantly different usage patterns.

Good point. In which case, we'd have to know the context of the conversation.

John, IMHO the statistics that you are citing which refute the estimates in the OP are not at all applicable here. Looking at the OP it should be clear that we are talking about an average across projects,

One industry savy individual claimed that 60% of all security projects use 6 cameras or less. Another engineer mentioned that systems with16 cameras or less account for 85% of the overall camera market.

not an average of average integrator deployment size. Using the average of an average loses all weighting based how many actual installs went in to that number to begin with.

For instance you claim 25% of integrators reported deploying an average of 16 cameras or less and that 39% deploy between 17-32 cameras.

But why does that mean that they have the same number of total deployments? Doesn't it follow that those integrators who sell smaller systems might sell more of them than ones who sell larger ones? This is the source of your "skew" in the mean. If you were to simply weight the size of the deployments by the number of deployments that would eliminate this systemic error.

FWIW, I am not saying your statistic is not valuable, just that it can't be reliably applied here. Further I want to make clear I am NOT objecting to the size of the sample itself, we obviously don't know every project from last year and so must extrapolate. But since one dimension is missing from your set, the extrapolation most probably will be highly distorted. Also note that what I am saying does not rely on my knowledge of 'what systems are actually out there', and rather is purely methodical and applies to all data sets without prejudice.

Finally, I do not claim to be an expert in statistics, but this is not graduate level probability either, and I gladly welcome any validation/refutation from anyone who would be so generous as to give one...

I don't consider someone installing a camera in their house to be a 'security project'. I believe most other professionals feel similarly.

If you do want to count home cameras, like nanny cams, dog cams, beach cams, etc. as a 'security project', that's your prerogative. We are simply choosing different populations.

This is not a statistical debate anymore, it's just someone who wants to redefine what the generally accepted notion of a security project is.

John, I am ONLY referring to the deployments which are represented by your graph, shown here:

Answer this ONE question directly, Yes, No or Unknown

Does this graph show that the number of deployments with 16 or less (25%) deployments is less than the number of deployments 32 or more (29% i.e. 15% + 14%)?

Again, I am only referring to your sampling of integrators, not nanny cams

I agree it shows the number of intergrators with an average deployment of 16 or less is fewer, but that is quite differrent, and not properly weighted by number of deployments.

If you want to get an accurate number for overall average deployment size (of your sample), mean OR median, you need to have the number of deployments, not just the number of sizes and the number of integrators. Correct me if I'm wrong but you don't know the number of deployments that went into each integrators average.

Rukmini, you are not going to be convinced and you don't know have enough real world security system experience to have any sense of what is truly being deployed.

You are free to think the average security project has 2, 4 or 6 cameras, however obviously wrong that might be.

I am not spending any more time on abstract debates with you. Any further comment here from you will be deleted.

John, I am an IPVM member who is sincerely and respectfully asking you a simple and reasonable question about what your survey results mean, with an open mind.

I honestly do not know the answer to the question, I assume you do. But if you (or anyone) would kindly tolerate my ignorance a moment more and answer just this one valid yet unanswered question, it would be appreciated.:) Either way I promise to move on without comment, and either way you can make good on your promise to remove this comment. Thanks.

One other big factor is whether one wants a mean or median figure.

Given that camera systems range from 1 to tens of thousands of cameras, the mean is going to be much higher than the median.

For example, let's say we had these set of camera systems: 1, 4, 6, 9, 16, 24, 60, 100, 1000 cameras

The median would be 16 (i.e. 5th of the 9 in the list).

The mean would be 149 ((1+4+6+9+24+60+100+1000)/9).

Now, I don't think it's this skewed, in practice, but surely the mean is integer multiples more than the median.

If I cared about total revenue, I'd be much more interested in the mean than the median. Conversely, if I wanted to know what is most common walking into a random building, I'd take the median over the mean.

Given that camera systems range from 1 to tens of thousands of cameras, the mean is going to be much higher than the median. For example, let's say we had these set of camera systems: 1, 4, 6, 9, 16, 24, 60, 100, 1000 cameras The median would be 16...

Now, I don't think it's this skewed, in practice, but surely the mean is integer multiples more than the median.

John, IMHO, the range of values itself has no effect on whether the mean or median will be higher, as evidenced by the following ideal series

  • [1..9] mean = 5 median = 5
  • [1..999] mean = 500 median = 500
  • [1..99999] mean = 50000 median = 50000

Of course these sets are not representative of camera deployments, for they are skewed with too many high numbers...

