Etsy Taste Test

I spent some time this morning playing around with an "experiment" our portfolio company Etsy has launched on the web. It is called the Etsy Taste Test.

It looks like this:

Etsy taste test

The idea is you click on stuff you like and Etsy builds a "taste map" for you. The end result is a set of recommendations for you. As Etsy has grown, the "discovery problem" has become an issue. Etsy has a bunch of stuff coming out to address that issue and this is an example of one of them.

Give it a try and let me know what you think.


Comments (Archived):

  1. William Mougayar

    I tried it. It’s pretty cool.The surprise element was more important than the precision part of whether results matched what I selected.This method can apply to anything, i.e. with a little human-aid, computers can do wonders.

  2. RichardF

    It’d be cool if you could link the result to a public profile or post a link in Facebook so that people could use it to buy presents for you. A little narcissistic I know but great if you are a difficult person to buy for.

    1. fredwilson

      Good idea

      1. Dave Pinsen

        Fred,OT, but I remember you mentioning the iPod Nano watch idea in the comments here a while back. I just saw a tweet about how much money they have raised to make those. Impressive.

    2. Fernando Gutierrez

      You already can. They give a permalink, you can tweet it and you can like it in facebook.

    3. Fernando Gutierrez

      If this were to be a really good guide for others you should be able to take thinks out of the suggestion list… that’s why I don’t post here mine, I know you all will try to send me presents and there are a couple I don’t like much 🙂

    4. Donna Brewington White

      Brilliant!BTW, narcissists are some of the most difficult people to shop for.

  3. LIAD

    Etsy have trail-blazed a whole new industry. The way they have empowered people is inspiring.The discovery problem is a source of discontent for sellers as well as buyers.They have a powerful API. Why haven’t they launched a rev-share affiliate scheme? Developers would be only too happy to muck-in if they got to share some of the upside.(doing pro-bono work on API’s where the platform doesn’t directly earn revenue from your effort, a la twitter is one thing, working for free when the company does directly benefit and could easily share revenue with you is another).If they shared the wealth, discovery and other problems would disappear overnight.

    1. fredwilson

      that’s a good suggestion LIAD

      1. Ed Lea

        Fred, you’ve acknowledged this suggestion a few times (which I hope means you like it :). What have Etsy said when you put it to them? How realistic is an affiliate cut on 3.5%? Or perhaps an additional layer of affiliate payout on top, paid for by the sellers themselves for whatever external service they use. We’re keen to know what the future could hold at

        1. LIAD

          Etsy themselves would have to take the hit on their revenue for the aff scheme to work. Developers would only embrace the api etc if they knew they could get paid on all transactions they facilitate and not just the ones which certain sellers had agreed to.In the UK £1 of every £10 spent on retail is done at Tesco – when they recently announced an api everyone began jumping for joy – what a great business to build an ecosystem around. Tesco claimed small margins prevented them for doing any kind of rev share. Their scheme stalled as you would expect. Finally they relented and offered a measly 10p per transaction regardless of the size of the transaction the affiliate facilitated. Whilst in my opinion this was derisory and I wouldn’t even consider my developers spending a moment looking into the api in detail the scheme as a whole began to take off.I don’t think the aff scheme needs to be a massive profit centre for affiliates, like tescos isn’t. It just needs to show the developers that the company has a little goodwill.

          1. Ed Lea

            “I don’t think the aff scheme needs to be a massive profit centre for affiliates”I do.

          2. LIAD

            Let me rephrase that.For hard and fast affiliates. Where ROI is paramount, Etsy are going to have a hard time convincing them to spend the time and effort building on their API as the rewards (50% of 3.5% – ?) are marginal at best. So it think that’s perhaps a non-starter.For the enthusiast developer. I think they would be happy to roll their sleeves up as they have a double bottom line when it comes to building things. They just don’t want to be taken for a ride. And therefore getting anything is really just a sign of good faith more than anything else.

        2. fredwilson

          those are the issues for sure

    2. howardlindzon

      yep…..thinking about that for stocktwits


      As an etsy application ourselves, I can say that I would love this. You can make money without Etsy directly giving it to app makers. Some Etsy applications charge money for using them. Others like us make money through advertising. As Etsy does supply a steady source of traffic, advertising does translate into money. That being said, I don’t think anybody would complain if Etsy started to share the wealth. I think it would be a bit more complicated for seller tools who indirectly relate to sales, rather than directly like a buyer tool. However, we (and other tools makers) made Etsy quite a bit of cash cash over the holidays. We did some measurements that showed that listings sold twice as often with our service. Etsy should certainly repay the favor with either money or by driving more traffic to our sites (which translates to cash) through better promotion. Either way would be great!I am not horribly worried about this happening or not though. Eventually it will, of that I’m sure. ( The sooner the better of course 🙂 )They will see the difference in their bottom line and they will make adjustments. They already laid out plans to track API transactions and such to gain information. This will likely be used in the next year for this or similar purpose.

  4. Tereza

    Taste engines tend to put me on edge, I’m not sure why except that I like to think I have my own unique style and opinions and they always feel overly simplistic for me.Part of it is that I have some quirks which if I could tell the engine ahead of time it could except out a bunch of stuff I’d definitely say no to. For example, I don’t do the colors black ever. And usually eschew white. Would love to tell them that, to not waste their time or mine.The noir lace bunny ears they recommended were just so wrong for so many reasons. Wish I took a screen shot of that. I mean, WTF? No, I’m not joking.So I went through the first ~4 pages having a tough time choosing between items (or more likely it was, “I like the form of that but I hate the color and I like that style but I do v-necks not crew necks so godammit I’m not sure how to answer that will be helpful to them”.Then I must’ve steered them seriously wrong because a screen came up where I hated everything so clicked “show me another” and I hated all of those and then fully hated another 4 screens.That’s when I got frustrated that I was wasting my time.And then I got served up a screen of things that were only OK, but included the black lace bunny ears.So the engine — may work well for normal people. Alas, not moi. Sorry.

