Our data shows "Amazon Choice" is essentially a join table between product and search keywords, ordered by a multiplier of conversions, ratings, and availability.
- Customers are happy when they get stuff fast - it needs to be available right now, and in the buy box
- Higher rated items are theoretically better than lower rated items
- "choice" badge is keyword specific: a plant light could be choice for "grow light bulb" and "grow light led" but not for "light bulb".
Our data shows the items are not correlated by margins for. We often see negative NET PPM on "choice" items.
Source: I run an Amazon consultancy that manages roughly $30M/year sold through Amazon.
So this would largely imply that Amazon's Choice truly is about giving customers the best experience (to the extent possible given the quality of keyword and review data)?
Basically, labeling products the customer is most likely to purchase.
but, pessimistically or realistically, I expect they are also training customers to blindly trust the black box and eventually the criteria will change towards maximizing short-term value.
Yes, also in my experience "Amazon's Choice" does not protect you from counterfeit, grey market or any other crap items.
I wrote to jeff (at) amazon (dot) com last week about this, because I got super upset. And some middle manager replied. I shipped US -> UK some underwear from a company that uses no latex (I'm allergic), sends items inside a ziplocked sealed bag and prohibits underwear returns (for hygiene reasons).
After waiting for the parcel for 2 days at home, as I was in a hurry to get more latex free underwear, I got an Amazon envelope and the underwear came inside without any packaging and clearly used (!). I was furious.
Not an isolated event. A few weeks ago, I ordered a shower with a chlorine filter from Amazon UK. I know this korean brand very well. Again, I got the item without any original packaging, and water inside (!).
So I don't trust Amazon anymore for items that are more than $10. People abuse the return policy, and Amazon does not care. Or it's even their business model: Ship it again, selling it as new. Online shopping is getting very tricky.
I'd gladly pay more at places that guarantee brand new items. Things are getting really scammy. Around 1/5 of the items I buy online have been opened and used, or have some defects. And I'm talking about expensive stuff and supposedly reputable sellers.
The jeff@ experience is really poor now. There doesn't seem to be a way to request real review by someone who might care about something other than ticket throughput. I tried it when I got a ding against my account because my wife and I had ordered something that in total was over a limit. I was accused of misuse of accounts. Ended up move all my personal data out of Amazon because of it.
It's been bad for years. I've heard from people who have gotten emails back that said that Amazon is actively pursuing a partnership with a company that they have never contacted (mine). Other people say they just get boilerplate responses that seem to be written by bad AI or non-native speakers who didn't understand the customer email.
I use amazon as much as I can, but I would never trust their supply chain with my life or my health. I don't think I would ever buy such an item on their website, nor anything I would apply to my skin or ingest. Even if they displayed a model suitable for climbing, how would you know it is guenine?
The counterfeit problem is real. Like half of their intel 10gbe NIC are counterfeit (I know because a Synology NAS will only work with a guenine model).
Looks like there is an inflection point where adversarial activities become more profitable than genuine activity. I wonder if this has happened for Amazon. In this case, I would expect counterfeiters would grow exponentially. That's just the laws of economics.
That’s nasty. They do point out in the article at least two examples where knockoff fashion items were badges instead of the real deal, so it seems somewhat pervasive.
Yes, I think their warehouse policies are totally crazy.
But that has a little advantage. When they send you used items, it is easy to notice.
If you buy at some reputable shops, they know how to scam you and resell used items as brand new, not as refurbished, grade A or whatever, which is illegal at least in EU.
For example, I wrote to the customer service of a famous fountain pen shop in the UK asking about return policies on a very particular model which comes sealed with stickers. I was shocked when they said I could not only open the box and inspect the pen, but also ink it and use it. If after a week I decide I no longer like it, I should just send it back. It will get sold as brand new. They only advised me to open the stickers really carefully to facilitate their (scam) task.
In fact, if you shop around their website, you'd notice there are practically no refurbished or open box items in proportion to their high volume of sales.
If you ask any fountain pen enthusiast, a pen that has been used for a week is not new. Just like a car that has 2000 km in the dial. Good shops even sell models have been dipped in ink, not even filled, at a heavy discount.
Cult Pens. The problem is that most manufacturers collaborate by not plastic wrapping and sealing pens.
For example, no Pilot is sealed except the 823. I don't think any brands other than MB, TWSBI and Dupont make it obvious to notice if a pen has been opened.
Same with good mechanical keyboards. E.g. all Leopolds are sent to EU and US retailers without stickers.
