Camera ownership on Flickr: 2013-2014

[lintoni]

The interesting numbers are growth from 2013 -> 2014:
Nikon +0.8%
Canon +0.4%
Sony +0.7%
Samsung +3.2% (mostly phones?)

In terms of %'s, this growth is:
Nikon: +6.8%
Canon: +3%
Sony: +20%
Samsung: +133%

So you might say that roughly twice as many people will buy a new Nikon or Sony camera as they will Canon.

If I was a Canon exec, that's something that would concern me.

You really don't have a grasp of how numbers work, do you?

As there are no absolute numbers given for the size of the sample from which the ratios are derived, to talk of growth is totally erroneous - the sample size could have diminished by half, or conversely it could have doubled.

Further, the numbers show that Canon has strengthened its position as the most popular camera of Flickr's users.

Now, I'm sure you're not really foolish enough to believe that these numbers are causing any concern, whatsoever, to Canon's executives, so why do you waste your time making such asinine posts?

 

[old-pr-pix]

The interesting numbers are growth from 2013 -> 2014:
Nikon +0.8%
Canon +0.4%
Sony +0.7%
Samsung +3.2% (mostly phones?)

In terms of %'s, this growth is:
Nikon: +6.8%
Canon: +3%
Sony: +20%
Samsung: +133%

So you might say that roughly twice as many people will buy a new Nikon or Sony camera as they will Canon.

If I was a Canon exec, that's something that would concern me.

What? How do you figure that? These are essentially "usage" measures of Flicker aligned contributors. It says nothing about sales of new models. How do you think the data would look if all the pro sports shooters and all the photojournalists put all their images on Flicker? And then, there is the question of time in market. Older models naturally have more hits. For example, where is 7D II on the list? Is that because it isn't selling well? I don't think so.

 

[lintoni]

Absurd.

 

[Don Haines]

Wow!

So much wrong on the survey! No idea what the methodology is.... Is it counting by pictures or by owners, what about multiples, why is only 42.1 % of the cameras in the " big 5 ", how many were undetermined, what is the margin of error, what is the sample size, what is the accuracy?

You can't say anything significant from the results, so why bother reading conclusions from it?

 

[Roo]

Wow!

So much wrong on the survey! No idea what the methodology is.... Is it counting by pictures or by owners, what about multiples, why is only 42.1 % of the cameras in the " big 5 ", how many were undetermined, what is the margin of error, what is the sample size, what is the accuracy?

You can't say anything significant from the results, so why bother reading conclusions from it?

Agreed. Any conclusions are made on some rather large assumptions.

 

[Orangutan]

Wow!

So much wrong on the survey! No idea what the methodology is.... Is it counting by pictures or by owners, what about multiples, why is only 42.1 % of the cameras in the " big 5 ", how many were undetermined, what is the margin of error, what is the sample size, what is the accuracy?

You can't say anything significant from the results, so why bother reading conclusions from it?

Agreed. Any conclusions are made on some rather large assumptions.

When people post Amazon listings showing Canon camera sales rankings, does anyone question that? No, everyone just cares about what the ranking numbers are when in fact the same questions could easily be asked of Amazon's reporting.

Dilbert, since you seem to be sincere on this point I'll try to explain why the data from Flickr are not definitive.

First, the good: they're careful to explain what they counted: "We estimate camera ownership per-week by only counting a camera once per-account, per-week, to compensate for community members uploading different quantities of photos." OK, that much is good

Because they use metadata from images, the method has some problems:

• They may have more than one camera, e.g. DSLR plus iPhone. We have no idea in what order their cameras were purchased, so we can't infer their "upgrade path."
• The photos posted to Flickr may be different from the photos posted on their portfolio site; e.g. SmugMug or Zenfolio, e.g. casual vs. serious photography.
• We don't know how may of the counted Flickr accounts are paid vs. free,which would likely correlate the kind of camera used.
• We know nothing of how well Flickr users represent the broader "market" of serious photographers: are they the serious ones, or is Flickr just a step up from FaceBook? Certainly there are some great photos on Flickr, but are they the rule or the exception?