    1. William Mougayar

      Relax…& take it with a grain of salt…or vodka.It’s partly a game more than anything.

      1. RichardF

        William, telling some women to relax about shopping is like swearing at them 😉

        1. William Mougayar

          I apologize if it came across that way. My point was not to take these recommendations literally. I saw them more like suggestions, and a gamification of discovery.

          1. Tereza

            No offense taken. But I don’t view it as a male/female thing per se.I love meandering but my life stage has me as a practical/utilitarian shopper and so when I saw this it was with the mindset that it was supposed to save me time and delight me by showing me things I’d want.I don’t have time for games and wouldn’t have done it if I thought it was a game.I clicked through and tried it because I thought it would save me time.I stepped away unhappy because I lost time and had poor choices.

          2. William Mougayar

            The fact that I and some others saw it as a loose “suggestion” list process, whereas others were expecting more precise recommendations or were annoyed by its inacurracy or length is probably a good data point for Etsy.

          3. RichardF

            William it’s me who should apologise, it was a poor attempt at humour on my part and I agree with you about them being only suggestions.

          4. ShanaC

            I still think the holy grail of explaining shopping is Paco Underhill. And I always get surprised that not so many places apply what he says. The exception seems to be Amazon with the recommendation engine -it always appears after you put something in your cart to tantalize you…

    2. Dave Pinsen

      You’re too unique.

      1. Tereza

        :-)I guess noir lace bunny ears are mainstream.

        1. ShanaC

          I have to wonder who is going to buy those

          1. Donna Brewington White

            Somebody’s husband?

      2. Mark Essel

        too unique, for shame Dave, unique is a binary quality.

        1. Dave Pinsen

          You’re right. Edited. Shaming successful.

          1. Mark Essel

            bonus points for the strike through formatting.

          2. Donna Brewington White

            Gee, you’ve got skills, dude.

          3. Dave Pinsen

            What skill — the ability to acknowledge grammatical mistakes? Give credit to my copy editor Mark Essel.

          4. Donna Brewington White

            Referring to your strike through. I am pathetically impressed by things people can do with html.And yes giving due credit to Mark Essel: A guy who knows his way around code as well as the English language is to be admired.

    3. Kelley Boyd @msksboyd

      I clicked show me another for at least 12 screens before I picked anything at all. And almost everything I wear is black or white. I guess that is left over from flying 200+ segments annually for many years in a row. Everything has to go with everything and two pairs of shoes! hahaha. Clarification: I do have khaki and red jackets for panache! Oh, and a couple of colorful scarves.I always like your style and have even asked where you got stuff. Maybe it is time for me to make a change. When you build or select an engine that drives your personal style choices to go with Honestly Now, I am pretty sure that will be of value to me. Keep the ball rolling forward sister!

  5. JLM

    I suspect it’s a good idea at the end of the day; however, it is a little like setting up a blind date — the results will color the messenger.The Etsy product offerings are so diverse that it may be very, very difficult to get to the 90% confidence level which is necessary to really get traction.I have some odd reading interests and I must say that I do appreciate when Amazon suggests similar books but their suggestions are not really the equivalent of my just searching for my topics.Happy Holidays to all!Merry Christmas and Happy New Year!

    1. ShanaC

      Happy Holidays to you too.Honestly, setting people up blind is much less of a bad idea than it looks. And a lot of people aren’t used to them, so that is why they judge the matchmaker. You have to keep certain expectations in check, and also hope that the matchmaker (or computer) isn’t bizarre because it sees you and a target as flat beings.So nuance is necessary in both fields.

    2. Mark Essel

      Happy holidays and Merry Xmas!As confidence goes over 90%, the value of the suggestion engine goes up exponentially. I want to see bulk filtering done with iteratively improving algorithms, and human specialists doing the really important last mile with a narrow subset and knowledge about the audience/end user.

  6. Tereza

    I just feel like they’re going about it all wrong.For my life it’s best for gifts, because that’s a pain point. Last year I bought a half-dozen gifts at holiday time on Etsy. Most worked out pretty well. And I largely enjoyed the process. I got tripped up when my husband didn’t like and asked me to return one of the items. Thing is it was handcrafted especially for him with a sweet note from the artisan saying “happy anniversary”. So I felt guilty about trying to return it, it sat on a shelf for nine months collecting dust and then I gave it to Goodwill.Last year being our ninth anniversary I wanted male gifts of leather and it was just very messy to find and the options were just OK.I used to get the daily email but cancelled because it’s too much clutter. But as a result I FORGOT to look on Etsy this year for my anniversary gift and as a result bought something non-aluminum (10th) at the Grand Central crafts fair for $120. That money should have flowed through Etsy.I want a “spousal subscription” for gifts. Enter name, gender, age, birthday, wedding date, and maybe some parameters for style. I should get pushed an email customized to me a month before our anniversary with curated recommendations based on OUR anniversary year (wood, leather, aluminum). And for the subscription include a card from Red Stamp and paper to wrap it in (don’t pre-wrap — let it look artisanal).This same thing goes for birthdays. And kids birthdays. Holidays. My 7-yr old daughter goes to 20 birthday parties this year for girls turning 8. Sounds like a subscription to me. Don’t need a fancy algorithm for that.Emails have lost their oomph. Between Groupon, Gilt, Lifebooker, Jetsetter, RueLaLa…..I route them all to Junk now. it’s my way of saving money when I shouldn’t be spending. But thing is when I DO need to buy, I’ve lost touch with the vendors. Unfortunately they didn’t try hard enough to get to know me — in very basic ways — when they had a chance.