If they were stupid enough to put that in in an email, forwarding to Trading Standards might get them leaned on. Though I admit these days it's a long shot.
Trading Standards used to be really great about responding to these sorts of complaints, and for larger sites and retailers they'd have their own test shopping as well. Tory austerity means they haven't had budget for years so are almost completely toothless. Only when Brexit stops being the only issue driving UK politics will the systemic destruction of the successful parts of the UK ever get noticed again.
> Around 1/5 of the items I buy online have been opened and used, or have some defects.
yeah. for some items, i've tried to avoid this phenomenon by ordering from, say, homedepot.com, or Lowes.com, but then it seems like they sometimes just send me items with damaged or blemished boxes, or opened product, which they couldn't sell in the brick and mortar retail stores.
Don’t see why that’s a bad thing, the less time I spend browsing amazon and reading reviews to find a suitable product the better. As long as people are consistently happy with the things they get they only have reason to rejoice.
Could I get something cheaper if I spend time looking for it? Maybe, but the time I spend looking is also money down the drain, for what usually amounts to savings of a couple dollars and cents. Rarely worth it.
I see two problems. First, the Amazon rating system is untrustworthy. There is simply far too much counterfeit reviewing and far too little done about it. This is a systemic problem that just needs to be fixed for the platform as a whole.
The more nuanced problem, I think, is the fact that it can create a feedback loop. Something can be popular simply because it's cheap, but "Amazon's Choice" implies quality, which makes people think they're getting a deal because it's also cheap. What sprung to mind for me was LockpickingLawyer's[1] series on Amazon's Choice for things like "padlock", lock, and safe, all of which were of fairly low quality, or at least lower quality than "Amazon's Choice" would suggest.
The problem is that, most of the time, people aren't going to know that what they have is low quality because it will almost never get tested and most people don't know anything about lock security. So they trust Amazon to tell them, when really it's the blind leading the blind
I agree with the spirit of your comment, but I think padlocks is a bad example.
The "quality" of a padlock genuinely does not matter. You should know this if you watch LPL! An experienced picker can get into any lock. And a determined but less dexterous enemy can trivially cut any lock you paid less that $100 for. And even then it won't take someone with the right tool more than a minute or two to cut through any padlock.
It's the same way with bike locks, or really any lock. They are a deterrent against the unwashed masses, but they never actually stop a determined actor.
And that is never actually the point. Most theft is not personally targeted. Sure there are the times that someone cuts a hole through a storage unit into a neighboring unit they know has expensive stuff, but mostly it's opportunistic. If you have a weaker lock, you have greater risk. Especially if your lock is so weak that it is trivial to make it look like it wasn't broken until you try to open it.
This happened to me once. I bought my lock from the storage company, but someone broke and then visibly reassembled the lock. Lost all the early prototypes from my startup, plus thousands in replaceable equipment.
> If you have a weaker lock, you have greater risk.
I think this is only true in limited circumstances. Someone went to the storage facility with the intention of stealing, and had the opportunity to window shop for the weakest looking lock.
Yup. There was a famous memo from Jeff that went round years ago that read something like "I want to buy a kettle. I search for 'kettle' and I get thousands of products. That sucks - now I have to comparison shop for kettles, when I don't particularly care what sort of kettle I get. Why can't Amazon just tell me 'buy this one'?"
Amazon's choice arrived years after my tenure at the big A, and I'm not sure it's directly connected, but it definitely helps solve this problem.
If I told a personal assistant "I want any kettle", they would know there's an implied <among buy-it-for-life high-quality stainless steel in America with a non-garish finish>.
They would also know me well enough to provide options when it's a non-commodity context, ie pressure cookers where the same implied values hold but I want to actually made the feature tradeoffs as a poweruser.
YouTube went down then global optimization path, and the experience was terrible...
Right, this makes sense to me. If you can cut down on customers' needs to make a decision, you'll increase the likelihood that someone makes a purchase. Building consumer trust by recommending actually good products is in Amazon's interest here
no, most users are still probably under the impression that "amazon's choice" means more than an automatic algorithmically derived badge. the wording implies a conscious decision was made, but there was none
it would be better phrased as what it truly is: a badge that says "popular item" or similar. even "shopper's choice". the intent of the badge is to sell (as the article states, combat the paradox of choice)
Amazon Consultants understand brand economics and translating those into Amazon retail economics.
By the time brands are talking to us, they already have products developed OR are actively launching them every year.
I am a software developer/eCommerce marketing guy, so I understand SEO and all the different levers that Amazon offers to get sales momentum moving faster and faster.
