Algorithmic Pricing On Amazon 'Could Spark Flash Crash'

DerekduPreez writes "Sellers on Amazon's retail site are increasingly using high-speed algorithmic trading tools to automatically set prices, which could lead to a malfunction similar to the 2010 flash crash. According to the Financial Times, prices on Amazon's website change as often as every 15 minutes, where sellers are using tools traditionally developed by data miners at banks to ensure that their prices are always below their rivals'. Third-party software is allowing sellers to detect a competitor's price and automatically undercut that price by, for example, £1. However, this could lead to a situation similar to the U.S. flash crash, where algorithmic trading was blamed for stock prices falling to near zero and then bouncing back within 20 minutes." At Slashdot's sister site for Business Intelligence, Nick Kolakowski has some more information on this possibility.

Comments

Falling to near zero??

[bhlowe]

I didn't read the article, but presumably the traders wouldn't allow their sale price to drop below the cost of the item plus the marginal expense to sell on Amazon.... so if anything, the prices will drop to at or near the cost of the item... Which is good for buyers, bad for resellers...

Re:Falling to near zero??

[jpmorgan]

And if they do, it's still good for the buyers, and the sellers aren't likely to make the same mistake twice.

With algorithmic pricing, the Amazon marketplace is just operating as an automated dutch auction. It's how markets should behave: raw supply and demand, with no collusion or other market distortions propping up prices.

Re:Falling to near zero??

[MortimerV]

Exactly, a floor price is used to prevent this sort of crash from happening. I'd imagine there could be some sellers on there that haven't set up their floors properly and they could lose money on a few products, but the entire site won't implode from this.

If those sellers don't honor the prices, they'll get bad user ratings and lose some future sales over it.

Re:Falling to near zero??

[History's Coming To]

You'd be surprised, there are many businesses built around a model of selling at a loss for the first year or two just to pressure the competition and build a reputation as the cheapest, then they ramp the price up once they have a sufficient chunk of the market. Businesses will also sell old stock at a loss simply to free up capital that's trapped in stockholding.

Re:Falling to near zero??

[jrroche]

With algorithmic pricing, the Amazon marketplace is just operating as an automated dutch auction. It's how markets should behave: raw supply and demand, with no collusion or other market distortions propping up prices.

Because everyone automatically undercutting their competitors by a few cents over and over until everyone is selling at cost and all but a couple players eventually have to shut down because they can't afford to run a profitless business forever, whereupon the few remaining players can finally raise prices ... isn't effectively collusion or a market distortion.

Re:Falling to near zero??

[roman_mir]

Your comment is so stupid, it's beyond words.

It's not the ECONOMY that 'ends' in a free market when somebody stops making a profit, it's that specific business model that puts somebody out of business ends and the market actually LEARNS something from that mistake.

Now contrast with the government ran economies - the same stupid thing is done over and over and over, the money is printed, the inflation causes savings to be wiped out or moved out and asset bubbles form while productivity falls.

Then the recession hits, which is free market attempting to reset the problem by erasing the fake credit, the fake values of assets, restore interest rates so that savings can be made again.

Then the government steps in with more fake money and fake interest rates, trying to inflate another asset bubble and call it 'economic recovery'. This is done in a loop over and over until either the government gives up and economy restructures (fat chance) or the free market stops playing the game, nobody wants to trade for that destroyed currency and there is a real economic collapse, which forces the restructuring anyway.

This is exactly the same as what happens when people try to prevent small fires that happen in forests and clean up the underbrush and all the rotten and fallen wood. For a while people prevent small fires from burning through the woods, eventually there is so much fuel collected in the forest that a huge fire destroys the forest and everything around it.

So what?

[Lev13than]

As long as Amazon forces the sellers to honour the price, then I don't see a problem. Pure market forces will balance the risk/reward for dynamic prices - if one or two consumers get lucky, then that's the cost of doing business.

The biggest mistake that the exchanges made following the flash crash was to cancel the errant trades - if you fuck up the pricing, you need to deal with the consequences. Getting rid of downside risk removes half the equation and blocks any incentive to play smart.

Re:So what?

[SirGarlon]

it charges you a higher amount for the same product because it assumes you have a higher willingness to pay

And if you do have a higher willingness to pay, then I don't see the problem.

So what?

[SirGarlon]

Who, other than the retailers, care if there is a "flash crash?" Presumably if they lose enough money on flash crashes they will stop with the algorithmic pricing.

Likewise, if I find algorithmic pricing makes prices unattractively high, I'll shop somewhere else. There has to be someone out there selling a product similar to what I want, without the algorithmic gouging.

In short, I think this is one case where we can trust the market to operate correctly.

Re:Problem?

[Kjella]

Approximately 99.999% of all shops I know both online and offline have some sort of "typo clause" in their terms and conditions, if their $100 item is suddenly $1 for some reason I think most will choose to exercise it if the algorithm goes completely bonkers. But if it's a minor mispricing relative to the item value or their total sales and they don't want the flurry of 1-star reviews that's bound to follow, they eat that loss. Been there, done that, got my order fulfilled - let's call it a surprise sale for both parties. If you're fucking this up so badly you can't take it, you really got no business running a retail store.

Won't happen. Here's why:

[Qbertino]

A little history:
I was the first to automate price wars on amazon marketplace. (True thing.)

A friend had just joined marketplace with a freshly founded internet media sales joint after it opened and two weeks in was adjusting prices of his sale books manually. Like, seriously, clicking through 200 items a night and entering new prices. I told him to stop that nonsense and built an automated scraper, parser and some other tools in Python that would parse the actual websites for each of our articles ISBN and compare our prices to those of the competition (this was before the days of publicly available Amazon APIs), readjust our pricing to the cent accordingly and upload the freshly generated updates once all 200 000 items were parsed.

Orders went from 3 - 5 per day to 120 - 150 per day. My buddies were packaging books and CDs all day while I was sitting there grinning and petting my script and ama-bot setup (still those right here in my project folder :-) ). We made 700 000$ of revenue the first year. A few months in competitors started to do the same - no suprise, the concept is quite obvious to any computer or programming guy - and a ruinous price-war started. My friend went out of business a year later. We could have fine-tuned the automated price adjustments like the marketplace vendors are doing today, such as upping the price of an item only you have got in stock, but after a few bad business decisions my friend didn't want to continue. That was all back in the early 2000s (2003-2004ish).

On the issue discussed:
Before a Flash Crash can happen on sites like amazon marketplace, the vendors involved will either die a painfull death before or finetune their algorythims to a much more complex model. Those still alive and well today have done the latter, and even if updates occur every 15 minutes, I'd bet money that they are still watching the sales and revenue with the appropriate tools and with their own eyeballs, because you can lose thousands within minutes if you don't. You can automate a lot, but you can't automate day-to-day business decisions, especially in such markets.

Bottom line:
Crashes don't happen here, only individual foreclosures for those who don't watch out well enough.

My 2 cents. .... Aaaah, the memories ...