    1. Fernando Gutierrez

      I like what you ask for emails in certain dates, I also would love those remainders. But I know many people who still think they can keep some privacy and would freak out if a shop ask for anniversary dates.

      1. Tereza

        I’m totally baffled as to why they’d try to engineer the best possible way to run around catching fireflies in a jar, when they could farm them in bulk.I’m telling you, I come with open checkbook and yet they are not meeting my needs.My tastes may be slightly advanced, but my needs are totally mainstream to the age 30+ female user (97-98% of their users are female)

        1. kirklove

          Is that true that 97%+ of Etsy users are female? Wow, I’m astonished by that. Mainly because I use Etsy from time to time.I’m curious if something like Svpply swings more toward males. My guess would be yes, but still nothing close to 97%.

          1. Tereza

            I got that figure from someone who’d know.But figure that you have sellers and you have buyers. Of the buyers, two types. One has lots of time but not much money. They like to wander around and explore. This sort of thing is built for them. Then you have a buyer who has very little time and lots of money to spend.If it were my company I’d say you need to nail the latter group. They do not. This solution is catering to the former group.

          2. ShanaC

            Of the guys I know who use etsy, it seems that they heard about it from women. I always get the feeling that guys don’t talk as much about where they shop for fashion, which may be the reason why you see such behavior.

          3. kirklove

            Perhaps. I use Etsy because I know some of the sellers and like uniqueone of kind stuff.And I don’t mind telling you what stores I like. That way you can getme a nice Christmas present. 😉

          4. romigli

            Funny – I was thinking about Svpply when I caught this post and was surprised it didn’t get more of a mention esp. since 1 part of its founding team (Zach Klein) runs product for another USV portfolio company (Boxee) amid the other not-quite-random connections (Hunch, Founders Collective).Zach’s blogpost explaining Svpply is far more elegant and articulate than anything I could say and so too is Svpply, thus far:

    2. ShanaC

      I get the feeling I should route more.More seriously though, why do recommendations bother you that much? (unless this thing isn’t dynamically filtered based on what you click on, buy)

      1. Tereza

        My time is scarce and i am constantly battling clutter.Every single day I am trying to accomplish 20 hours in 10. Every second counts, I have no margin for error in my day. I barely have time to go to the bathroom. In that, I am not unique. Ask any working mother.Recommendations that are completely off base are a colossal waste of my time.They also make me feel like I’m not being listened to, and that pisses me off more. I’ll just go find someone that will.

        1. Mark Essel

          My experience with recommendation engines is that there needs to be a human element.Our first take at extracting entities from twitter resulted in 50% good, rest garbage. We refined it to be 60-70% good and rest bad by throwing away keywords/noise. But algorithms have trouble getting to know us, we’re complicated critters ;).Algorithms must reduce clutter to humans specialists who can use their judgment to suggest perfect items, information, apps, etc.

          1. awaldstein

            MarkI consider Amazon ‘if you liked this you should try this” and their self publishing platform for consumer reviews the original and still one of the best recommendation engines for books, films and less so music.Where would a person play into this if you think that engine is off whack?

          2. Mark Essel

            Always right before giving that suggestion to other people. Let’s say company X algorithms yield a few dozen suggested products. A human could read just a few of your blog posts or many twitter updates to gain a real social profile and narrow the automated results down to even fewer better suggestions, maybe just one per visit or weekly/periodic email. Why stop at products?How many social profiles could an individual build? A few dozen, a few hundred based on recent updates. The mechanical turk comes to mind.

          3. awaldstein

            True…sometimes it is easier to solve a large problem with people rather than technology but adding the human factor, almost a personal shopper, creates by definition scale and cost factors.I love shopping from folks who know their stuff and know me. Can’t really think of a model where this works unless the ticket item is large cost and margin though.

          4. Mark Essel

            If we knew of an easy answer we’d be building it Arnold 😀

        2. Blsavini

          “Recommendations that are completely off base are a colossal waste of my time.They also make me feel like I’m not being listened to, and that pisses me off more. I’ll just go find someone that will.”Thank you for expressing this sentiment. A good portion of the “disgruntled Etsy sellers”, share this oppinion.

    3. fredwilson

      this is just the tip of the iceberg TerezaEtsy quadrupled the size of its engineering team this year and we are just starting to see the results of thatthere are all sorts of ideas to make gift giving to others work and your ideas are similar to some of the things on the roadmap

      1. Tereza

        It feels more like oooooh aaaaah value at a Tech Meetup but not high value to high value customers.I would’ve sequenced the roapmap differently. A recommendation engine that doesn’t work leads to fan leakage. They need to leverage and build on the fan base of customers that can pay. Not sound like they’re not listening to them.Every interaction has to elicit delight and there is great risk that they get it just plain wrong. Plus they’re going after mood purchases and not need purchases. Why not kill on the ones that the customer is definitely going to buy, it’s just a question of from whom?Etsy has great customer base and terrific sellers and products. Execute well and they could be at and Groupon level. I see an engine like this as possibly supporting, but not the center of, such a strategy.