Our target market are the large brands, where sales to Amazon is still a rather small portion of their total GMV sales. Them hiring a sales consultant is better ROI (results / investment) than hiring an employee/team full-time internally.
Recently when we were deciding among a couple vitamins to buy on Amazon we noticed that the "Amazon's Choice" label got moved from one product to another during a screen refresh. We thought that was strange and were curious how they determining the labeling.
I sell a product on Amazon and have seen it be AC one day, not the next day, and then two days later, back to AC. Return rate and star rating were constant throughout the period, so I'm at a loss for how this "choice" is bestowed.
A factor may be availability and not wanting to overload a choice.
Let's say you are way down in list, and pop up to be the Amazon Choice. That drives real sales increases. They might look at how many in warehouse local to you. Persons A, B, C put in cart - they slow rate at which AC shows up until they get a sense of whether all those will become sales or some will expire from cart etc.
Amazon has some places that feel like they have softer consistency guarantees. Ie, something shows up, but with some time passing things like price might change even while in your cart etc.
Those things remind me of when I'm at a home hardware store.
They'll have a sale on say multi bit screwdrivers on the end cap. It will look like a fancy screwdriver that is on sale with lots of features and bits.
Inevitably if I by these ... they suck and are actually expensive for the quality and often have "features" that get in the way (some sort of fancy grip, more bits).
If I go down the isle I'll find for far less a simple multi bit screwdriver that is both cheaper, and higher quality.
Accordingly I never trust these signs / actually avoid them as the same rule apply to cooking tools and so many other things.
Note that Amazon says that Amazon's Choice products are "Amazon's Choice recommends highly rated, well-priced products available to ship immediately." so that would go against the idea that these products are junky or expensive.
> Why don't they only let you review items that you have purchased using the same account? Too simple I guess.
Doing that would help, but probably not as much as you'd think. A lot of the fake review scams have the reviewers purchase the products legitimately and then get reimbursed through Paypal.
They do highlight reviews from users who actually purchased the item. That's why there's a thriving business in free products in exchange for good reviews.
Giving away a few hundred or thousand items can lead to huge sales numbers.
The idea is that people may have bought it somewhere other than Amazon and to let them post a review. Amazon does distinguish between “verified purchase”s and not
Lots of people assume end caps are on sale, but they usually never are, its just stuff that the store wants to sell to you really badly, which usually means high margins (because of crap quality, or overpricing), or the brand is paying them for the privilege.
Back in the days of Windows 3.1 when people bought software in a store on floppies, end-cap positioning was discussed as if it were under the control or influence of the 'distributor' layer of the retail food chain, and therefore the developer (the company) would design the collateral for the end-cap and work with the distributor. The end-caps were not usually bargains but "important" such as newly available OS's.
I don't doubt they could be reasonable products, but stuff like a rice cooker or say crock pot isn't usually a product that has a lot of breakage... well shouldn't.
If I were an evil decision maker at Amazon, I would have the algorithm pick products that are well rated with low return rate, but have high inventory that doesn't clear fast enough. Easy way to free up space in the warehouse. Because why call out actual crowd favorites if they are already "organically" selling well?
My friend works at a liquor store. His boss one day told him to label the poorly-selling wines as "Staff Picks". So your fear isn't exactly unfounded :P
We did exactly this with some stock that was about to expire (and no one drank) at our college bar during freshers’ week - we made a ‘college special’ cocktail containing it and advertised it on the first night.
Backfired as it proved immensely popular and I had to buy hundreds of pounds more of it from the wholesaler the next day and pick it up ourselves as delivery wasn’t for another week...
I guess that isn't wrong, your boss is part of the staff and they have "picked" the poorly selling wines to highlight. It didn't say "Staff Recommended" or "Staff Favorites"
I think the average shopper would interpret "Staff Picks" as meaning "wine that the staff recommends" instead of "wine the store is hoping to sell." It's disingenuous at best.
For the record, my friend initially refused, and then compromised by labeling them "[Manager name]'s Picks".
I was being facetious. It definitely is extremely misleading based on any common understanding of what a “pick” implies in terms of quality or attributes of a product.
I look only at the negative reviews. You can trust them to highlight the weak points of a product. So I know at least what would be the "worst" case scenario and if that works for me I buy it.
I prefer to 2 to 4 star reviews. Too many of the 1 star reviews are people who received a defective product or had other unique issues that I'm unlikely to encounter.
When buying a product I look for the reviews about it breaking. Once I know all the common ways the product might break I have a bit of an idea how long it will last. I can then also look up how much the replacement parts for the most likely to break parts cost as well as how hard they are to replace.