In short, the data are "accurate" for Flickr users, but we have almost zero correlation to the greater market due to the confounding factors. I.e., we can't use these data to tell what kinds of cameras people want.



Regarding Amazon: Amazon's sales rankings are hourly (that's a problem); however, when multiple snapshots are taken over time, it can provide a reasonable pictures of Amazon's overall sales, though not perfect.

• Though Amazon is not the only online retailer, they're huge.
• Amazon's market tends to be the general online-shopping community: there's no reason to believe they're a niche camera retailer, so their sales numbers are likely a good approximation of the purchase habits of the online-purchasing market.
• There are probably a good number of people who like to buy local rather than online; I have no idea what fraction they comprise, but I know that a lot of local camera stores have closed over the last decade.
• Amazon is not a true global seller, so is best a representative of the U.S., and other markets only to a lesser extent.
• Amazon's rankings represents sales of cameras, period. No inference of camera from image.

In short, Flickr can't be used at all to infer what cameras people want and will pay for. Amazon data, though far from perfect, goes some distance to that conclusion.
I hope this helps.

 

[DominoDude]

Fully agree with a bunch of you who think before you type.
If the statistics are derived from submitted EXIF information, then it could be skewed because some photographers don't publish their EXIF. If it's based on tags on images, then it can be just as biased.

Plenty of smartphones comes with preinstalled apps that makes it very easy to synchronize the images taken on the phone up to a Flickr-account. In some instances that even takes place on all images taken on the phone, whether you publish them or not. How are those photos counted? With an ordinary dSLR you have to put in some manual labour before your shots gets uploaded and visible to the general public.

 

[DominoDude]

I don't think we can draw conclusions about sales numbers or the number of owned photographic devices. Numbers collected are more likely coupled to usage and how that's distributed among the Flickr accounts.

I leave it to lobbyists, politicians and propagandists to wrongfully use statistics. To collect data with one purpose, but trying to apply it to another scenario is more than prone to errors.

 

[Orangutan]

but if the sample set size from flickr is large enough (for example their entire userbase)

I think this is the problem: for the sample to have any validity at all it needs to be a representative sample. The point I'm making is that, regardless of size, there's insufficient information in Flickr's data collection to infer beyond the Flickr user base.

So whilst all of the above issues you mention may be in flickr numbers, there is no way to account for them specifically and similarly there is no reason to expect that they would favour or disadvantage

More importantly, though, there's no reason to believe they wouldn't. In order for a statistical sample to be valid, you have to be demonstrably confident that it isn't skewed. While the Flickr survey is interesting for what it is, it simply doesn't contain enough data, and there are too many confounding factors to allow me to exclude the risk of bias to use it to infer camera ownership and/or recent purchase.

 

[Roo]

I imagine its easy to skew the numbers on flickr. I shot with 2 different 7Dii cameras at a tennis tournament the other day. Both were supplied to me by CPS while I shot 2 sets for the day. I've posted a few images on flickr while others that were there could post their images from the same cameras and how would that be reported by flickr? My conclusion is that there are no useful conclusions that can be drawn about camera ownership or purchasing from the data. Amazon data is different matter.

 

[old-pr-pix]

Seems like most of us are convinced the data is interesting, but not a forecast of future buying trends. For those still unconvinced, consider this:
1) Top 5 manufacturers only total 35% (2013) and 42% (2014) and only 3 of the 5 are "camera" companies. Who are the other 60-65%? Must be a whole lot of companies (probably a lot of smartphones?) since they each have to have less than ~2% share (2013). Or, is there 50% of data that the EXIF is missing or can't tell manufacturer? Maybe there are a significant number of scanned film images? That would be a huge bias in the data.
2) This data is still a reflection of how often people post pictures using cameras they bought sometime in the past. Recent models are overwhelmed by the field population (and usage) of long-standing camera models. It doesn't indicate buying preferences for the future. Go back several years in the Flicker data -- could it be used to forecast the onslaught of smartphones when Apple and Samsung were not even reported in the data? Only if you look over several years data can you see the smartphone invasion coming.

