Renegadetim: I am no longer posting regularly on woot, but someone let me know that you'd responded to a previous post of mine and I might want to take a look. I think there are some significant misconception of my analysis in your post, so I hope that you'll forgive me for dropping by to clarify a few things, but not sticking around. I'll try to hew closely to "facts" versus "discussion" (or commentary) in offering a few clarifications.
renegadetim wrote:In all honesty, this data isn't really a definitive measure at all in regards to sales.
Sales numbers are, in fact, a definitive measure of sales.
Yeah, it's kind of surprising and may say something, but regardless of what it says, it doesn't say the redesign, the blanks, or the price increase are the reason for it.
I agree with you 100% here. I just re-read my post and I didn't make any claims about the reasons for the decreases in sales: just said that it larger than I expected. (We've been tracking sales data for over a year now and by the time that I made the post to which you're responding, I'd already demonstrated that sales had dropped again, so my starting position for this analysis was based on a previous analysis. That data may not have been posted here: I've been involved for a while in maintaining Narfcake's spreadsheet, but he doesn't let me post the nightly numbers, so it may not have been clear that I'd already calculated how much the numbers had dropped.)
The only possible way to come up with something like that would be much more in depth and maybe knowing you and Narf, you'd be willing to do it. But it would require some kind of ratio to how popular those particular designs were in relation to these. Not to mention price being an issue(1st day sales vs sales during week, not including woot off sales).
I'd like to see a week vs week comparison, and a month vs month comparison around the time of the price changes, blank changes and then from the site redesign as well.
I can't recall seeing your name around the Reckoning thread much, so I'm guessing that you haven't been following the full analysis that I've been doing. If you have and I don't remember your name, please accept my apologies and skip reading the following summary. 
I looked at:
- price increase
- artist
- debut blank for women
- debut blank for men
- debut blank for children
- blank changes (i.e., shirts that debuted on AA and transitioned to Anvil before they were Reckoned)
- color
- day of the week
- reason for being on sale: Daily, Derby1, Derby2, Derby3, Editors' Choice, or random
- first day sales
- sales including first day until Reckoned (through midnight on Monday of Reckoning)
- weeks available in the Reckoning
I've shared a small portion of those results with other wooters off-woot, but I've chosen not to post that information on woot because, frankly, I can either make money off the analysis indirectly (by writing it up and submitting it for publication, thereby increasing my chances of earning a merit raise at my job) or I can give it to woot for free but receive no benefit myself. Since I put about 60 hours into this over two months, at the cost of some productivity at my job, I'd like to salvage something out of it in terms of my job!
I have also, to the best of my knowledge, not referred to those results on woot. The analysis to which you are responding was a simple calculation of sales achieved by adding up sales numbers (and dividing sales numbers to achieve percentages).
This information you are presenting is faulty because there were some VERY popular shirts in that particular reckoning.
Part of what I do for a living involves teaching statistics and research methods to both undergraduate and graduate students, so I'm going to make a really important point before I end up largely agreeing with you. I hope you'll consider what I'm saying, even if it sounds a bit like splitting hairs, because it actually matters quite a bit.
- The information that I presented was not faulty at all: the information was based on sales data provided directly from woot. You may disagree with the analysis or the metric, but the information was not faulty.
- I think what you're trying to say is that, in your opinion, comparing sales from the week of June 8, 2011 to the week of June 8, 2012 is meaningless, because the shirts that are being offered now aren't as "popular" (i.e., selling well) as the ones being offered then. That's an absolutely fundamental flaw in your argument: you're saying that the shirts selling a year ago were selling better than the shirts selling now, so it's unfair to say that the shirts selling now are selling fewer shirts.
I'd also argue that one week's worth of shirt sales is a fine measurement to compare to another week's worth of shirt sales, as long as one specifies the units and doesn't extrapolate beyond one's data. In this case, I firmly believe that - according to the best practices of professional statistics - I was quite accurate about my units, my metrics, and my conclusions.
I think your basic, fundamental point is that comparing sales from a week a year ago to a week now might be interesting but is essentially pointless. I agree with you 100% there.
Even then, it's based on the popularity of the designs, you can't blame/pinpoint any of it on price, blanks, redesign.
Actually, no: you can, using sufficient data and appropriate statistics. As I mentioned, I've chosen not to post those results here and I think it would be utterly unfair to "tease" them by saying anything about them.
I feel like you are trying to prove a point as to how much sales have dropped because of the blank change and redesign, but using variables that favor your argument when there are many other variables in play here. Also, didn't you say there was a woot-off during that particular week?
To answer your second point first: no, there was not a woot-off during the week of 216. There was a woot-off during the week of 215, which is why I chose not to analyze those sales.
It sounds like you might be a little confused between my references to the data Narfcake collected during 201 and my analysis of the data Narfcake collected during 216; several of the shirts that you mention as having been popular and thus skewing my analysis of sales during Reckoning 216 were long Reckoned by Reckoning 216. Although I mentioned that the data from Reckoning 201 wasn't sufficient for analysis so our best comparison was data from Reckoning 216, I can see how easy it would be to get confused by that. [And yes, there was a woot-off during Reckoning 201, but that doesn't affect the sales data from Reckoning 216 for obvious reasons. ]
That would clearly skew the sales for that particular week, ESPECIALLY since there were VERY popular shirts at that time. There were at least 5, maybe 8, sell out first day shirts(Liederkranz, Poison, Library, Acquired, and UnNinja were definite sell outs, and I think In harmony, Meh, and "they're watching" were too) on that list and 8 or more shirts that were #1 on the reckoning at some point in their lives on that list. Nothing of which you could say about current shirts, but that's because of the designs themselves, not everything else.
Once again, you can't have it both ways: you're saying that it's unfair to compare sales because a year ago, some shirts were selling better than some shirts now. Yeah, that's pretty much what I said.
If you've taken any statistics, you're familiar with the concept of "correlation is not causation." Of all the posters on woot, I believe I don't cross that line. Some wooters, including some posters that I think I'm associated with, have insisted that the decrease in sales is the result of the change of blanks. I'm kinda wondering if you're conflating my analysis demonstrating correlation with the claims by others of causation, because you'd have to perform a personality transplant on me in order to get me to acknowledge that ANYTHING causes ANYTHING else.
(Oh, I guess I've said a few times that there must be something fundamentally wrong with me because anyone I care about ends up hurting me, resulting in this personality. I maintain that particular correlation must involve causation somehow, but I fully admit that I can't demonstrate it quantitatively!)
I hope that clears up a few things. I've been teaching long enough to know that I rarely explain things as clearly as I thought that I explained them so my apologies if I've made things worse instead of clearer.
