Subscribe to DSC Newsletter

Shopper Alert: Price May Drop for You Alone | NewYorkTimes

When your grocery store bill is based on your profile...or another way for retail companies to optimize ROI via customized pricing (something all consultants, merchants and sales people have been doing for thousands of years, so it's not really new stuff)


It used to be that with dedication and a good pair of scissors, one grocery shopper could get the same coupons — and cheap prices — as another.

Now going to the grocery store is becoming a lot less egalitarian.

At a Safeway in Denver, a 24-pack of Refreshe bottled water costs $2.71 for Jennie Sanford, a project manager. For Emily Vanek, a blogger, the price is $3.69.

The difference? The vast shopping data Safeway maintains on both women through its loyalty card program. Ms. Sanford has a history of buying Refreshe brand products, but not its bottled water, while Ms. Vanek, a Smartwater partisan, said she was unlikely to try Refreshe.

So Ms. Sanford gets the nudge to put another Refreshe product into her grocery cart, with the hope that she will keep buying it, and increase the company’s sales of bottled water. A Safeway Web site shows her the lower price, which is applied when she swipes her loyalty card at checkout.

Read full story at

Views: 1151


You need to be a member of AnalyticBridge to add comments!

Join AnalyticBridge

Comment by Vincent Granville on August 18, 2012 at 5:36pm

@Lance - my background (in terms of price optimization) is limited to selling consulting services and advertising. It's a different framework than retail, but differential pricing is standard - partly because we can offer lots of extra free stuff to our best advertising clients. But I did an interesting study back in 2001 to show that had to offer a price that was a function of the minimum prices offered by its 3 large cheapest competitors, in order to increase total leads. We recommended a non-random price, but the price would change 3 times of month, sometimes 3 times a week, for any given product.

Comment by Lance Olson on August 18, 2012 at 5:24pm


I am with you on the auction and how most of the auction operates and what you propose.  I am stuck on your phrase "Even you if you check the price and check one second later, it will be different.  Part of what determines the price is a random generator, ..."

I am fine with everything else but those words.  This is exactly what Amazon did according to Bezos

-- Begin Excerpt from BizJournal Link below.

"What we did was a random price test, and even that was a mistake because it created uncertainty for customers rather than simplifying their lives." He said the company's new policy is that if it ever again tests differential pricing, it will subsequently charge all buyers the lowest price.

-- End Excerpt from BizJournal.

A single customer sees different prices after a very short amount of time goes by, say less than one day.  This would be a PR nightmare for any retail shop, in my opinion.

I am fine with your "...random price for a given book increase on average over time..." - who is to say there is some methodical method retailers go through to determine price.

That is a pill hard for anyone to swallow.  Maybe if it was a little amount say less than $5 on a $100 purchase, that would be a  little better.  But nothing less than a day on the elapsed time.

Comment by Vincent Granville on August 18, 2012 at 4:18pm

@Lance: simple solution: the price changes all the time: all your friends get a different price. Even you if you check the price and check one second later, it will be different. Part of what determines the price is a random generator, and the fact that the more you check it, the more expensive it gets, forcing people to not try too hard.

Indeed, this model could detect people who do not have time to engage in lengthy cheating activity, or even offer the pricing model to people who enjoy this auction model - as long as it offers a remote chance to win. The system would use the auction model exclusively with them (I'm one of them, I have better things to do than save $5 on a computer science book especially if it takes me an hour to beat the system).

Indeed if the random prices for a given book increase on average over time (say by 5% a year, although it could decrease for books that are hard to sell), it will force people to buy books ASAP and boost Amazon revenue.

Comment by Lance Olson on August 18, 2012 at 3:36pm


How would you prevent collusion in discovering the secret minimum price?  For example: I can call several of my friends that might be interested in the product.  We start low on the price and we take turns, singly, entering in a new increased price until Amazon accepts the new price immediately.  Then once we discover an acceptable price we start telling everyone.  "Go to Amazon for product A and use this price $."

If the burden of proof is on me to figure this out, I would tier the minimum price so that if we get "a lot" of low bids on one item then we would increase the minimum price to the next tier. 

So basically the minimum price would have some sort of profit built in, I would think.

I have my doubts with an auction after typing this in.  Not that it cannot be done BUT this would induce unnecessary uncertainty in revenue.

One of the aspects of business is continuity, doing an auction provides little continuity in revenue and makes product planning very difficult.   Continuity is essential on Wall Street in providing forecasts for the investors - that is if someone wants to attract investors.   Investors and/or Wall Street likes the forecast very much...see the news when a company misses their forecast - they get penalized with a massive sell off. This takes me back to loyalty cards and/or compensated demand.  People would know the profits on the first purchase the second purchase would be a different story if the consumer decides to take advantage.  This is a little more solid continuity in business.

