Monday, March 11, 2013

The State of the Stock(inette) Market: February 25th-March 10th

So, as you might have guessed from previous posts, I'm a big old nerd.  Most of my college career was focused on the quantitative analysis of ethnographic, anthropological, and linguistic data (told you I was a nerd), and it was inevitable that this would eventually bleed over into my work in the knitting world.  Special thanks go to my friend Sarah for the final kick in the butt a few weeks ago to get this started; after a conversation in which she described my ambitions as "wanting to be the knitting Nate Silver," I knew I had to pursue my interest further!

For the past couple weeks, I've been working on a project that I hope will prove beneficial and interesting to other knitting designers, people in the industry, and just plain old statistics nerds like myself.  I've always been interested in the numbers of the knitting world, and there's a wealth of raw data available on Ravelry--so I thought, why not combine the two?  The Stock(inette) Market is the result!

My method has been simple: ideally twice a day, ideally 12 hours apart (ideally, because I am human after all), I take a look at the first page of one Ravelry's search algorithms, Hot Right Now.  There are 48 patterns on the first page, representing the patterns that are getting the most attention at that time on Ravelry.  Within those 48 patterns, I've broken down my analysis into five different classifications: Garment Type, Fabric Type, Yarn Type, Color, and whether or not the item is on a model.  After collating all this information every two weeks, I'll present an analysis of where things stand in  knitting design in that timeframe, and how it compares to previous analyses.  With this information, I hope to get a sense of where trends are shifting in the knitting world, thus enabling myself and other designers to best use the resources available to us and further make our way in the industry.  (God, the academic language, it just comes right back, doesn't it?)

I'm definitely interested in as much feedback as possible; are there further analyses you'd like to see? Things you'd change? I'm personally thinking about splitting a few garment categories further down into their component parts (neck accessories is the big one I'd like to split, but that gets into very semantic territory), and am open to further discussion.  I've also taken screenshots of the page each time I've accessed the data, and so would be able to go back and look into things further.  I'm interested in eventually incorporating knit vs. crochet, yarn weight, yardage, date published, price, and a few other factors, but I'm also trying to be realistic about the time commitment! I'll most definitely take into consideration any suggestions, though, and welcome them.

Without further ado, then, here's the first two weeks of The Stock(inette) Market!

Note that this chart is an incomplete view of all the patterns that were on HRN--this represents only the garment types that averaged more than one appearance per data collection event.  Not pictured: toys, men's, kid's, mitts, mittens/gloves, dresses, jewelry, vests, ponchos, or unmentionables.  It's worthwhile to note that there was a spike in vests starting on the 6th and continuing through the 10th, partly due to a number of them in the most recent Knitty and the release of Splitstone.  Also spiking over the last few days were ponchos from various sources, which indicates that there was either a popular forum post or blog post on the topic. (EDIT: dergugelhupf wrote in the comments that there was indeed a forum post in For the Love of Ravelry by someone talking about ponchos with sleeves, or "swonchos" [shudder].  That would definitely explain the spike in ponchos!)

On those garment types that do appear on the graph, neck accessories (scarves, shawls, and wraps) are still incredibly strong, following a trend that's been going for at least three years at this point.  There was a large spike in the last week due to the release of Quince & Co.'s Scarves Etc. 2013, a collection of 17 scarves, shawls, and cowls.   Garments took a bit of a backseat due to this spike and dropped accordingly.  Homegoods, hats, and socks remained relatively steady.



Fabric type is determined by what I feel to be the prevailing motif of the item.  Lace continues strong, a trend that started right about when shawls did.  Textured knits, however, ranging from simple garter stitch to slip stitches and knit-purl combinations, climbed over the weeks due to a strong presence in the Q&C Scarves, Etc. 2013.  Colorwork (defined as anything with more than one color, be it stripes, stranding, intarsia, colorblocking, and so on) rose steadily, but didn't have a huge presence in either Knitty or Q&C.  I'm interested to see where that one goes! Cables and stockinette were steady in the background, with some popular and well-loved patterns (the Featherweight cardigan and Aidez, for example) staying on the front page consistently. 


Yarn type is a pretty interesting illustration that solid and semi-solid (i.e. kettle-dyed) yarns are still standing tall, though there were some big spikes in 2- or 3-color projects throughout the weeks.  There is an almost direct shift between solids and semi-solids when Q&C's Scarves, Etc. met Knitty, which came out a few days later; the semi-solids were definitely very well represented in this issue of Knitty.  Tweed, variegateds, and self-striping yarns remained steady, with longstanding favorites such as Hitchhiker, Storm Mountain, and Wingspan appearing most every day.