        1. David Semeria

          Sequence is not a verb, and if “every second counts” in your life, how do you find the time to post multiple page-length comments on AVC?Nothing personal Tereza – I’m a bit grumpy today.

          1. Tereza

            whatever.usually it’s because i’m waiting for something else. I type extremely fast.what’s annoying is in the setting i come off as a lone voice but then multiple people email me in the backchannel tell me they agreed.i take the time to say it because if I don’t it won’t be heard.

        2. fredwilson

          this is an experimentlike google labsit has not been launched in the core producti believe i mentioned that in my post

          1. Tereza

            yes you did mention it, i get it.i brought this up last holiday time. it felt like a big hole that was very easy to fix, if they acknowledge and respect the problem. (which is marketing/merch, not tech)it doesn’t feel like they tried. meanwhile etsy’s dropped off my radar.i want to support etsy and their sellers, because it fits with my ethos (and time constraints). but i literally forgot about etsy because they’re not trying to reach me in the right ways.that’s money lost for the 5-10 sellers i would’ve supported. that bums me out.

  7. ShanaC

    I’m getting the heavily architecture and classic clothing that I love so much and wish I knew where to buy…but the sunglasses you all are offering me may be very of the moment, but they look ridiculous (those cheesy eighties styles hyper architectural shades that look ridiculous on actual people.As for gifts, that would be kinda difficult. it means judging my friends taste, then letting the recontamination engine do its work. That feels awkward to me…

  8. Fernando Gutierrez

    I tried it and looks good, but it was too long. Many times I clicked that I didn’t like anything and some of the items appeared again.

  9. Kelley Boyd @msksboyd

    After 6 minutes I was only 37% done. It felt “hunchy” and I just don’t like the “please spend your time telling me what you like so I can mass mail you stuff you might look at or buy” Another time suck.I realize, I am not the standard or predicable profile so if I were, what would I value? Etsy could possibly partner with a new site integrated into facebook, I love this because it is an idea I had and actually wrote a business plan for a couple of years ago but could not get any buy in….of course I did not know any of you at the time…anyhooo. Rock on people of the list. Have a great day.

  10. Sebastian Wain

    The problem is hard to solve. It took me ~10 minutes and the result was regular. It seems the algorithm is using some metadata (text, tags, etc) instead of some kind of image/colors recognition? because I choose one robot and in the end a lot of robots appeared (I just liked that robot!). But I think this it typical for recommendations engines.Some suggestions- Study the possibility of doing something like the Netflix Prize.- Study other complements like and their competitors:…- Open Source alternatives like Add some kind of (simple) color recognition, mainly for pictures.Also, always add noise to the suggestions.

  11. RJ Johnston

    The Taste engine performs as expected. No jaw dropping, the product recommendations were mediocre. (I.e. I selected several Men’s clothing items and no more than those selected were displayed) Having said that, this is by no means a metric by which I measure the Etsy experience.Etsy is a fantastic medium for individuals to market, sell and distribute their goods.An alternative I would explore is allowing users to input their taste in products by providing manufacturer/designer/whatever then selecting an image of the style they prefer by the provided manufacturer followed by an Etsy decision of shops that fit this criteria (similar). The old “If you like [big name product] then you will love [independent product]”.One more suggestion I have is expanding and enhancing the “Shop Local” site. Two points – First, you have to guess whether entering your zip code or city/state will produce a result. After providing my zip code an error message was displayed advising city/state is the preferred method (why did I have to guess first??). Second, shop_local.php IMHO should be a replication of the front page or the category search result page ( with auto detection of the user’s general area then allow the user to enter zip and/or city/state etc.Depending on your business model this method would absolutely provide valuable site real estate for shops to market their goods and provide up to date event information where users can come see them. (Assuming many shops attend local craft shows, flea markets, etc) To offset the amount of revenue lost when an Etsy shopper visits the shop/person in person Etsy would be creating a new revenue stream through its local “promoted or featured” shops/products. In addition the conversations generated at these shows will most likely attract new shop owners.I could go on but I think I’ve bloviated enough. :)Thanks for asking Fred!

  12. Aynsley

    I usually make quick decisions and don’t like a lot of options. I finished in less than a minute and loved my results. I’ve never used Etsy before unless I got there through a blog link and checked it out quickly. Discover is probably an issue for me but this tool is great.I got in and out so quick – but I’d love to know if the tool will save all those recommendations of my test so I can go back and browse when I have more time!!!Love your blog – FYI – I added you to my reader yesterday after reading a post you wrote in 2005 about the Attention Crisis!!!! cool huh!

  13. palbi

    Etsy is just amazing and of course I’ve used it also for my xmas shopping this yearThe issue I had is that most of the recommendation tools tends to present you a large majority of pretty low priced items. Which is fair but sometimes doesn’t work when you search xmas presentsOf course you can order the results by price but the experience is not as seamless

  14. kirklove

    Interesting. Kind of like what Chris and crew are trying to do at Hunch. I think that’s smart.I sound like a broken record (that no one really hears mind you) but I fully believe the future winners on the internet will succeed at killer curation. That’s the holy grail. I don’t want to wade through page after page of stuff to find the one or two items that interest me.The internet doesn’t work when it’s work.Quickness, ease, and relevance are the holy trinity moving forward. Good to see Etsy trying.

