Slashdot Mirror
Price Optimization Software Big in Retail Business
prostoalex writes "Even if you spent only a single day in an economics class, you're probably familiar with a concept of supply and demand. The Associated Press is running an article on retailers employing mathematical models for price optimization, where some products are priced higher to generate higher margins, and some are discounted to generate larger volumes even at the expense of per-product margins. DemandTec, Oracle and SAP are some of the companies producing those mathematical models for retailers around the country, with AP listing some of the pricing optimizations employed currently."
Seriously - this is NOT new. Not even in the software field.
and make it up in volume! That's what I always say.
this is more about consumer behavior than straight economics. the optimizations referred to aren't just adjusting pricing to supply and demand, but, as noted in the article, address perceived value as well. I'm no economist, nor do I want to be, but it seems to me that such analysis can uncover otherwise unexpected responses to price adjustments.
Customers vary in their willingness and ability to pay. If a company charges one simple price for each item, it creates a situation in which some people get a great deal (they pay less than they might be willing to) and some people don't buy the product at all (because the price is more than they want to pay). But if a company can find a way to separate the customers that really value the product from the customers that value it less, then more people will be able to buy the product and the company will earn more profit. (You mathematically prove that this increases what is called consumer surplus which is the equivalent to the consumers "profit" on the purchase and the seller's profit). Both side benefit, as does society.
The amusing fact is that this is nothing more than a capitalist version of taking from the rich (those are willing and able to pay more) and giving to the poor (those aren't willing or able to pay more).
Two wrongs don't make a right, but three lefts do.
As other posters will doubtless already have said, price point optimisation by software is neither new nor interesting. What is interesting IMHO is the scale of the whole system. Big box superstores employ an army of psychologists, ergonomics experts and statisticians to try and control your behaviour and squeeze as much cash as possible from your pocket.
:)
There was a quite fascinating article published the other day in a Digg linked blog that I am sure many here read (I don't have the link unfortunately). What is really interesting is that by knowing the system and subverting it you can make HUGE savings in your shopping. The layout of the store is carefully crafted to expose you to the products they want to push. Color schemes and shelf placememt are designed to confuse or lead you to select certain products. Prices and product sizes are carefully designed to make comparative math very difficult to ordinary folks. Bargains are placed outside the normal lines of sight.
In other words, the very existence of a cold and calculated system is what enables you to game it.
Some bits of strategy I remember:
1) Make a list and stick to it. Impulse purchases account for a huge amount of profit and the stores rely on you buying things you do not need.
2) Never look at the products at eye level, they are the most expensive and worst value.
3) Move as fast as you can to the back of the store. Start at the back of the store and work your way forwards.
4) Do not stop unncecessarily. Deliberate impedences are put in isles to slow you down.
5) Don't take a cart or basket unless you really can't carry what is written on your list.
6) Use the bathroom before you go shopping. They place the restrooms to make you walk as far as possible past tempting impulse products.
A couple of my own...
7) Eat before you shop, never go to a grocery supermarket when hungry.
8) Take cash, just as much as you need and no more, and use the cash only fast checkout.
Perhaps someone who knows the systems they use in detail should write a piece of open source software in their spare time to calculate the optimal path through a store
I wonder how much this software costs. Does everyone pay the same price for it?
The more you regulate a company, the worse its products become.
Lots of companies have been using it for a long time.
s /priceoptimization.mspx
I have used a prize optimization solutions based of MS SQL Server back in 2005
http://www.microsoft.com/industry/retail/solution
I have a better way to game the system: don't participate in "the system". Shop at your friendly local, independent retailers. In our store, we put things where people can find them, and price them competitively.
I don't respond to AC's.
You see. A lot of stores would like to charge each and every customer a different price. Those prices being the set that maximizes revenues from that particular customer. But in practice that is very difficult. Changing differnet prices on the same item for every sale would be cumbersome, and customers who see the person in front of them get a better deal than they do might get pissed. The stores response to this is customer "value" cards.
In an idea scenario, they set all prices on the high end - but then give the customers "value" cards that offer varing discounts and rewards so as to optimize sales and profit. For example, if they know you won't pay more than $2 for a soda, then your soda will always be $2. For example, they might do something like use buying habits track your period. If you buy tampons on the day of your period - you will get reamed hard because they know you need them right now, but if you buy them in the off cycle then you get a good deal. If you buy just milk in the early morning, you will get reamed hard because they know you might need it for breakfast right then, but if you buy it later on you will get a competitive discount. If you buy a phone today, but the last phone you bought was two years ago and had a two year average lifespan, then you get reamed hard because they know you need a replacement right now. Otherwise you get a deal. If you buy condoms on friday night, you get a nailed hard, but if you buy them on wednesday morning you get a great deal.