Color has been one of the most interesting things to track over the last two weeks; as you can see from the graph, it was an incredibly diverse selection every time.  Neutrals made a strong showing, with a spike in grey on the first day and a steady climb in brown and white/off-white over the weeks.  Red also climbed, but wasn't a large portion of either Knitty or Q&C. Alternately, green started off strong, but dropped steadily over the two weeks.  Other colors remained relatively steady, and have me interested to see where these will go in the long term.  Are color trends indeed seasonal?


The last option I explored was whether or not the item was modeled by a person.  Most discussions of how best to photograph your patterns encourage doing so on a person, but approximately 1/5-1/3 of patterns were without a human model.  Quite a number were pictured on a dressform or similar set-up, however, so that might be something I track in the future.  An interesting aside is the dips below the trendline on the weekends; far fewer patterns were modeled during those times.  Something to keep an eye on!

So, to sum up, shawls/scarves/wraps and lace remain strong, with a rise in 2 or more color patterns and texture.  Semi-solids and solids still rule the yarn market, partly thanks to the Quince & Co. Scarves, Etc. 2013 and the most recent Knitty.  Green is down, and reds and neutrals are up.  Where do you think things will go in the next few weeks?

61 comments:

  1. this is so interesting! just the kind of statistic i love, love, love to read.

    ReplyDelete
  2. What a fun and cool idea! I am definitely on board for this ride. I think we'll see a rise in green while lace will remain strong. Or maybe I'm just designing a green lace shawlette right now. :-)

    ReplyDelete
    Replies
    1. Haha! I am a green girl through and through, and so am hoping that it'll track back strong. We'll see what the next few months will bring!

      Delete
  3. This is fascinating! Thanks for putting together all of the charts.

    ReplyDelete
    Replies
    1. My pleasure! I'm so glad it's as fascinating to others as it is for me!

      Delete
  4. Brilliant! Can you factor in weather trends? For example, did recent the East coast cold snap shift focus from Spring greens back to warming reds? I suppose you'd then have to balance it for the proportion of Rav users in various parts of the globe, but it would be really interesting if there was a correlation. Thanks for taking the time to do the analysis!

    ReplyDelete
    Replies
    1. That's an interesting thought, and would be completely fascinating to try to correlate. I'll look into the feasibility!

      Delete
  5. This is doing it for me in so many ways! Awesomesauce Bristol! love it!

    ReplyDelete
    Replies
    1. Thank you so much--I was hoping there were other data nerds out there! :)

      Delete
  6. Interesting analysis. I was wondering if you could clarify something about your sample size. You mention at the beginning 48 patterns on the first page of HRN. Later, you state that it is an incomplete representation as you only represent garment types with more than one representative per data collection point and that certain categories are completely excluded. This leaves me very uncertain as to what the actual sample size at any given collection point actually is. On the comparative line graphs, that is not so bad as the comparison there is mainly between the different lines. On the modeled graph, however, it would be helpful to either have a non-modeled line plot or a total items sampled baseline to which we could compare the modeled shot data.

    ReplyDelete
    Replies
    1. Those are both good points! I had originally intended to graph all garment types, but the list was so long and so many averaged below 1 per data collection event that it would have made the bottom half of the graph incomprehensible. I will keep looking for solutions for how to incorporate them, or perhaps to also offer a bar graph with percentages.
      On the modeled/non-modeled, I think I will also switch that to a bar graph with trendlines, and factor in dressforms vs. unable to be modeled (i.e. toys) as well. This is definitely a work in progress!

      Delete
  7. Wow, this is just so interesting! Thanks for being such a nerd - I'll keep an eye on this and on your whole blog, too.

    ReplyDelete
    Replies
    1. Thanks so much--I'm so glad that other people find it as interesting as I do!

      Delete
  8. Love it! Dates are cut off on the x-axis though; is that intentional? (Math textbook editor here...of course that would stand out for me.) Also, this is Tina (of the smaller variety). See you at KPG brunch this Saturday?

    ReplyDelete
    Replies
    1. Ah yes, the limitations of the Numbers screencap. :) I'll be fixing that in future editions! And booooo, I have a prior engagement during brunch, which is PANTS. Next time!

      Delete
  9. wow - I love this! look forward to future installments... and I'll do my best to keep the pullover and cardigan counts at least present on the charts!