    1. mike gilfillan

      Agreed. Started a job site in ’95 and all our clients asked for quickness, ease and relevance back then too. It drove our business until we sold it at end of ’06 (regretfully). Also believe that human involvement is a required part of that holy grail.

    2. ShanaC

      It sounds like you are looking for the madison avenue boutique on the internet. My guess is those businesses have thin margins to play with. I get the feeling while the internet will thicken them -it may not be enough because of scale issues.

      1. kirklove

        I’m actually looking for the opposite of that. I want a site (likeAmazon say) that has everything but only shows me the >1% I’minterested. Everything else that distracts or frustrates me from thatsale should be eliminated.Kirk(sent from my iPhone)

  15. ian_peterscampbell

    I tried it twice. The first time iI didn’t see the “don’t like ny of this” button, so I always picked the top item on each screen it showed me. A lot of sets had nothing interesting (though there were a couple inspired items). The results were mediocre.After noticing the “none of this” button I tried again, and it took a long, long time. Feels like maybe it would be better if the learning was applied as I went instead of just at the end, I feel like I had to wade through pages of junk to find one interesting item. Maybe 25 choices on each page instead of 6 would also help me find things that suit without having to do the tedious “not this” click over and over. It would also be nice if I could vote the negative in addition to the positive (I will never ever care to buy jewelry or knit mittens).

  16. NJ Braithwaite

    Maybe the ‘taste map” should include a short list of personal “profile” questions for whomever you are gifting. (Male, female, ethnic origin, age, hobbies, etc.) Combining that personal profile with items you “like” should help Etsy make more informed/personal recommendations. This may also require Etsy to brand each item with a short “profile” of its own. With that, maybe Etsy should give users an option to provide a more extensive “personal” profile that Etsy can use to recommend items for them as well.

  17. Joe Lazarus

    I liked some of the recommendations. Many were off the mark. It’s a tough problem to solve. I did get a good laugh from this one that I posted to Tumblr: top request for discovery on Etsy would be an “interestingness” sort for search (a la Flickr)… something that surfaces things that are universally interesting to lots of people for a particular search. While not personalized, it would help to at least remove a lot of the junk. I suppose the problem with offering that feature is that a relatively small set of Etsy stores & items would receive a disproportionate amount of attention, but maybe that’s not such a bad thing. Create larger businesses for the talented craftspeople and weed out the duds.I do love all the new innovation coming out of Etsy lately.

  18. Mark Essel

    Is there any effort to correlate Hunch’s Q&A powered API with Etsy gathered taste/style information? The more orthogonal measures of information, the merrier when it comes to suggestion engines.How about extending the input sources to social web shares, blog posts, status updates on Facebook/Twitter/Buzz? Then finding tie points or important taste discriminators: Apple fans (design and experience over versatility)Google fans (function over design)Startup fans (rebellious, rooting for underdogs, seeks contrast with status quo)Microsoft fans (old)

  19. shanesnow

    This is something I’ve been thinking about a lot. In e-commerce, the “where should you buy it” has largely been addressed (price comparisons, sites like ebay, etsy, and amazon, etc). But the “what should you buy” is a little trickier. I’ve heard the idea “pandora for shopping” tossed around by a few entrepreneurs here in New York.I always thought the Zappos Smackdown (… was a great idea because it turned product discovery into a game. I played it until I bought a pair of socks and then that was the “Ah ha! They got me” moment.That actually was the fuel for an experiment my friend and I did this fall called a shopping discovery game. We both since moved on to join a startup, but I was happy to see this post as Etsy’s Taste Test has a similar feel to what we were toying with. Great idea, IMO. 🙂

  20. Andrew Greene

    I think it’s a great experiment. I love the design. I like seeing how the products you pick add up on the bottom. Sometimes the photos are so vague that it’s hard to tell what the product is. But overall, I like the design, and I think the recommendations are good.The drill-down might be a bit too quick however. I chose a wallet and a few clocks, and now I’m getting lots of recommendations for more wallets and clocks. Do we generally use recommender systems to discover more products of the same category? Or to discover similarly designed products in different categories?Another lesson: photographs are so damn important. If you don’t have a good product photograph, then no one will buy it. What kind of new tech can improve product photography?

  21. Dan Abdinoor

    I think the recommendation engine still has a ways to go. I tried it out and I’d say it was maybe 30-40% accurate. Cool idea, though, and not an easy problem to solve.

  22. Kevin

    I started out clicking “Stuff for Men” and after a few screens I got eye shadow. Seems like a point of negative training data, at least.

    1. Kelley Boyd @msksboyd

      I actually switched to the stuff for men after too many polished rock and wire things. I like organizational tools etc. But that is just me. The first screen in the men stuff had a dress? Maybe they should have put the black lace bunny ears there as well!?!

  23. Jeremy Black

    I actually happened to give this a go last night. I made my first Etsy purchase about a week ago. It was a travel bag for my wife for Christmas. The bag was here within a week and the quality is fantastic.With respect to the Taste Test, It was fairly good at picking aesthetically what I would be interested in, however, the inability to categorize the discovery is missing. That’s obvious and would present more complexity than was intended so I understand why it has been left out.Overall, a great way for Etsy to experiment with creating a “taste map” for users. In the end it’s fun and that’s what Etsy does best.

