A graphical analysis of women's tops sold on Goodwill's website | Hacker News Reader

A graphical analysis of women's tops sold on Goodwill's website

goodwill.awardwinninghuman.com

394 pointsjjmccoolguy84 comments

> There's still a high number of items from various in-house brands from Florida department store Bealls (namely Coral Bay, Reel Legends, and Dept 222)

Looking at the chart, it appears these 3 brands were all in the top 4 most commonly sold (or at least available for sale) brands at Goodwill in by 2019, and they've been in the top 10 or so for years. I can't imagine organic amounts of resale from a Florida-only department store chain could account for this. So is Bealls just straight selling their old stock to Goodwill for cheap? Or maybe even giving it away as a write-off (and to reduce cost to store items they don't think they'll ever be able to sell). I wonder how much of Goodwill's stock actually comes from things like this, as opposed to houseful donations.

>I wonder how much of Goodwill's stock actually comes from things like this, as opposed to houseful donations.

Anecdata. I worked at a goodwill in the early 2000s for community service. A lot of people donate soiled rubbish disguised as clothing that goes to landfill. I would have no doubt that Goodwill has partnerships with department stores...or else they wouldn't have anything to sell.

Most people do not.

There's some (easy) analysis to be done to look at if these brands are coming only from only one or two Goodwill sellers, which could help us better form a theory. I might take a look after work. Even then, I didn't want to go too far down that particular rabbit hole since understanding might (shudder) require me to get off the internet and call someone or something.

I liked the visualizations a lot by the way. Maybe you mentioned in the write-up and I missed it, but what did you use to make them? Checked the skills listed on your CV but didn't see any not-static visualization tools listed

Thank you! I used d3.js, my first love.

I just went to Goodwill last weekend. They seemed to have more “new” items than I remembered. I assume it’s cheap overstock that’s donated vs marked down.

Had limited experience of this while doing strategy work for a large UK fasion retailer.

A significant proportion of items from a new range had been returned as faulty. When they investigated they realised a new factory they were using for this range had slipped a huge amount of bad items through the reatilers QA process. They ended up writing down the whole line and donating it to Oxfam/Redcross to be sold in their charity shops, very similar to Goodwill. It was 100,000's of items.

They were already writing the goods down as a loss against their balance sheet and they managed to recoup a small percentage of that loss as a tax deduction for the charity donation.

It didn't happen often, but it wasn't the first time they'd done it

FWIW this is also a relaitvley common practice for administrators when dealing with bankrupt companies. Donating goods they can't sell is often cheaper than storing them or paying for them to be disposed of!

Being wrote off an entire 787 Dreamliner and donated it to the Pima Air And Space museum. It was going to cost to much to fix and have recertified after it went through testing, so they gave the whole thing away.

Possibly the second largest donated item ever, after the aircraft carrier Russia "donated" to India (I think it was in need of $2B worth of repairs...)

The family that started the Patagonia brand donated 1600 mi^2 of land to Chile to form a national park. That's way bigger than a boat(and probably cheaper to maintain too)!

The first 3 flight test aircraft were donated:

ZA001 Went to Nagoya, Japan.

ZA002 Went to Pima Air and Space Museum (as you noted).

ZA003 Went to the Museum of Flight in Seattle.

As for the other three:

ZA004 may still be with Boeing.

ZA005 was scrapped.

ZA006 appears to have been sold to Mexico for a VIP transport.

If you try to return a mattress you bought online this happens too. They'll have a truck from the Salvation Army stop by. Small write off I suppose. I "returned" an Amazon mattress but my wife didn't like it, and because of COVID they just refunded the money and told me I could keep it. I don't know what I'm gonna do with this extra king sized mattress though, guess I'll make a rather luxurious guest bedroom.

I wonder if any of the price increase OP noticed is from liquidation retailers (both brick and mortar and online) buying and reselling a greater fraction of these runs of "faulty" items since that lets them recapture a greater percentage of what would be a loss.

Goodwill does alot of deals with retailers for excess inventory and reshopped returns. Target used to use them, for example.

This is what TJ Maxx/Marshalls used to do. The market is pretty diverse for this type of thing. There's a whole industry around it.

