5 ms·

This is a point I bring up when casually talking with non-tech minded friends when they start talking about machine learning taking over everything... I bought

This is a point I bring up when casually talking with non-tech minded friends when they start talking about machine learning taking over everything... I bought a Dyson last summer. Amazon is still suggesting that I buy more vacuum cleaners.

itissid 8y ago
And selling Hawaii vacation packages to people after they have been there. And I live on the east coast.
fastball 8y ago
Tbf Amazon has the least experience with ML of the FAANGs.
- hawski 8y ago
  It seems that it would pay out a lot for them to do a few simple heuristics by hand. For example: if just bought something do not show a thing from the same category. At most show a thing from accessories category, which I think they do.
  - rusk 8y ago
    Maybe that just goes to show just how much money they're making that they don't really need to feel bothered!
  - tmoravec 8y ago
    Nah. That's not AI enough!
  - hashhar 8y ago
    Up until recently i used to work at a major Indian eCommerce and can confirm that most companies use a mix of the simple heuristics you suggest together with some data mining. So, they would take your recent purchases, clicks, searches and wish listed products and recommended products from some accessory category or related products. ie. Traditional mining combined with heuristics. It also helps to have separate widgets for similar products and for cross-sell products.
  - aurelwu 8y ago
    I always wondered why that happens at all but then I thought maybe it is because of their historic origins starting out as book seller - where showing another book after you bought a book makes absolute sense. While for refridgerators it makes no sense, they should put stuff in categories to distinguish that but then they are a super big company and I am single selftaught developer trying to make my living out of it.
- jhayward 8y ago
  It depends on how you measure that. They certainly were the first to make serious money by using simple regression to suggest 'people also bought'. In general I think that all of the personal surveillance / categorization will end up being more truly profitable (as opposed to 'twinkie calorie' profits of ad networks) as they are used to suggest price points for products and services rather than to direct attention to products.
SilasX 8y ago
"We showed him an ad. He bought the vacuum cleaner." 'In that order?' "Shhhh, just give us money." Half-joking there -- in a previous discussion, the claim was that the ad-buyers don't care enough to check for whether these product-buyers were ones that had already seen an ad, or it was after the fact.
- semi-extrinsic 8y ago
  This makes sense. While the "you might need a second vacuum cleaner" argument maaybe checks out, the "you might need a second sous vide machine/espresso bean grinder/welding machine" definitely doesn't make sense.
  - dsr_ 8y ago
    There are some people who end up buying all the best coffee bean grinders in search of perfection; there are people who can't resist a classic MIG welder. But those people were going to buy them anyway, and represent too few sales to justify any expenditure. What you need are complementary goods. These people bought coffee bean grinders? Show them ads for travel mugs, for high-end coffee subscriptions, and fancy grinder-cleaning brushes.
    - semi-extrinsic 8y ago
      Agreed; just to be clear, when I've had this happen to me it's showing an ad e.g. for the exact same model coffee grinder from the exact same store that I already bought one from. Clearly no intelligence involved there.
pbhjpbhj 8y ago
Dyson users probably disproportionally have second homes, holiday homes, etc.. You're more likely to recommend to a friend of you see a deal "hey, I bought this awesome vacuum, you should get one, I saw a deal" ... It's not as silly as it seems. I buy repeat items as gifts (I like this, X will like this), or for my parents (home stuff), or when I bought something for one child I know and it turned out to be well made and now another child needs/wants one; or tech stuff that turned out well when someone else needs the same repair (eg new HDD).
- chipotle_coyote 8y ago
  It's not as silly as it seems. I'm not sure I agree. I mean, I see the theory you're getting at, but the Dyson example is built on a few assumptions we'd need more data on -- starting with the notion that Dyson owners are disproportionately more likely than owners of other vacuum cleaners to have second homes. And even if they do, that's, well, one more sale at most, and one that's more likely to be influenced by their experience with the first Dyson rather than ongoing advertising. When you say that you buy repeat items as gifts because they turned out well, you're tacitly confirming that further advertisements for that product aren't necessary to reach you. At best, the ads can sway you if they happen to be running an unusually good deal on the product you've already decided you want to buy. The big problem targeted ads have now is, as other people have pointed out, that they seem to be targeted with knowledge of what you've recently been looking to buy, but not knowledge of what you've recently bought. If I search for polo shirts, I'm in the market for polo shirts, but once I buy polo shirts, I'm probably not going to be in the market for them for a few months. Once I buy a car, or a television, I'm probably not going to be in the market for another one for years.
  - not2b 8y ago
    No one has the data for both sides of the equation. Google dominates search advertising, Amazon dominates online buying. If those companies merged, one company would have all the data needed to use deep learning to take all our money. They'd be able to distinguish products that, once purchased, means the buyer won't buy again for years, and items that are regularly bought, and they'd have the timing between seeing ads and buying. It wouldn't be perfect, because some people would buy elsewhere, but it might be frighteningly effective.
magicalhippo 8y ago
This also puzzles me. If I buy a household appliance that's not a consumable product (most shouldn't be), why does it recommend more of the exact same? Why doesn't it instead recommend other types of appliances? The key take-away should be that you're comfortable with buying household appliances online, usually not that you need another vacuum cleaner within the next year.
- manigandham 8y ago
  It's data, and a lack thereof. It's easy to know you're browsing a site. It's almost impossible to know you bought something. Banks and merchants and publishers and advertisers and all the layers in between create dozens of data silos and nobody is interested in sharing. Add to that the weird expectation that people don't want to be tracked but yet want adtech to know when you've bought something.
  - magicalhippo 8y ago
    Well surely Amazon knows what I bought on Amazon...