Yes, you are right that they are separate but related. (Your post is not rude and I hope my response is not rude, either). The lead example concerned pricing of drills. The three models (perhaps from three different makers) define different market segments as far as the retailer is concerned. Optimizing the price on the three models gives the retailer a chance to maximize both revenues and profits even if the retailer doesn't do anything to market the products differently. Similarly, markdown optimization is both a price optimization and a segmentation issue -- segmenting the "I'll pay anything to be the first person to own this" customers from the "I'll buy it later when its discounted" customers.
The classic example, that I was thinking of, is the revenue management strategies of airlines that attempt to optimize prices across presumed underlying segments of price-sensitive leisure travelers versus price-insensitive business travelers. Technically, it's the identical seat that's being sold for radically different prices (up to 10X different) depending on issues such as how the customer buys the ticket, when they buy the ticket, whether they book a saturday-night stay, etc. The result is that the business customers pay for the plane and the vacation travelers only pay for the fuel and variable costs. The ability to differentiate does benefit the business customers because the added volume of travelers means a more frequent schedule of flights, larger planes, and more destinations.
You are right, though, that true market segmentation involves much more than just price optimization.
Two wrongs don't make a right, but three lefts do.
It's not just price optimization that's relatively easy to learn. There's also location theory and optimization (e.g., retail, essential and emergency services, noxious facilities), routing and scheduling, order and inventory levels, ... The math is relatively simple (most of the time) and if you can set up the problems there are some decent OS software packages with very good solvers. COIN-OR comes to mind but I'm sure there are several.
The real issues are getting the data and interpreting the results. I've got a land use decision model for urban fringe areas that's doing a reasonable job of presenting sets of potential solutions for decision-makers. The tough stuff is the pre-processing, e.g., defining what's feasible and desirable, and interpreting the results, e.g., what solutions are or aren't significantly different. For this type of work geographic data is readily available. I doubt that will be the case for most business decisions.
I was the purchasing agent for a chain of auto parts stores, and we used a method called GROI: Gross Return on Inventory.
The original pricing theory in traditional auto parts stores was based on four "turns" per year: after opening a parts store and filling it up with stuff to sell, you needed to sell each stock number four times, at 35% gross profit, to make an adequate gross profit to cover your expenses -- and pay off your inventory in twelve months.
This resulted in some items being priced much higher than at mass retailers, and caused stores to lose sales on popular items: people would go elsewhere for oil, antifreeze, and the most common spark plugs, brake pads and filters, because they were much less expensive at places like Pep Boys.
The GROI method constantly recalculated sales and adjusted the prices downward on popular items, thus increasing sales and lowering the prices still further. For example, the best-selling oil filters would sell at under 10% gross profit, but we would sell out our inventory of those items twelve to fifteen times per year... thus making a larger profit on the initial purchases we had made to stock a new store. By setting a minimum gross profit percentage, runaway sales on an item would result in higher profits instead of ever-lower margins.
This was all calculated by an incredibly expensive 200 MHz Pentium Pro box, running proprietary software atop SCO Unix.
(I was the only one who could work the thing... which led to me running a newly purchased Netware 4.1 network in the chain's offices... which led me out of that filthy auto parts business altogether, thank Jeebus).
Advice for Wal*Mart shoppers:
Never buy produce or fresh bakery goods at a Wal*Mart. The premium at the true grocery stores often corresponds to the produce & bakery goods actually being better quality.
Also, since you actually have a choice, try to memorize the routine sale prices at your other local stores. Sales tend to be cyclical. Wal*Mart has lower prices on the most popular items; for more obscure stuff they can go higher because those items are harder to find elsewhere, or fewer people are looking for them. I learned this when trying to buy a rare iron syrup (which could've had a proof number).
Wal*Mart is a good place to shop for low prices, but other places have different selection, and it's a good idea to give at least token support to its competition.
There is a fine line between recklessness and courage... -- Paul McCartney
First, a disclaimer. I was employee #4 at KhiMetrics, the company founded by Ken and Tim Ouimet (employees #1 and #2). They're mentioned in TFA. SAP bought KhiMetrics in January of 2006. Ken had been my office-mate in grad school. That said, I haven't seen Ken and Tim in years, and I have no financial stake in KhiMetrics or SAP anymore (SAP bought out the KhiMetrics stockholders with money, not shares of SAP stock).