    ReplyDelete
    Replies
    1. Thank you! And yep, I'll be doing my part on keeping sweaters popular, too. :)

      Delete
  10. This is very interesting! Thanks for doing this!!! How will you be "announcing" future installments? Please don't say F***B**k, as my only option....
    pvb

    ReplyDelete
    Replies
    1. Thank you so much! I'll definitely be announcing it via Twitter, and if there's enough interest, in the Designer group on Rav.

      Delete
  11. So much fun!

    I think you need a backup data person. If there's a point where you're going on vacation (or something) and you know you'll miss several data collection points in a row. In fact if you're looking for coverage let me know!

    I agree breaking down neckwear would be helpful. Rather than getting tied up in semantics, what if you just accepted whatever label the designer applies to the pattern in Ravelry? Since whether the ravelry tag is for shawl, stole, scarf, etc... is going to effect the search results accepting that term shouldn't skew the data off that course. Essentially, I'm encouraging you to accept whatever bias the designer assigns the project.

    ReplyDelete
    Replies
    1. That's a good point about the self-selecting bias; I'll look into that and see if I can divide it that way. I may take you up on your offer of data collection--Casey at Ravelry steered me to the API group, which, if I work up the courage to give it a try, would eliminate a lot of the grunt work. I'll let you know how it goes!

      Delete
    2. Well figuring out a way to collect the data automatically would be the best solution. Computers are much better at that sort of thing than humans. Then all you'd have to do is the fun charting bits making the whole process much easier on yourself!

      Delete
  12. Hi Bristol - you might be interested in the Ravelry API. The pattern search API provides all the same options as the on-site search (sort by HRN, whatever you want) and if you are on a Mac, grabbing a page or two of search results is as easy as running a command in Terminal.

    http://www.ravelry.com/groups/ravelry-api

    ReplyDelete
    Replies
    1. Thank you so much, Casey! I had looked a little bit into API, but got intimidated at first by the coding aspect. I think I just need to be brave and give it a try!

      Delete
  13. Interesting! Thank you for doing this!


    I will occasionally track the top selling patterns as well- to see what seems to be selling well, average price, etc. But I only ever do it weekly (when I remember), just to get a recap!

    ReplyDelete
    Replies
    1. My pleasure--so good to know there are other stats nerds out there!

      Delete
  14. w00t! Bristol you are such a rockstar! I love your analysis!

    ReplyDelete
  15. Wow! This was very interesting. Thanks for taking the time to put it together and explain it so clearly!

    ReplyDelete
  16. This is so interesting! I wonder what the reason is for the drop in modelled patterns on the weekends. My guess is fewer professional patterns are released then (because of expected lower traffic)?

    ReplyDelete
    Replies
    1. That would definitely be an interesting hypothesis! I've been asked to start tracking in terms of industry pubs vs. self-publishing as well as modeled vs. non-modeled, so I'll see if there's a correlation between the two!

      Delete
  17. "Also spiking over the last few days were ponchos from various sources, which indicates that there was either a popular forum post or blog post on the topic."

    There was a forum post about ponchos with sleeves in FTLoR during that time period. A Raveler posted hoping to get a specific attribute for that type of poncho and listed many patterns of that style. The click-throughs created a ripple effect on my friends radar too, and certainly on others'.

    ReplyDelete
    Replies
    1. Here is the link to the discussion - the OP calls them swonchos, as some publishers do as well.

      http://www.ravelry.com/discuss/for-the-love-of-ravelry/2501866/1-25

      Delete
    2. Thank you so much--I've added a note to that part of the post!

      Delete
  18. Karen in upstate NYMarch 12, 2013 at 6:27 PM

    What a mind exercise...have fun. I am not a daily visitor to Ravelry, but I have noticed something. There is often in the top twenty a pattern that is being temporarily offered for free....say a designer is celebrating something and offering one or more patterns at no charge for a short period. This might cause a spike in popularity for other than purely design,etc. reasons.

    ReplyDelete
    Replies
    1. I am definitely adding free vs. paid vs. on sale vs. free for a limited time to my analysis--it's definitely a big factor in people viewing the pattern!

      Delete
    2. Agreed - this practise of offering a pattern free for a limited time has snowballed in the last few months and when unexpected patterns show up in the top five on rav this is almost always the reason.

      Delete
  19. Replies
    1. So sorry for the late reply--such a lovely compliment! Thank you so much!