  24. Meghan

    I’m going to go ahead and say that I think this is a good thing. As we all know, Etsy has been wildly successful thus far, even with their discovery problem. I guess they’ve learned from facebook that people love to procrastinate via browsing, and when you add in that community element — Etsy’s ten times more intriguing than any other online shopping experience.Of course, for some the discovery issue is a deal breaker — they don’t know where to even begin, they search for something, don’t like their first few hits, and go to the flea market instead. Having an interactive quiz is a cute way to try to capture these people, even if the results aren’t perfect (I actually thought it was better than Google’s Boutique, which got me all wrong and was even longer). People love quizes — look at eharmony! teen magazine! and the idea that selections are custom // tailored to them specifically. Will everyone do this? Probably not. Is it a nice way for a new user to get acclimated to Etsy? I think so.Also agree with Tereza’s comments re: anniversary gifts, kids bdays, etc. People should be able to enter that info into their profile and subscribe to newsletters that address special occasions that require a gift. Also after new user makes their first post, it would great to get an email w/ other recommended sellers based on your first purchase. I think Etsy can pull this off without being annoying, bc they’re ultimately helping the underdog by making the recs/introductions, which is different than Saks wanting me to buy something new where they get all the $$$.

  25. lance

    I played with this and perhaps more importantly had my 12 year old daughter play with it yesterday. She loved it.

  26. JeremiahKane

    Interesting tool but the results aren’t very relevant. Especially given how much there is on the site that I am interested in.Etsy’s discovery problem/search challenge is especially hard since by their nature, the items are unique and transitory. If a book sells on Amazon you have new information to apply to the other copies still available (other books the buyer purchased etc.). With Etsy you now know something about the item, but the item doesn’t exist anymore (some sellers have multiple numbers of an item but most are unique). Tag abuse is also a problem.For search and discovery to be successful one of two things will have to happen.1) Gather good/verified information about item attributes. I believe this is very difficult and the wrong approach- tags are easily abused and other information gathering is intrusive.2) Re-focus search on the people and shops behind the items not on the items themselves. The attributes of the artists who create the items should be the main source of inheritable item information. This happens naturally with Etsy when you talk about the site with friends – “I found this amazing artist who makes xxx.” the focus is on the artist. A shop is a lot more analogous to a web page, it’s permanent, indexable, rank-able and searchable. I think a focus on the seller would also strengthen the discovery process.

  27. kris

    The recommendations I get include quite a lot of cats because one of the choices I had to make was between a cat and a few other things I didn’t like. Had I been given the choice I wouldn’t have selected the cat. That way I wouldn’t get cat recommendations.I’m being shown a lot of other animals as well. A duck, a cow, an ugly frog, a meerkat. I think that’s wrong. I didn’t want the cat, nor do I want to see other animals. I would never buy photographs or products with an animal print on it.I selected a few paper sculptures as well and this gave me a several good recommendations but that’s something I know I like, so if I needed it I would do a search for it. But I’m a guy. I can see why this could be useful when I have no idea what gift to buy for someone, but still, it’s a lot easier to just browse around on the site. I don’t like to do the work of having to click through all kinds of products I’m not interested in.It’s a tool to discover lots of products but it ain’t for me. Sorry.

  28. Donna Brewington White

    To my utter surprise, the Etsy Taste Test pretty much nailed me — at least the me I’d like to be if I could spend more of my time and money shopping for me.I like that it took me less than 5 minutes.I saw stuff I loved, but am not an impulse buyer so wasn’t motivated to buy anything in the moment. There was no easy way to save the information for later reference for when I AM ready to buy. I only want to go through this exercise once. Unless I copied and saved the URL, I’d have to start over the next time. I would feel frustrated — like backtracking. I did tweet the results so I could get the URL later from Twitter, but that’s not ideal. I would like to be able to have the URL emailed to me or saved on the site for me to access later.They are probably going to get more of my business for gift purchases. Maximizing the use of this as a gift purchasing feature might be highly beneficial for someone like me. An easy way to send this to someone (gift recipient) with the results coming back to me would be helpful.The one way that this does work for a shopper like me is that it gives me some of the benefits of boutique shopping. Once I find a store or shop that suits my tastes, I tend to do all my shopping in those few select places — mostly online. I barely have time to shop with even less time to shop for shops. This feature will make me much more inclined to shop via Etsy — if I don’t have to start from scratch each time in setting preferences.Overall, I think they are on to something with this — but this feature in itself is not enough.

    1. Gayle

      I spent some time on etsy for the first time this Xmas and bought one item. I found that I have little patience for searching endless pages of products much less all the stores…and I am very much a discovery and impulse buyer (but with caveat).Also spent time on ArtFire and loved how I could save an item on my Amazon wishlist to come back to later. The “discovery problem” is an issue here too though for whatever reason, I found myself more williing to click on random sites in the featured section on the front page as well as refine my searches using their categories.

  29. Susan

    I played with taste test last night and really enjoyed it. When I first discovered Etsy in 2007 I was fascinated by the fun ways to explore. I’m thrilled to see these options grow. Shopping can be fun, and these nifty tools really help.

  30. Paula_G

    Well, I gave it a whirl and it pegged some very nice Japanese-styled nature print items as well as brightly colored glass, but I also clicked on a vintage camera in my preferences and maybe it threw off the algorithm? I’m not sure if that’s what made the recommendations include a skull ring and a primitive, melting head mask.As far as the Taste Test — Nice idea, but it still needs work.And I agree with Tereza 100%. This stuff is fun in an exploratory “stumbleupon” kind of way, but not really useful in a “lifehacker,” make my life easier sort of fashion. If I’m really pressed for time and overwhelmed with options, which method do you think I’d prefer?I really like her idea about the anniversary, birthday, etc. emails. Now that’s useful! In the age of social, email is so under-appreciated. I mean, come on, we’re not all 18-24.

