Amazon Most Wanted Monitoring.

Symbios63 said:

Ok, as I was suggested, here is a correlation between my "amazon most wished for items" points vs actual sales. Thank you to the guy who gave me the idea...

 

No problem.  Thank you for doing them!

 

Symbios63 said:

(I can't read his name for some reasons...)  :)

 

Don't worry, it's not just you.  I used angle brackets in my username, not realizing the site would think it's an HTML tag.  I was initially going to start over but it's kind of become an identifying feature in its own right now.

Anyway, my username is "ColdFire" - it's in my sig.

 

Symbios63 said:

For sales, one point = 1k

 

What exactly are the sales figures you used here, by the way?  Are they NPD numbers, global sales, etc?

 

Also, please forgive my inclusion of your charts.  I know you wanted them deleted from our replies, but I wanted to comment on each one.

 

Symbios63 said:

 

This is, I think, the best match of them all.  It really only misses one major sales spike, and it begins to trail off a little at the end, but it's otherwise quite representative of the overall trend.

 

Symbios63 said:

 

Here, the Amazon numbers appear to anticipate the big sales spike, but then they fail to drop as sales decline.  They remain fairly level, even rising a couple of times, while overall sales steadily decrease.  They do eventually start to drop a little before shooting up again, apparently anticipating the second rise in sales.  Overall, this one seems to do a good job of predicting sudden bursts in the market, but not so much at representing gradual changes.

 

Symbios63 said:

 

Hmm... I'm not quite sure what to make of this one.  The Amazon figures start to rise well in advance of the actual sales spike, and then continue to rise after the spike has ended.  It's not until the actual sales bottom out that the Amazon figures peak and begin to decrease.  Shooting around the target a little, but overall, not bad.

 

Symbios63 said:

 

Yeah, the absolute values are off, but to be honest, that was never what I wanted to compare (I'm actually surprised the absolute values in these charts match as well as they do, to be honest); I was more interested in whether Amazon could be used as a guide for overall market trends.

This one doesn't really have any massive sales spikes to compare, but the most dramatic rises and falls do appear to be slightly anticipated by the Amazon figures, and there's definitely a broad correlation between overall trends.

 

Symbios63 said:

 

WTF?  This one is all over the place!  The Amazon figures drop while the sales figures rise, and then when sales top out and start to fall, Amazon's figures start to climb again.  I'd say that the Amazon figures were anticipating (or perhaps even lagging behind) actual sales, but then they do seem to match the sales trends on the right side of the chart.  They match the spike in the middle (though they remain high a little too long) and they appear to, once again, anticipate the rise in sales towards the end.

So, yeah, this one has a bit of a bumpy start, but overall, it manages to pull it together by the end.  Not a terrific correlation, but also not as bad as it might seem at first glance.

 

Symbios63 said:

Well, so far, exept the 3DS, I am rather nicely surprised by those correlations. What do you think ?

 

Don't be too hard on the 3DS chart.  It actually correlates well trend-wise.  And that's what I was interested to know - if Amazon was representative of overall trends.  I honestly never expected the absolute numbers to match.  I just wanted to know if Amazon could give you a snapshot of what the market is doing at any given time.

Overall, I'm kind of surprised by how well it works.  I realize this is just a small slice of the data, but based on what we have here, I have two major takeaways.  First, it does seem that Amazon is at least somewhat representative of overall sales.

And second, Amazon appears to be slightly predictive of sudden changes - big spikes (or big drops) in sales tend to be preceded by a sudden rise (or drop) in interest on Amazon.

Again, this is all based on the admittedly small sample we have here, but assuming this data is all accurate, I don't think any of this is too unreasonable.  And now that we know how the correlations look, it will be particularly interesting to see if these ideas hold up going forward.  So, I hope you'll keep working on this because we'll get a better look at the big picture as more information comes in.  You're off to a good start and I look forward to seeing how the data builds over time.

Thanks for your work on this.