      Delete
  20. Love the analysis! As a fellow geek and number cruncher I can't wait to see how trends change over time, such as color. Do we see trends towards brighter colors during the last few slogging weeks of winter (at least here in Colorado)?

    Thanks for having such an inquiring mind.

    ReplyDelete
    Replies
    1. Apologies for the late reply! I think color is such a fascinating trend to watch, especially now that I've got almost a year's worth of data. It will be interesting to see where things go in the next couple months to see if it's cyclical, but I do think there's a push towards more color as the winter wears on. I know I definitely need it!

      Delete
  21. This is really neat, my inner stats nerd rejoices. One question though: have you adjusted for the "Yarn Harlot Effect"? I've noticed anecdotally that anytime there's a post on Stephanie's blog that that pattern rockets up the popularity chart, so have you adjusted for this in the analyses? Alternatively is it a real, statistically significant effect?

    ReplyDelete
    Replies
    1. So sorry for the late reply! I haven't discounted for the Harlot Effect, solely because I'm trying to record without judgment. I don't have first-hand experience on whether the press from the Yarn Harlot affects sales, but notes from a Harlot-ed friend indicate that it does make a huge difference.

      Delete
  22. I love this! I found you from the mention in the Knitty blog today, so your readership may go way up :) I agree with maget that there's a "Yarn Harlot Effect" - I'm not sure there needs to be an adjustment for it though, it just helps explain some of the trends.

    ReplyDelete
    Replies
    1. Apologies for the late reply, but thank you so much! I definitely agree that the Harlot Effect should be monitored and not discounted, just as any other trend. It's an important part of the knitting world!

      Delete
  23. This is a genius idea! So useful for designers.

    ReplyDelete
    Replies
    1. So sorry for the late reply, and thank you so much!

      Delete
  24. Awesome!!! You've done an excellent job :)

    Are you measuring projects started or completed? Have you thought about comparing the two?

    ReplyDelete
    Replies
    1. Thank you so much, and so sorry for the late reply! I haven't folded in any other data but Hot Right Now, as it's the most easily obtainable and clear raw data available. If I'm able to expand this in the future, that'd be a great thing to add!

      Delete
  25. I would love to see the seasonal variation that Elizabeth mentioned above. I wonder if knitters are anticipating the upcoming seasons, or knitting to catch up to the season that has burst upon them?
    Also, have you thought to include yarn weights in the analysis?
    I'm delighted with your work, not as a designer, but as a interested consumer!

    ReplyDelete
    Replies
    1. Apologies for the super-late reply, but thank you! There definitely is a good amount of seasonal variation, both in terms of "okay, I'm ready to be done with winter now" and reflexive "oh my gosh it's fall I'm cold I need a sweater". It's interesting to figure out which is which!
      I'd like very much to add yarn weight into the discussion, and will definitely do so if it becomes feasible.

      Delete
  26. Wow, this is so cool. Loved looking at your analysis!

    ReplyDelete
    Replies
    1. So sorry for the late reply, but thank you so much!

      Delete
  27. Outstanding. One of those "why didn't we think of this before" ideas. Clearly you have stirred more than a few brain cells among us. I look forward to more in the future.

    ReplyDelete
    Replies
    1. Thank you so much, Susan, and my apologies for the late reply! I am pleased to get people to think about knitting in a different way. :)

      Delete
  28. As a fallen away number cruncher this is fascinating. I postulate that seasonality is an important variable. And that a longer time span is needed to see solid trends. Time may be better than sampling every 12 hours for noise reduction.

    ReplyDelete
    Replies
    1. Thank you so much, and so sorry for the late reply! The upshot to me replying nine months later is that I can agree that yes, more data has definitely brought some interesting things to the forefront and eliminated some of the noise. :) I'm looking forward to seeing how it continues to develop!

      Delete
  29. What a mighty effort in the step toward trend analysis - and gathered and plotted manually, too! I just have a couple of questions. What are your x and y axes for each of the charts? How do the solid lines correlate to the individual data points? As indicators of slope (y=mx+b) for trend directionality?

    Like I said; very intriguing.

    ReplyDelete
    Replies
    1. Thank you so much, and my apologies for the long wait in reply! The x axis is the date, and the y axis is the number of patterns, out of 48 possible. The solid lines are connectors only between the points, and the trend line was auto-generated by Numbers, a spreadsheet software, so I would assume they follow a standard formula for slope. Hope this clarifies, and thank you again!

      Delete