At best, a casual look at the data over only two years shows changes in the posting habits of the existing population of cameras/users. Only very sophisticated analysis over several years looking deep into the data and how it is collected could anyone draw any conclusions about future sales trends. And that would require an assumption that technology wouldn't change. Any new innovations would potentially destroy the analysis.

 

[Don Haines]

Survey taking is both an art and a science. It is very hard to frame questions properly to avoid bias, and it is very hard to mine an existing pile of randomly collected data for trends.

To have any significance, this survey would have to somehow only look at pictures taken from cameras bought in 2014.... and at that, it would only have significance as a representation of flicker users.... not the general public.

I have no way of verifying this, but my suspicion is that a disproportionately large number of advanced users will both be shooting with a DSLR and that those users will be editing with software like photoshop or lightroom and stripping out the EXIF data.... this would result in an under reporting of "fancy" cameras and a large "unknown" group. This factor alone destroys the accuracy of this survey.

 

[rfdesigner]

Seriously! you guys are still on this.

There is only One rule, One truth : Statistical Analysis is for the [size=14pt]Ignorant.[/size]

Not true at all.

It simply matters what you're measuring and how.

I've used statistical analysis for volume electronics production test. No problems with stats there, a very handy predictive tool.

The problems come in (mis)using stats when measure people as they are highly complicated and there must always be some kind of simplificaion somewhere which degrades reliability.

Don't blame the stats, it's not their fault they're misused.

 

[Zv]

Flickr is a graveyard -- has been for some good time now.

That really isn't true you know. It achieves that balance between being centred on the photographs while retaining a very important social element better than any other site I know. There are other photo gallery sites, some of them visually very good, but socially they are pretty sterile. There are obviously sites which major on the social side but the photographs come a poor second.

I don't have any actual statistics, but anecdotally I'm seeing people who moved away from Flickr drifting back.

Yup, I'm one of them who drifted back. Tried 500px but decided Flickr was good enough for a place to store pics in decent quality incase someone wants to see my work. All I needed to do was clean it out a bit. The social aspect of it has been dead for a long time though, haven't noticed it getting any better.

 

[Orangutan]

More importantly, though, there's no reason to believe they wouldn't.

Oh come on and quit with the conspiracy garbage will you?

I can tell you're sincere, Dilbert, but apparently you've never taken a class in statistics, even a basic one. The word "bias" in this case is not an accusation against Flickr, it's a potential flaw in the data. I'm not saying at all that Flickr may be biased, I'm saying the data may be biased in a way that prevents us from drawing certain conclusions. Again, this is not a conspiratorial assertion, it's a statistical assertion. I'll let you read the Wikipedia article on it.

And let me pose some questions for you...
- how would flickr introduce bias? I'm not saying they're introducing it, I'm saying they don't have enough data to eliminate pre-existing bias in their data.
- what does flickr have to benefit from bias? Again, I'm not saying it's intentional, I'm saying it's unavoidable due to the data they have.
- what outcome would flickr see as a result of it being biased? Actually, I don't think they were trying to be biased: they were very clear about what they were trying to do and how they did it. I simply believe there is insufficient data to go beyond what they've done to draw any conclusions about what future cameras purchases people are likely to make.

Answers to the above are in-line in blue.

In summary, Dilbert, Flickr's data can lead to a valid conclusion summarized as: "among all Flickr account holders who keep metadata in their images, the prevalence of cameras is given in the chart below..."

What they did NOT say, and which would NOT be true, is that they can extrapolate that to the entire camera-buying community to infer what kinds of cameras people would like to buy.

 

[lintoni]

More importantly, though, there's no reason to believe they wouldn't.

Oh come on and quit with the conspiracy garbage will you?