But your example suffers the same distortion, although to an even greater degree... Since either intentionally or not, you leave out any duplicates in the lower values. For surely for every single 1000 camera deployment there exists many, many multiples of 1,2,3,4,5,6,7 and 8 camera ones, agree?

It should start out more like 1,1,1,1,1,1,1,1.1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2.....

So, assuming even a rough Gaussian distribution with continuous datapoints, I think the duplicate low values will keep the mean under the median. But I can way suck at math and not realize it sometimes. Is this one of those times? :)

I think you are overthinking this.

The mean will definitely be over the median simply because there is a fixed minimum point in the range, but an infinite max (i.e., a system, by definition, never has less than 1 camera but it can have 10,000, 20,000, 30,000, etc.). It is the rare but super high end camera counts that pushes the mean over the median.

In your example, modified:

1,1,1,1,1,1,1,1.1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2,10000

The median is still 1 but the mean is now over 200.

Same pattern emerges when looking at the mean vs median wealth of individuals. Most people's wealth is low, but there are some mega millionaires and billionaires that skew the mean up compared to the median.

Everything you wanted to know about mean & median, but afraid to ask. Wow guys, it's 6am and my eyes are already glazing over....thanks for the math lesson. ;o)

Let me say that you may be right about the mean being higher than the medium, it certainly can be, (tho I do find it unlikely to be 'multiples' higher) I'm just trying to understand why you think it has to be that way. FYI, to simplify I am assuming a best case scenario where we have perfect knowledge of all camera systems worldwide and their exact integer values, i.e. not 'buckets'.

The mean will definitely be over the median simply because there is a fixed minimum point in the range, but an infinite max...

Firstly, this infinite max term is what as known as a potential infinite, but you are treating it like an actual infinite. Because in any actual sampling there is a max actual value, but what that could have been is irrelevant. And so the actual range is what is used.

1,1,1,1,1,1,1,1.1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2,10000

The median is still 1 but the mean is now over 200.

Nextly, you take my example and turn it pathological. Don't you see my point was that you need to put in all the other numbers (including many, many dupes) between 3 and 10000. That is unless you believe there are no systems with 3-9999 cameras. And depending what numbers you choose you can end up with the median higher or vice versa. Consider this counter-pathological:

1,1,1,1,9,9,9,13,25 Median = 9 Mean = 7.666

So extreme outliers do not necessarily make the mean higher than the medium.

Although there is no denying that enough of them, if large enough will have that effect. The textbook example of this is US average wealth distribution. But in that case, the wealthiest 1% have 35% of the total wealth. Do you believe that the biggest 1% of camera deployments account for 35% of all cameras combined? If that's even close to true then you're probably right, otherwise it depends on the size and number of the outliers, not the fact that they are theoretically unconstrained in max value.

Your example is unrepresentative of real camera deployments:

1,1,1,1,9,9,9,13,25 Median = 9 Mean = 7.666

You can only reach that by assuming a very small max number, close to the median but that's not realistic.

There are quite a number of truly huge camera deployments. And the top 1% of deployments easily are hundreds of cameras each, included in that are ones with thousands and some with tens of thousands of cameras.

Those will inevitably shift the mean far over the median.

Your example is unrepresentative of real camera deployments [1,1,1,1,9,9,9,13,25]

No more than yours though (to which it was a stated pathological counter-example):

[1,1,1,1,1,1,1,1.1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2,10000]

Regardless we now have a statement that can be tested, since you state:

You can only reach that by assuming a very small max number, close to the median...

No, I don't think that's true, but I would rather not waste anymore ink on explaining why.

So instead why don't you tell me your max number and tell me your median and I will be glad to provide the rest of the series so that the median turns out greater than the mean, and if I can't, I'll concede. Just give me two numbers as big and as far apart as you want them... Fair enough?

I know you can create more hypotheticals, but you've already demonstrated that you do not understand what types of systems are actually out there. Any hypothetical series you create would be unrealistic to the distribution of systems deployed.

No more on this.

Do I get 3 credits (Accredited) for Statistics 101 now?

John, I am guessing the respondents of your survey were larger integrators rather than small installers. Based on our experience of working with a very broad spectrum of integrators/installers/dealers the overwhelming number of projects are less than 16 cameras even after netting out consumers. And we don't sell any low-end analog systems.

Rukmini, thanks for reminding me why statistics gave me the hives in college.