Yes, it's true that humans doing pricing try to do the same things. But the thing is that software can do things a human mind cannot. Yes, the opposite is also true, but here software has a lot of advantages. In the case of the KhiMetrics (now SAP) software, it works on the category level, optimizing profit for the category as a whole, which can include taking losses on individual items. The software never makes the common mistakes human beings make. For example, different "flavors" of the same size package of the same product should come out at the same price, and the unit price of a given item should go down as the amount bought increases. I can tell you that I have seen examples where humans have screwed this up this week. When there are two sizes of a given product, let's say a certain laundry detergent, then the price per weight of the larger package better be less than the price per weight of the smaller package, or there's never any incentive for the customer to buy the larger package. Still, I see examples where the pricers have gotten this wrong. I've even asked people at the stores if they were trying to move the smaller packages because of having too much of that size in stock or something, and they told me that no, they had no such problem.
The other thing is that the KhiMetrics software uses actual sales data to determine how sensitive the customers are to the price of a given product. This can be done down to the SKU (individual item) level in the product dimension and down to the level of customers of a specific store in the geographic dimension. In other words, the KhiMetrics software is capable of determining the sensitivity of the customers of each individual store to the price of a specific product. No human being could do that at all, much less in the time the KhiMetrics software can do it. Even with a pricing team for each category in each store, which would end up costing a fortune in human resource costs, the result would not be as good as what KhiMetrics can deliver. Additionally, since the Ouimets "grew up in retail," the KhiMetrics software, since the beginning, has been compatible with things like Category Management and Efficient Replenishment, and able to take into account things like having different goals for different products in a category (loss leader, profit generator, traffic generator, etc.). The software takes into account complex factors like seasonality, promotion, and product visibility. Since I have a reasonably good idea of the internal workings of the software, I can tell you with some confidence that I, a Ph.D. in theoretical physics, would not even want to try to tackle the problem of optimizing the prices for a subcategory of 20 products in a single store, much less the dozens of categories and tens of thousands of SKUs in the dozens of stores in a retail chain. KhiMetrics can do all that, basing itself on years of actual sales data, before breakfast.
There are experienced people in retail who are good at such things, but the software was created with people that have the same level of understanding of retail pricing, plus it has all the advantages of being able to do high-speed computerized analysis of huge amounts of price and sales data. I don't work for KhiMetrics anymore, nor for SAP, but I can say that if I were working in a retail company, I would definitely want us to be using software like this for pricing. And experienced retail people agree with me. One thing we saw back when I was with K
"It is nice to know that the computer understands the problem. But I would like to understand it too." --Eugene Wigner
Although, WalMart is NOT a good example of profit optimization; there strategy is more likely to take cost decreases and pass them on. Which may not optimize the short term. And certainly only works if you are the more efficient retailer, which scale tends to help. But it is very pre-consumer and relatively dificult to compete against based upon price.
A MUCH better example would be airlines; where what is practically the same product - a coach seat on the same flight might go for 5 to 8 price-points of $100 to $1000 each way. Grocery and department stores might sell comparable products for 10 to 50% more than WalMart but rarely can get the 1000%
"NOTHING about the individual case; we're just plain hard to predict..."
Don't care. While the individual case can't be modeled the group case can be. Plus a WM can price a DVD at one price in LA and another in SF, seen the trend, reverse it next week, see what happens then and adjust the price in all of the stores accordingly. Do variations of the above for 1,900 stores nationwide. See what happens when you advertise price A vs. price B and seel how many are sold. Do variations of the above for 1,900 stores nationwide. Compare against the entire history of every DVD you've ever sold. Correlate against box-office receipts. Find out that I'll sell more copies at $14.95 and make 5% more than if I sell each copy at $14.99.
Do the same thing for 140,000 other products. Rinse. Repeat.
"people do indeed communicate and discuss the prices / merits of their purchases"
Indeed. When was the last time you and your friends discussed the fact that baked beans are $1.23 at store A and $1.25 at store B, while the dial soap was $1.45 vs $1.42?
"Stores that do this well find that they can reduce the number of SKUs they stock and make even more by only stocking the items with the highest markup / fastest turnover."
They're smarter than that. By your logic my grocery store would only sell sugar water. They don't. And they understand long-tail economics and price elasticity and loss-leaders much, much better than you do.
"Nobody likes to be manipulated."
True. However, if you're selling your house or car (and all other things equal) you probably want the most money you can get for it. If you're on the other side, you want that house or car for the fewest number of your hard earned dollars as possible. Your employer would probably like to get the same work done for less, you, OTOH, would like to do the same work for more money, benefits, vacation days, whatever.
Some people will pay more for convenience, others will drive ten miles out of their way for bargins. Some people think Apples are more than worth the price, while others think that anyone who buys anything other than the cheapest beige box on the shelf is nuts. Horrses for courses.
As long as all of these variables are in play prices will be adjusted and raised and lowered and fought over... and in some cases, manipulated.
Any sect, cult, or religion will legislate its creed into law if it acquires the political power to do so.