  31. kgu87

    Half-baked idea, at best.I’d suggest putting more science in the backend before calling this “a set of recommendations”.

  32. Dinesh Vadhia

    I’ll assume that the Etsy Taste or Discovery system is based on collaborative filtering (CF). If not, then ignore comment. If so, problems with CF are well-known particularly as the data size grows ie. CF scales poorly. CF is also a batch system where the predictions are calculated offline which makes it impossible to offer dynamic suggestions. Most orgs who use CF end up with mediocre suggestions. It is a technology issue.

  33. panterosa,

    I am a shopper, a designer who makes things, a stylist, and a potential Etsy vendor. I have several takes on this new idea, and I am glad to have read this post. I have a specific question to ask after my comments.As a shopper, I took the test and felt it did very well for me, though I did have to search thru after doing some refresh all, and I felt the test got ‘stuck’ when it floundered, and did well when it was successful. But I liked my results.As a visual decoder/stylist, I help people match what they already have and find new to match to what they own, I can tell you that decoding what people’s eyes like is tricky. I do this in the fields of interiors, gifts, graphics/packaging, and color, and I am good decoding because I understand all the visual styles and the taste which underpins them. Making sense of people’s taste, even making their taste cohesive, takes this knowledge base and as a person I have a better rate of success than an algorhythm. My work is akin to branding, building a visual word cloud.Taste can be a minefield when people perceive that edge of superiority. Many people’s taste is eclectic, or mismatched, or not developed. Those who have the money often hire someone else to give them taste for the very reason that wading through can be so confusing. It is like a self-discovery/self-definition game/challenge. For this reason black and navy and beige rule the main spectrum of products, and I am hired to help people get out of their post-Navy Moment, which resulted in pandemonium. .I can see those with developed taste doing better in the taste map test. The suggestions to have the multiple profiles saved is a great one, and, as one user did, have his daughter go on, helps build that profile with the ‘target’s input. Perhaps a personal shopping service, and/or questionnaire, would help refine even more. A wish list helps too. I can see doing a search if I am looking for something for a client. I already sent my client something from my results. I have not used Etsy as a source in this way before due to the time needed to cull through the ever growing products.As a designer, my own shop on Etsy has not been built due to the effort I would need to make compared to the return needed to justify the time it would take. It is not my primary business, making things to sell on Etsy, so I simply sell directly to those who like what I make, and at a small scale this works. To grow would require making a time investment.This will be of more interest to me in the coming year having just quit my day job yesterday. I thought of the talk you gave about the 3 highly addictive things – heroin, carbs, and a monthly paycheck.Specifically, Fred, what metric does Etsy seek to hit – what kind of sales increase they hope to gain from this effort? That would help define how deep Etsy will go to solve the task of wading thru the ever expanding ‘product line’ they offer now thru their success in offering this service. As a potential vendor it would be good to know. Having the vendors tag their merch better, alongside this effort of Taste Mapping, would be of great value to the vendor and the customer. Will they address this in newsletters?

  34. eromer

    I found it a little hard to find stuff I liked, having finally done so I was not at all impressed with the taste algorithm. Most of what I liked in the recommendations were the actual items I had already picked, and the other stuff not so much….Back to the drawing board I think.

  35. paramendra

    Great move on Etsy’s part.

  36. moumou

    I agree with Tereza. It can be too hard to search, unless you know in advance what you’re looking for (and sometimes even then). I DON’T HAVE TIME for lengthy searches. If etsy understands my likes and needs and can make suggestions, great. I’ve often bought on etsy after some other entity has done the weeding out, like I also often use etsy for gifts. If they can simplify the search, I might stay on long enough to complete the purchase.

  37. Roger Ellman

    I love the ideaThe whole prospect of recommendationsI think the reality is that to get to have a hazy (fuzzy?) idea of what one person wants you also need to know what they don’t want/like – I can see that consciously rejecting/sidestepping items would add as much to refining the formula of what I would like as choosing/approving some itemsHappy to say more as I have created several websites/systems/trainings by git feeling/instinct that have worked out wonderfullyBest wishesand Happy Christmasfollowed by a profitable and rewarding, joy-filled New YearRoger

  38. Ghorwood

    This isn’t very satisfying. A simple enhancement would be to treat each item like a Facebook treats ads and allow users to rate/demonstrate a value judgment. Something like this:You have removed this item. Why didn’t you like it?UninterestingRepetitiveNot Unique to EtsyOtherType your reason here…If you did something like this you could leverage the information to “merchandise” and create collections that would be more appealing. You could also make the information available to sellers and help them zero in on the items people really want.There are other things Etsy could do such as focus on a constituency that isn’t represented at all. That would be the gift receiver. While not every gift receiver would come to the site to interact after getting the gift those that did shape a gift guide from a completely new perspective. Also you would be further linking the Etsy name to the gift itself. “This gift was found on” Tell us what you think about it.”

    1. Kelley Boyd @msksboyd

      In theory great idea Gail, but why would a person spend time doing that to benefit the seller if I am the buyer? Always looking for the WIIFM. Massive time suck.