Fun hypothesis: a giant Florida Goodwill moved into an old Bealls in 2019. Bealls may have run a cost-benefit analysis and said "you know what, it's actually cheaper for us to sell all this old merch to Goodwill at a discount than for us to move it to the new store."

https://www.willmeng.com/goodwill-store-moving-to-shopping-c...

Bealls is based in Florida, it is not Florida-only.

Walk into a goodwill, and your answer will be obvious. Plenty of things are sold packaged and new in goodwill. Usually it's things like homegoods, in my experience.

Yes, stores often sell excess stock to goodwill.

I am not versed into fashion or thrift stores, but the author managed to make it interesting. I appreciate the attention to details, the quality of visualisations and transparency about data acquisition (isn't scrapping 50% of data science?). Good read, thanks!

I want to applaud the author for loudly calling out the likelihood that their own data is bad and the reasons for that. You hardly ever see articles like this mention that, and if they do it's usually a quick aside.

I suspect this has something to do with the author's intentions. The author here has no reward attached to people believing her conclusions drawn from this data. Instead she may wish to just show people "look at this dumb cool thing I made" or she is using this to pitch her skills at potential recruiters, in which case honesty is a good policy to filter for good employers.

For scientists and commercial interests, the quality of the data could be fundamental to the point they are trying to make. So admitting their data sucks would basically ruin their whole argument, or at least make people more skeptical about the conclusions drawn. In science, the bad data eventually gets called out and everyone else is left wondering why the miracle panacea for discovering the genetic basis of complex disease still hasn't solved schizophrenia.

For those who don't know, reading the product descriptions on shopgoodwill.com listings can be downright comical. They're written by random employees of random Goodwill stores, of items which they may not understand or have a very pointed opinion about. If you're really bored it can be a goldmine of mild entertainment.

It's sad that they revamped their website. It used to look like a 1990's e-Cart website, which was so wonderfully functional and compact. Back to 2010's "endlessly scrolling through giant type faces and no content" design...

No kidding, from the frontpage:

HOLY CANNOLI !! 100 Charms 925 Ultimate Bracelet

THIS IS THE FINAL BRACELET YOU HAVE TO FACE AFTER DEFEATING ALL OTHER PUNY BRACELETS - "BOSS BRACELET"

I got curious about the redesign, and some timeline scrubbing found that the switchover happened Aug 30 2017. Here's the 29th: http://web.archive.org/web/20170829001028/https://www.shopgo...

Why is the title “weirdly detailed”? Apparel, especially women’s apparel, is a massive market segment. Even if they’re only talking about thrift stores it’s an important economic subject.

I would have said it's impressively detailed.

This is describing what I imagine a few savvy bargain hunters know from experience and interest in the market. Being able to see massively undervalued items that don't fit in the catalogue, kind of like being able to pick horses at the races.

Titled like clickbait. "You won't believe what analysis of Goodwill women's tops sales reveals"

This article is very much not clickbait.

I think it’s a bit of clickbait. My very initial impression was that this was going to be a strange sexualized analysis of how women of varying body types would look in tops sold by Goodwill. That was followed by “but that’s insane and this is on HN, it must be charts and graphs, hmm, better to click to confirm.”

Agreed on clickbait, I thought it would be some near pervy analysis of how the tops made some guy feel..

That sounds like a better article.

... so you clicked on it ...

"for research purposes"

It wasn't the original title, but this is explained in the closing sentence:

> Hopefully, though, they'll see this content as it's intended to be seen: As a very weird love letter to thrifting from a very weird person.

So, the author is admitting they are weird for doing such an unsolicited analysis. But weird in a good way.

I do love all this conjecture on the titling (and your comment that women's clothing is an important subject helps me feel a bit validated in my time spent on this). I categorize it as weird because it's quite niche, there's no actual call to action or news story, and I spent waaaaay too much time on it.

One of the reasons I like Hacker News is seeing all the people who will commit to days/weeks/months of work just to satisfy their curiosity. You’re in good company, this isn’t weird by HN standards.

fwiw, i think you can drop the '.com' from the title for a smoother punny title without ambiguity: "Goodwill.com Hunting" -> "Goodwill Hunting" since the movie is "Good Will Hunting". =)

Ha! I thought as much.