Comment by Vincent Granville on August 18, 2012 at 1:00pm

@Lance: I'm thinking at model where you log on to your Amazon account, see an interesting book, it is listed with a suggested price of $150, you offer $130, and because $130 is above the secret minimum price of $125 that Amazon is willing to sell, your $130 offer goes through right away. 

It could work with the one-click purchase feature. However, I'm afraid people will cancel their order right away, try a new price (say $127) and try yet another one until they've managed to discover the magic $125 price. Maybe a way for Amazon to eliminate this problem is to have a truly interactive price: the first time you try, Amazon will sell it for p1=$125, but on your second attempt, it is now p2=$128, and then p3=$130 on your third attempt. The goal being to make more profit on you than if the book was listed at $125, by optimizing p1, p2, p3.  

Comment by Lance Olson on August 18, 2012 at 12:07pm


I am not sure that I understand the auction model.  Maybe explain in more detail.  With an auction, if the transaction has a lag time to settle on the amount and the customer would go some where else to get the product quicker.  

I guess a quick dutch auction meaning the customer would be able to make an offer then the settlement would be for the retailer to accept or reject or make a new offer - that would be an immediate reply by the retailer.  At most, I think the back and forth would be no more than 2 per party.

The pay per click model would have more possibilities, from my perspective.  For example: the retailer could offer different prices to the different advertising channels.  This would presume that advertising channels are abundant to get a good market response, or any kind of response.  The trick is that there is no overlap with advertising like the same advertising goes to the same person with different prices on the same product.

Comment by Vincent Granville on August 18, 2012 at 11:49am

@Lance: I'm wondering if Amazon could sell books and other stuff using a pricing model that is an hybrid between

  • traditional static prices and 
  • an auction model where a minimum bid (unknown to the purchaser) is required to win the auction.

Indeed, a model similar to pay-per-click bidding on ad networks (e.g. on Google) to purchase advertising spots.

Comment by Lance Olson on August 18, 2012 at 11:15am

This is reminiscent of problems a number of years ago, I think it was in 2000.

Amazon was trying to do a "Differential Pricing" to gain insight into what customers would pay. They did their study by openly advertising the different prices to the masses. Differential or Dynamic, the name may be different but the practice is still the same.


See also wikipedia for controversies - .

Then in 2005, CNN ran the following article:
According to the CNN article, Dell does the same type of "Dynamic" pricing.

All that set aside, someone else can pass judgement.

Technically, this "dynamic pricing" would be a cool project. I am definitely curious, especially about the buying habits. Just a side note, the information is already available from list brokers that collects all sorts of information about consumers.  

But considering the fact that consumers can text their friends across the country in minutes only to find that a business is playing mucky muck with prices is a bad idea. I can text my friends and family, who are out of state, to find good prices on the same online retailers. Then if my friend has the better price then I will ask them to buy it for me and I will send a check to them.

Silly online retailers thinking they can openly charge different customers more money for the same product.

This is nothing new, the retailers already charge differently at different locations across the US, charge more at the different regional markets that can take higher prices.

Silly consumers thinking that retailers charge the same at all geographical locations.

Today, the only difference is that consumers are more aware of the practice through the internet and texting.

Retailers just need to stop openly practice and think of better ways to get the information.  They need to invent a better practice for the new century and stop relying on what has worked in the past.

I personally would offer some sort of loyalty card/compensated demand offer. This would give more details across the transaction and be able to provide insight in a lot of detail. Then the retailer would be able to provide customer specific rebates to test the different elasticities. If the loyalty card is swiped then the retailer can track ALL transactions, even the cash only customers. Then you don't have to worry about the appearance of "discriminating" and be under the microscope of your customers.

Comment by Vincent Granville on August 15, 2012 at 10:01pm

Could it work with an online retailer? Amazon could do the following test: set book prices that change in real-time, even for a single user. You check a book twice within a 5-minutes time period: the fist time it is listed at $32.87 (say), and the second time at $32.91 or $32.68 depending on time of day, IP address, user and some random component, Maybe the price would be based on your browser too: are Chrome users willing to pay more than Internet Explorer users for technical books? If yes, charge them more.

Amazon could optimize the prices (in real time) using machine learning with metrics such as time of day, IP address, random generator parameters, prices from competitors, user profile, geo data, tax parameters, browser etc. Nobody might even realize that "user profile" or "purchasing history" is part of the algorithm because of the tiny random component added to the price, unless some data scientists spend time reverse-engineering the pricing data.

Anyway, this gave me an interesting idea: crawling with various IP addresses at various times to see how prices change over time, and check if there are patterns. The data (book prices) is public - you just need to run a crawler to extract it, on sampled books.

On Data Science Central

© 2021   TechTarget, Inc.   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service