  39. ericstern

    Ok, I went through it twice. Having an Etsy store myself, I am happy to see Etsy trying new things. Some quick feedback:I like the minimalistic design, but it would be a lot more useful if there was some customization, for example if I could choose a few categories or something like that. I have zero interest in most of the items being presented – if I could choose to see only vintage clothing, housewares, and accessories and then go from there it would be a lot better / more relevant for me.It took a long time to find anything I really liked and I found it pretty frustrating (especially cause it kept at 83% done for several selections). In general, I find these active selection recommendation engines very tiring. I much prefer getting recommendations based on whatever I am already doing on the site/store, like what songs I listen to, what purchases I make and how I rate them, and what ratings I give on Netflix movies (I usually only rate them as I see them, not through the longer rating process).So what I would really like are recommendations based on Etsy items I have already purchased, favorited, or based on the types of items I’ve been looking at recently. I’m sure Etsy is already working on these types of tools, and I’ll be happy to use them – however, I’ll probably pass on the Taste Test for now. But thank you Fred for bringing it to our attention.PS – as an Etsy seller, it would be really interesting to see recommendations based on items I have sold – there are so many sellers on Etsy that I bet a lot of people would find this useful and end up making purchases as a result.


      Seconded. I think it would be best if Etsy recommended based on previously purchased or rated listings. Maybe also based on favorited and viewed listings. The simple act of viewing something says a lot about taste.


    I think that this is a great step in the right direction. However, I do believe that there is a way to do an entirely automated recommendation system that is very accurate. I also believe that in order to be successful it must be automatic to reach regulars and not just Etsy addicts.I do have some ideas about how a fully automatic method could be achieved through some non-obvious methods. Its on my roadmap to try out maybe next year. The primary difficulty in my eyes (why standard methods are hard to map to the problem space) is because most every item is unique. So, you can’t for example say, somebody who liked iPads would also like these iPad accessories or even an iPhone automatically. Unique products makes the problem space a bit troublesome, however not impossible. And actually I believe it is possible to get a very high percentage (maybe 50%) success rate in automatic suggestions. I’ll post more about this later after I try a few of my ideas out, for now though you’ll either have to e-mail me directly or wait. I don’t usually prefer posting in a public forum about ideas that are not fully materialized.In other notes, I couldn’t comment on this blog w/ Safari.

  41. Dave Snigier

    I think this is the first time that I’ve seen many things on Etsy that match my taste and haven’t been directed there from a blog or tweet first. I really like fractals and offbeat colors which the recommendations reflected.Amazon’s recommendation engine for art? Count me in, so long as things don’t become so mechanical that we miss out on the human elements of discovery and never are exposed to pieces outside our comfort zones. Finding the balance between the two extremes would be something of immense value.

  42. Steven Kane

    great ideabut suggestions are a bit too linearchance and surprise need to be factored in, maybe. a little more unpredictable noise, a little less predictable signalimho, shoppers get delighted discovering things — and discovering their own tastes that they didn’t really know they had — as much as by having their known tastes reinforced

  43. robka

    We do something like this on and use it to match brands to research panels. I’ve also seen a good consumer oriented adaptive conjoint / discrete choice analysis on I think to make these work well, you need to tradoff between something that the consumer really likes and that polarizes them. This is tough to do with taste. We have some tricks, but rarely do you get a polarizing effect in most product categories. I think this would work better with clothing and then using the clothing taste to match people to each other and using those segments to drive clusters of taste. This uses accessories and home furnishings which I don’t think really polarize people enough to make them connect with the consumer’s taste. It’s good, but it’s not up to the level of or some of the more refined versions of this that I’ve seen.

  44. PhotoHand

    I second Tereza’s opinion. This feature will be a nuisance. Women shop for gifts more often then they shop for themselves and they go out of their way to find an item that would reflect the receiving person’s lifestyle, interests and needs.

  45. Mark Essel

    Local everywhere, I like it. Many news startups are gunning for that architecture for routing.

  46. Dave Pinsen

    What would be cool would be something like Etsy, but for manufactured products produced by local SMEs. For example, elsewhere in this thread I linked to the Kickstarter page for the designer who sells watch bands to hold iPad Nanos. He says that they’re manufactured in the same factories where Nike makes its watches (presumably, in China), and that the straps are made of “high grade silicone rubber”. Why not a stainless steel link band, and why can’t a small company in Lancaster manufacture it? Apple fans don’t seem to be penny pinchers when it comes to accessories; why not offer a higher-quality band that justifies a higher price, so that you can manufacture it in the U.S.?If Etsy could facilitate something like that, it would be performing a valuable service for the U.S. economy and for American workers.

  47. ShanaC

    Something to think about:'s not an issue of space. It’s cheap labor doing rote tasks versus same labor doing innovative tasks. Apparently it is true for manufacturing as well.Probably one of the best podcasts I have ever heard about the subject.

  48. JanC

    If you live in NYC, then buy local is very much supported by Goodzer NYC (

  49. Andrew Greene

    I love the idea, but I’m not sure how it could work. For such a market to work, you need volume. Massive volume because margins will be low. But you also need massive investment for facilities to respond to customer needs. Low margins and high capex is not attractive to investors.So the question is whether American manufacturers have the facilities for small jobs and rapid prototyping. Maybe we already have these machines going unused?For the product I’m most familiar with, bicycles, the answer is no.I suppose it’s possible that my assumption of low margins is wrong. Custom products would be selling point. “Designed by Me” or “Made for Me” could be a very attractive marketing slogan.

  50. Andrew Greene

    Charlie, can you get in touch with me?my first name @speedwellracing.comThanks!

  51. Dave Pinsen

    High quality products can have high margins, which in turn enable manufacturers to pay decent wages. That’s proven to be a viable path for first world manufacturers in Europe, Japan, and even in the U.S.