Women’s apparel is a 600B industry. It is highly segmented, and literal armies of people analyze category and product performance across millions of skus. So spending a lot of time on something like this is definitely not unusual.

It's really fun to see the moment people realize things like this exist outside their bubble of knowledge.

I had some first and second-hand knowledge around pricing household appliances. There's a common assumption that "the manufacturer says 'charge X'", the retailer charges X+Y% and you all go home happy. And you could not be further from the truth if you tried: pricing is an insanely complicated process with negotiation from both sides and monitored by an army of secret shoppers to keep both sides honest.

I think it means 'weirdly' as in 'obsessively', or 'nerding out over'.

Something can be both a not normal thing to do and a wonderful or useful thing to do.

Looks like a moderator took the weird details out of the title above. Submitted title was "Show HN: A weirdly detailed graphical analysis of women's tops sold by Goodwill".

We took Show HN out too, because while this looks like an excellent submission, it's not a product or project that people can try out - see https://news.ycombinator.com/showhn.html.

In the "Goodwill Tops by State Over Time" blocks there's a state called Michegan right above Michigan. Made me look up to see if it was called that at any point in the past, but it seems to be a typo :)

Ahhhhhhhh, sorry mate, good catch. I'll fix after work. I'm Canadian if it gets me a little off the hook.

Didn't Macklemore's 'Thrift Shop' come out in 2016? Maybe this explains the price increase in 2016

more like 2012

Ah yeah you're right

The days are long but the years, short.

I liked your theory while it lasted.

I can't make sense of ‘Price relative to state average’. That's relative inside each state, right? But how can things in Carolina, Florida, Texas and West Virginia always cost more than the average?

Meanwhile, ‘Price relative to overall average’ is all white due to one point in 2019 in Missouri at 245 bucks. But how many of those 245-bucks items were sold that they aren't absorbed by the overall average?

Also, since Pennsylvania apparently dominates 2018-19, perhaps it skews overall data considerably.

You make some good points, I made some bad choices. To address: > Yes, I meant relative inside each state but phrased it poorly > With median price, I probably should have limited it to periods and states selling more than x items > And, yeah, Pennsylvania, and in particular the seller labelled "Goodwill Industries of North Central PA, Inc." dominates the market and should maybe have been excluded. I thought I'd get away with this by doing inner-state comparisons but it's still unideal

My thought after reading several paragraphs is that the author may have chosen women's tops for analysis because that's the group of items where the greatest depth of data exists across all of Goodwill's categorized SKUs.

The article begins with "After 10ish years of second-hand shopping, I've started to ask myself a lot of questions about the clothes I've been buying..." but never says that the author buys this sort of item in particular.

Based on my bunch of visits to local second-hand shops (not the US), women's clothing overshadows the supply for men, so much so that those are basically ‘women's second hand clothing stores’. And I'd easily expect tops to dominate the selection.

Consider yourself lucky that I cut the over-long biographical introduction about my views on shopping and the effect of going back to school on my budget. I can tell you right now: neither funny nor interesting.

But for your curiosity, I chose women's tops because it's an item I buy, it represents a good portion of Goodwill sales (though I don't know how much), and it gives some consistent area for comparison more than if I was looking at, say, everything from old TVs to ceramic knick knacks.

I'm pretty thrilled you read several paragraphs though, haha.

I like they you have a section about data quality, but because this is only of things listed on their website there's going to be a lot of underlying bias of what was chosen to be listed on the website. There might be some kind of company policy that is driving the big shifts in # listings, brands that are considered good enough to put online, and price.

Just fixed the typo on "disproportionately" that was pointed out, and the duplication/misspelling of Michigan as "Michigan."

Will probably not address the other suggestions tonight and just chill.

Thanks everyone for your interest! I honestly thought this would die in "new."

> This project was done without Goodwill's assistance or permission.

May I ask about this - why no permission? Did you attempt to reach out to Goodwill to explain the kind of project you're embarking on and just didn't receive a response?

Is the graph at the bottom showing volume of sales increasing, or lack of data before 2018? I didn’t entirely understand the comments about gaps in data, what specifically is being referred to (e.g. the big dip in 2019, or the noise, or the 5x jump from 2017 to 2018?). While reading, I was wondering what the volume of sales were and if that explained price increases. If the volume really increased more than 10x, it’d be surprising if prices didn’t go up even more than they did, right? But I guess volume didn’t jump this big this fast?