itissid 8y ago

And selling Hawaii vacation packages to people after they have been there. And I live on the east coast.

fastball 8y ago

Tbf Amazon has the least experience with ML of the FAANGs.

hawski 8y ago
It seems that it would pay out a lot for them to do a few simple heuristics by hand. For example: if just bought something do not show a thing from the same category. At most show a thing from accessories category, which I think they do.
- rusk 8y ago
  Maybe that just goes to show just how much money they're making that they don't really need to feel bothered!
- tmoravec 8y ago
  Nah. That's not AI enough!
- hashhar 8y ago
  Up until recently i used to work at a major Indian eCommerce and can confirm that most companies use a mix of the simple heuristics you suggest together with some data mining. So, they would take your recent purchases, clicks, searches and wish listed products and recommended products from some accessory category or related products. ie. Traditional mining combined with heuristics. It also helps to have separate widgets for similar products and for cross-sell products.
- aurelwu 8y ago
  I always wondered why that happens at all but then I thought maybe it is because of their historic origins starting out as book seller - where showing another book after you bought a book makes absolute sense. While for refridgerators it makes no sense, they should put stuff in categories to distinguish that but then they are a super big company and I am single selftaught developer trying to make my living out of it.
jhayward 8y ago
It depends on how you measure that. They certainly were the first to make serious money by using simple regression to suggest 'people also bought'. In general I think that all of the personal surveillance / categorization will end up being more truly profitable (as opposed to 'twinkie calorie' profits of ad networks) as they are used to suggest price points for products and services rather than to direct attention to products.

SilasX 8y ago

"We showed him an ad. He bought the vacuum cleaner." 'In that order?' "Shhhh, just give us money." Half-joking there -- in a previous discussion, the claim was that the ad-buyers don't care enough to check for whether these product-buyers were ones that had already seen an ad, or it was after the fact.

semi-extrinsic 8y ago
This makes sense. While the "you might need a second vacuum cleaner" argument maaybe checks out, the "you might need a second sous vide machine/espresso bean grinder/welding machine" definitely doesn't make sense.
- dsr_ 8y ago
  There are some people who end up buying all the best coffee bean grinders in search of perfection; there are people who can't resist a classic MIG welder. But those people were going to buy them anyway, and represent too few sales to justify any expenditure. What you need are complementary goods. These people bought coffee bean grinders? Show them ads for travel mugs, for high-end coffee subscriptions, and fancy grinder-cleaning brushes.
  - semi-extrinsic 8y ago
    Agreed; just to be clear, when I've had this happen to me it's showing an ad e.g. for the exact same model coffee grinder from the exact same store that I already bought one from. Clearly no intelligence involved there.

pbhjpbhj 8y ago

Dyson users probably disproportionally have second homes, holiday homes, etc.. You're more likely to recommend to a friend of you see a deal "hey, I bought this awesome vacuum, you should get one, I saw a deal" ... It's not as silly as it seems. I buy repeat items as gifts (I like this, X will like this), or for my parents (home stuff), or when I bought something for one child I know and it turned out to be well made and now another child needs/wants one; or tech stuff that turned out well when someone else needs the same repair (eg new HDD).

chipotle_coyote 8y ago
It's not as silly as it seems. I'm not sure I agree. I mean, I see the theory you're getting at, but the Dyson example is built on a few assumptions we'd need more data on -- starting with the notion that Dyson owners are disproportionately more likely than owners of other vacuum cleaners to have second homes. And even if they do, that's, well, one more sale at most, and one that's more likely to be influenced by their experience with the first Dyson rather than ongoing advertising. When you say that you buy repeat items as gifts because they turned out well, you're tacitly confirming that further advertisements for that product aren't necessary to reach you. At best, the ads can sway you if they happen to be running an unusually good deal on the product you've already decided you want to buy. The big problem targeted ads have now is, as other people have pointed out, that they seem to be targeted with knowledge of what you've recently been looking to buy, but not knowledge of what you've recently bought. If I search for polo shirts, I'm in the market for polo shirts, but once I buy polo shirts, I'm probably not going to be in the market for them for a few months. Once I buy a car, or a television, I'm probably not going to be in the market for another one for years.
- not2b 8y ago
  No one has the data for both sides of the equation. Google dominates search advertising, Amazon dominates online buying. If those companies merged, one company would have all the data needed to use deep learning to take all our money. They'd be able to distinguish products that, once purchased, means the buyer won't buy again for years, and items that are regularly bought, and they'd have the timing between seeing ads and buying. It wouldn't be perfect, because some people would buy elsewhere, but it might be frighteningly effective.

magicalhippo 8y ago

This also puzzles me. If I buy a household appliance that's not a consumable product (most shouldn't be), why does it recommend more of the exact same? Why doesn't it instead recommend other types of appliances? The key take-away should be that you're comfortable with buying household appliances online, usually not that you need another vacuum cleaner within the next year.

manigandham 8y ago
It's data, and a lack thereof. It's easy to know you're browsing a site. It's almost impossible to know you bought something. Banks and merchants and publishers and advertisers and all the layers in between create dozens of data silos and nobody is interested in sharing. Add to that the weird expectation that people don't want to be tracked but yet want adtech to know when you've bought something.
- magicalhippo 8y ago
  Well surely Amazon knows what I bought on Amazon...