In my experience a lot of products have a single design flaw that causes almost all of the breakages excluding drops/water damage. I had a set of skullcandy headphones that the little arm thing on the ear part would snap just from the force of putting them on and taking them off too many times. Got a replacement and it happened again. Checked reddit and saw everyone has this same problem. Ended up getting the store to exchange them for a different pair.
This is something I do too. Recently, I stumbled upon a product where this was exploited. It read something like that: "Are you like me and read the negative reviews first? If that is the case I have to tell you that this product is really great!..."
My approach is to only look at the reviews with photos. That's usually a good way to find high-quality, detailed reviews. Just click in the customer photos section and you can scroll through the reviews.
This is purely anecdotal, but I've noticed recently a trend of reviews that start with "I was sent this product for free in exchange for an honest review..." and they almost always include photos. True, these are easy enough to skim over, but people planting fake reviews could also be getting smarter and making their reviews seem real by including pictures.
They're not necessarily fake, but they may be incentivised. I've had offers of a £5 voucher for a photo review or £10 for a video (the product was a £15 solar panel). No idea what happens if you slate the product, but amazon would probably refund you if it was that bad.
Sellers are well aware how much of a boost a photo or video from a verified purchase can give.
It bothers me because the term pretty strongly implies that a "choice" involves a human doing the choosing. When I first saw it, I'd hoped it was a sign that Amazon was going to be investing time in curating their junkyard swapmeet of a store. Not so much.
I think an interesting point to keep an eye on is recommendations with negative side effects, e.g. amazons choice for a high security lock has a terrible flaw. https://youtu.be/kJ1_P5oqf6Y
This is a good example as it really does lay the Amazon gig bare. LPL knows his stuff but, for customer recommendations, do you really need LPL level product knowledge?
Nope.
I have sold locks as just one of thousands of products in the bicycle trade and you don't need to sell that many to know what you should be recommending to customers. It is not difficult. You don't have to pick/angle-grind them, you just need to listen to the customers.
There is a forest of locks out there, all different for slightly different use cases and customers do need a little bit of advice to get the right one. It is one of those situations where you can have two out of three of these things: security, a nice price or a manageable weight.
There is unwritten moral obligation to provide some customer service, as a shop assistant it would be pathologically wrong to sell someone a lock that is not suited for their use case. The same goes for any other product.
Sure there may be times when you might try your hardest to shift inventory, e.g. the models with last year's packaging that have got a bit dusty, or the ones that are over-abundant in the warehouse. But you can do that responsibly so the customer knows what the deal is.
In normal retail with a finite amount of shelf space you do not have sub-standard clone products taking up tonnes of space with little product differentiation. It makes no sense to sell these products as the profit from the sale is nowhere near as good as what you get selling a legitimate product. Besides, how on earth are the customers supposed to choose when the options are mind-boggling?
With the LPL Amazon locks series it seems clear to me that Amazon are selling highly branded products that are useless as 'choice' as well as off-brand junk as 'choice'. If a regular shop assistant was doing what they were doing then you would wonder whether they were trying to destroy the company!
In regular retail if you realise that some locks are sub-standard then you get the customers coming back with some horror story about how their shed was raided. Or they come in with the remains of what the thieves left behind. After profuse apologies they are cool about it, they have insurance and you help them get what they need. You then decide not to stock the offending locks again or to sell them only to people who are okay with the compromise.
Amazon just do not have this type of a feedback loop, they are not trimming down their range to only offer the customer the good stuff. They just don't get this aspect of customer service even though they excel in other areas.
I have seen some really questionable products recommended by "Amazon's Choice." I have assumed that in addition to customer input, the choice is probably working against me (recommending things with higher margins, items that Amazon needs to get rid of, or kickbacks themselves).
I bought my parents some outdoor pathway solar lights marketed as "Amazon's Choice" that turned out to be cheap garbage. Their first set that they bought decades ago lasted over 10 years. Then next two sets they bought fizzled out in 2 years or so. With solar, battery & led technology moving forward exponentially I figured it would be easy to find nice solar lights for a good price. I set out to buy a decent set of lights for them & didn't mind paying a little more for some quality lights. I couldn't decide between a few models & Amazon kept recommending their "Amazon's Choice" that didn't list as many technical specifications as the ones I was considering. Time marched on & I needed to make a decision & decided to trust Amazon with their choice. I figured that their AI would know I analyze many of my purchases extensively & pay extra for quality, so I assumed that info would probably be in my account profile. (yeah I remember the old saying too!) I was never so embarrassed to give somebody a gift then when I gave the "Amazon's Choice" solar light set to my parents. Some didn't work at all right out of the box. Within 6 months only two worked. They were dim, really cheap (stakes broke) & were that annoying artificial light spectrum. I don't buy much from Amazon anymore for this & other reasons, but when I do, if it says "Amazon's Choice" it doesn't encourage me to buy, it discourages me from making a purchase.