It might also be nice to adjust for inflation over the last decade, which hasn’t been huge, but it makes a little difference in the price curves over time.

Gaps might be due to the fact that scraping requires fiddling with the code every time the site changes even invisibly—or data just stops coming in.

Can't answer conclusively. The big dip and the fall-off at the end were probably errors in my scraping. The site went down for a bit at a couple points, and the way dates were formatted changed a bit, all of which I thought I handled correctly but maybe not. And at a certain point of combing back over gaps, I just decided to be done.

I strongly get the impression that sales volumes did increase from 2014 onward, but sales in 2014, particularly in the early range there, probably appear lower than they are. IDs in that range sometimes returned normal Goodwill item pages, and sometimes returned 404-type pages. Maybe they migrated systems or something around that time?

Quite a nice job on visualizations!

This looks like all d3 or was there anything else you used?

Agreed. Those are awesome visualizations.

According to the Chrome console output they all appear to be D3 visualizations.

The way the layout uses the right margin for legends or auxillary information, or how the charts sometimes bleed into that space looks straight out of a Tufte book. :thumbsup:

Aw shucks! It is indeed d3, and, as per the commenter below, uses very very lightly adapted tufte.css for layout.

This was weirdly fascinating for a topic I've never been curious about.

Neat! This may be a dumb question, but OP, have you been scraping price data since 2014, or are all those old listings still around on the site?

Sorry, I was just vague. I scraped data for old listings around the site which seemed to go back to around 2014ish. Most data I collected last summer and over Christmas, and then it sat around for a long time.

The "Number of Women's Tops Over Time" graph may be partially showing the Marie Kondo effect.

shopgoodwill has a weird way of doing shipping and handling. I'm assuming that is not included in these results, because I didn't see it mentioned.

The last time I used the site many of the stores would set high handling prices, which basically acted as a starting bid. I did notice that some stores started to lower their handling price, if that caught on it could account for some of the changes. The handling price is directly related to how much people are willing to bid on a price. Shipping is also be a factor, but would be harder to gather the data as it is based on the destination.

In the "Goodwill Tops by State Over Time" section, it seems like Missouri is skewing the scale when I select "Price relative to overall average".

It's a shame I was really interested in that particular viz!

My wife gets nearly all of her clothes and our kids clothes from Salvation Army. There seems to always be a good selection for that. Men's clothing is much harder to find second hand in my experience.

Angle I didn't see in your article: the price increases are a result of flipping finding true market value for an item. The ebay-et.al. effect.

Why are USED t-shirts selling at 8 dollars wtf??

Because all the super cheap stuff goes to Africa and is then sold by the bale to small resellers. These small resellers then sell every shirt for the same price. Then local women buy the shirts they like and sell them at a markup.

Please do not donate to GoodWill. Altho their mission is good, its all hard core capitalism from there. Just Google "Goodwill owner house" to see his multi million dollar mansions all over USA. He was caught giving himself $250,000 raise while battling employees for disability claims worth pennies to the company. If you have clothes to donate, meet any homeless person they will get most from you or tell you where to go to give it all out to other homeless people.

> If you have clothes to donate, meet any homeless person they will get most from you or tell you where to go to give it all out to other homeless people.

That sounds like an excellent suggestion. And even if they can't use any, the kindness will do you both good.

https://www.snopes.com/fact-check/executive-salaries-chariti...

So the guy made $750,000. Whats your point ??

Top notch visualizations!

The (very attractive) charts do not display without JavaScript enabled. Surely this is a textbook example of where graceful degradation would come in handy, as a simple image without interactivity is still useful.

Also the annotations, which are simple text, do not show up without JavaScript. This is even worse: if there is anything HTML is capable of doing, it is putting text on a screen!

Prices have gone up? Tell the federal reserve, I'm sure they will be shocked /s

I don't think quality has anything to do with it. Employees are more expensive because life is more expensive.

The price of clothing has gotten significantly cheaper since the 80s.

Reply on news.ycombinator.com