I can tell you're sincere, Dilbert, but apparently you've never taken a class in statistics, even a basic one. The word "bias" in this case is not an accusation against Flickr, it's a potential flaw in the data. I'm not saying at all that Flickr may be biased, I'm saying the data may be biased in a way that prevents us from drawing certain conclusions. Again, this is not a conspiratorial assertion, it's a statistical assertion. I'll let you read the Wikipedia article on it.
...
In summary, Dilbert, Flickr's data can lead to a valid conclusion summarized as: "among all Flickr account holders who keep metadata in their images, the prevalence of cameras is given in the chart below..."

What they did NOT say, and which would NOT be true, is that they can extrapolate that to the entire camera-buying community to infer what kinds of cameras people would like to buy.

Sure, you're concerned that there could be bias in the source of the data (for example, people expunging EXIF data.)

My assertion is that the expunging of EXIF data is not specific to a particular manufacturer and that any inaccuracies in data equally effects all vendors thus the percentages and changes in them are representative of the actual market.

Which is an entirely unwarranted assertion. Personally, I'd hazard a guess that the average smartphone photographer doesn't even know what EXIF data is, let alone how to strip it from their photographs.

 

[Orangutan]

More importantly, though, there's no reason to believe they wouldn't.

Oh come on and quit with the conspiracy garbage will you?

I can tell you're sincere, Dilbert, but apparently you've never taken a class in statistics, even a basic one. The word "bias" in this case is not an accusation against Flickr, it's a potential flaw in the data. I'm not saying at all that Flickr may be biased, I'm saying the data may be biased in a way that prevents us from drawing certain conclusions. Again, this is not a conspiratorial assertion, it's a statistical assertion. I'll let you read the Wikipedia article on it.
...
In summary, Dilbert, Flickr's data can lead to a valid conclusion summarized as: "among all Flickr account holders who keep metadata in their images, the prevalence of cameras is given in the chart below..."

What they did NOT say, and which would NOT be true, is that they can extrapolate that to the entire camera-buying community to infer what kinds of cameras people would like to buy.

Sure, you're concerned that there could be bias in the source of the data (for example, people expunging EXIF data.)

My assertion is that the expunging of EXIF data is not specific to a particular manufacturer and that any inaccuracies in data equally effects all vendors thus the percentages and changes in them are representative of the actual market.

Which is an entirely unwarranted assertion. Personally, I'd hazard a guess that the average smartphone photographer doesn't even know what EXIF data is, let alone how to strip it from their photographs.

I have to agree with lintoni: it certainly might be true; however, there are too many "but what if" situations to be confident that it's likely to be so. Dilbert, we'll just have to disagree on this point.

 

[Orangutan]

Original blog link:

http://blog.flickr.net/en/2015/01/13/camera-ownership-on-flickr-2013-2014

Note that it is "Camera ownership" that they're reporting, not a % based on photo count.

Although where it doesn't make sense is:
2014 Top 10 Cameras (Canon): 600D -> 5D2 -> 7D -> 60D -> 5D3
2014 Top 10 DSLRs (Canon): 600D -> 7D -> 60D -> 5D3

Those two lists should align!

Why? There are all kinds of possibilities to explain the data in other ways. While your explanation is plausible (i.e., could be true), there's nothing to distinguish it from other plausible explanations as the more likely option.

 

[Don Haines]

...
I have no way of verifying this, but my suspicion is that a disproportionately large number of advanced users will both be shooting with a DSLR and that those users will be editing with software like photoshop or lightroom and stripping out the EXIF data.... this would result in an under reporting of "fancy" cameras and a large "unknown" group. This factor alone destroys the accuracy of this survey.

Only in so far as it concerns phone cameras vs fancy cameras.

Amongst the fancy camera ownership, there's no reason to suspect that any one group would use EXIF stripping more than any other group.

agreed... this crosses all manufacturer boundaries....

 

[zim]

Maybe it's just me that finds 2/3rds of the original posts images from flickr no longer available ;D

and of the image thats left it would seem from this argument that Apple sell more phones than Samsung, by a lot. :