In Ye Olde Dayes, these sorts of promotional considerations were paid for by the manufacturer. When Amazon was still mainly selling books, 'coop' funds paid for favorable placement on the intro pages for genre sections (the same way bookstores arrange endcaps and window displays).
But it sounds like this has been replaced with a system that doesn't require any sort of human input.
TLDR; looks like none of what I thought "Amazon Choice" meant is true. This is what I thought: (1) they had verified seller through some background checks, (2) made sure the seller was selling genuine brands/product, and (3) the product was the best selling among such verified products.
Amazon is just a mess. Search for "apple watch charger cable". Among the results is one "Amazon's Choice", this one by ATETION [1], sold by ZDAGO and fulfilled by Amazon, with Prime shipping. Amazingly, all the reviews in the listing seem to actually be reviews for an Apple watch charger cable, which is a rarity for that category. Almost every other third party cable listing is rife with reviews and questions for other products.
So, in this case, the "Amazon's Choice" label seems to be very useful--it is marking pretty much the only third party Apple watch charger cable on Amazon that doesn't show obvious signs of being sold be a scammer.
However, if like me before you go look at the search results you go over to the side and check the "Prime" box to refine the search to just listings with Prime shipping, the "Amazon's Choice" ATETION listing from ZDAGO goes away! This makes no sense.
The results still do contain what appears to be the same ATETION cable (at least the specs are the same) [2], except now it is sold be WEIZY and fulfilled by Amazon, and is not an "Amazon's Choice", and is slightly cheaper. The WEIZY listing is a cesspool, like pretty much everything else in the category except the ZDAGO listing. It's full of questions and reviews that have nothing to do with the charger cable.
Sounds like their algo has reached youtube recommender/ranking algo level of "we don't know what the black box does, we just try to fix it when it fucks up"
That the company won't say that it's hand-curated, with the observations that the article makes, seems to suggest that automation is being used to some degree. Which wasn't obvious to me as a customer.
Yes, it's the #1 search term by their algorithm, with some badging to psyochologically nudge you toward confidence to buy. It's conceptually no different from a floor salesman telling you that whatever item you are currently looking at is the best choice.
It's similar to how almost every Amazon product is a "best seller" in an extremely narrow category.
Presumably it's like any other store, products with slight defects ("it was on offer, not worth returning"), excess products taking up warehouse space, old versions, highest profit item in a class, paid promotions.
Why would Amazon do it differently, they clearly don't care about product quality; how else would they use the badge?
> Presumably it's like any other store, products with slight defects
What you are describing sounds like "clearance" which is a pretty upfront designation (this didn't sell very well, it's here at a discount before we chuck it).
But I think the message that "Amazon's Choice" is intended to convey is "this product meets certain quality and value standards and is unlikely to disappoint you."
Of course, in practice, they could apply it in order to clear inventory. But with FBA, the inventory carrying cost isn't primarily theirs.
No, clearance is different. At least in UK stores that's marked down from the shelf price.
Example, Battenburg cake, when it's on promotion it always has one of: mismatches (not quite square), loose seams, cake edges (crust that's darker coloured, not burnt), too much jam. It's not old, but without the promotion it would probably end up in clearance because people are more choosy with "full price".
Similarly in restaurants, the waiter will recommend food that's not inferior, but would become so if it's not sold. Or to use a product their supplier discounted.
Managers special might be "we got excess inventory at our central warehouse and I get a bonus if we shift more of these, which are perfectly good product".
That would be burning Amazon's reputation for dubious short-term gain. If they're thinking about the longer term, it would make more sense to play it straight: make the Amazon's Choice recommendations as honest as possible, and seriously try to help their customers find good products.
- Customers are happy when they get stuff fast - it needs to be available right now, and in the buy box
- Higher rated items are theoretically better than lower rated items
- "choice" badge is keyword specific: a plant light could be choice for "grow light bulb" and "grow light led" but not for "light bulb".
Our data shows the items are not correlated by margins for. We often see negative NET PPM on "choice" items.
Source: I run an Amazon consultancy that manages roughly $30M/year sold through Amazon.