Here's the stats from SSI:

Their overall average is 16 cameras (as noted directly above).

These are two broad surveys that both show far higher numbers than the anecdotes being shared.

The respondents to the SSI survey were integrators with an average staff of 54 people. This suggests that there were few small installers in this survey and these are the guys that typically installing the smaller systems. I may be overlooking but I don't see anywhere in the SSI survey that states the number of respondents. They mention that the survey was distributed to thousands of integrators but don't mention how many responded. Strange. Without knowing the sample size it is impossible to determine whether their data is meaningful. If there sample is less than say 200 integrators then I would argue that anecdotal information may be more valid depending on the source.

"If there sample is less than say 200 integrators then I would argue that anecdotal information may be more valid depending on the source."

If there was less than 200 data points in a set, it would be less valid than an anecdote - i.e., a single data point??? :)

Just because the average number of employees were 54, does not mean they serve large businesses. SSI readers tend to be alarm dealers, who may have lots of employees, but tend to serve smaller to medium sized businesses and with smaller camera counts.

Ultimately, there is no way to sample the whole world and there's lots of variance between countries, regions, market segments, etc. so we won't find agreement.

However, I will continue to emphasize the difference between mean and median, which is a fundamental mathematical point and a key practical issue when discussing the 'average' number of cameras deployed.

I don't want to beat this to death but if your single data point was a company like ADI and they were able to provide you data on average number of cameras on order then, yes, this single data point would be more valid than a survey of 200 integrators. And even if they couldn't provide you with hard data but someone within the company with good visibility on sales orders provided general observation information this also might be more valid. This is why I said that it depends on the source.

ADI skews low, due to the nature of people who buy cameras/recorders there. We both know this.

It depends on what your core business is. I have been with a company where the average had to be around 20-40 and with companies where the average was probably between 3-6. When I was a rep, the average was probably between 20-40.

In fact, a main camera manufacturer direct rep once told me that he only gets involved with projects that are 50 or more cameras.

The difference is probably contract work and refurb work versus service work and Small and Medium Business work.

Reading comments and seeing various polls in this site, I expect the majority of people here work with more Small to Medium Businesses. Something you should probably take into consideration is the sites poll figures will be skewed more that direction and may not be the true representative of the national or global average. It is the site average. So if your're a contractor, the numbers may not seem right to you and seem pretty low.

Update - Wow, the IPVM average has gone up since 2011 I beleive. Those #'s seems a lot higher. - Stil I note that forum commenters seem to be for small to medium business.

Here's our 2011 numbers:

That's fairly close to the 2014 ones:

Pull up some old VMS Camera license polls you did. It may have been this pole that I was surprised at the average low amount. It could be one of the polls you did when talking about the PSIM marketing.

We have a bit more than 5000 cameras delpoyed in the field that we monitor, maintain, etc... Job average is almost exactly 10 per, at about 90% IP. With that said I want to sell for the 14% installing 65 or more.

Are there other useful metrics of complexity? Number of users viewing video perhaps? fully stand-alone vs. integrated with existing infrastructure?

Rodney, on the stand-alone vs integrated, see: IP Video Network Deployment Survey Results

As for users viewing, that's an interesting question. We don't have stats for that but will consider it for a future survey.

John,

The manufacuring industry usually produces these figures and uses the "averages" to set price points and sales strategies and drive the manufacturing and technology decisions at the recording side. As a consultant and former sales/sales manager I know the response to the question you pose depends greatly upon two contributing factors: the installer (large integrator vs. local dealer) and the industry served by them. I know of system integrators that market almost exclusively to the USG defense industry who won't even bother to participate in small system projects and other dealers that are only comfortable with small, fast turnaround projects such as light commercial, retail, etc..

Though the total averge number of cameras per recipient might be useful in some way to a professional consultant, I'd like to see more granular informatiomn that added the industry sector addressed. For instance it would be helpful to know the average camera install for rail transportation projects vs. high-rise commercial real estate, vs. USG non-defense, etc. These stats would have more value to both the dealer an the consultant who wishes to target their marketing efforts.

As for averages for my own trade I couldn't even begin to estimate. I led a design/build team that recently installed over 1500 cameras, designed for at least one untra high net worth residence that included the install of 68 cameras, and did a small retrofit for a huge transportation authority that averaged 4 cameras at each of four facilities (added to the 30+ cameras per site existing).

Just food for thought.

Jim