Introduction

Price points are extremely important in business as their selection can make a huge difference in gross revenue.

To illustrate the importance of price point, consider a product which might be priced between $5 and $20. The number of customers that purchase at these different prices will change for different reasons:

$/month	% that buy	score	comment
$5	5%	25
$10	7%	70	higher price sometimes yields higher perceived value by customer
$15	6%	90	higher price eventually drives customers away
$20	3%	60	higher price eventually causes many customers to balk

So $15 is head and shoulders above the rest: more that 29% higher grossing than the second most profitable price point.

Selecting a price point is done by almost all major businesses that have goods and services to sell. Often a combination of surveys, consultants, spreadsheets and software are employed by businesses large enough to afford all that. Price testing is also conducted.

It's notable that Amazon's initial random price testing caused an uproar among customers. Some reports state that Amazon refined their approach by giving customers the lowest price available at the end of the test and others state that Amazon ceased testing indefinitely. See: http://www.google.com/search?hl=en&q=amazon+dynamic+pricing

Proposed Approach

A dynamic web site offers a unique environment to determing price points dynamically, e.g., to adjust prices automatically as they need adjusting.

One approach could be to set a base price and after a certain number of potential sales, adjust it. If the last batch scored less than its predecessor, the price drops. Otherwise, the price increases.

A period is defined in terms of the number of visitors that actually get to the pricing information for the opportunity to purchase. We'll call those visitors the potential customers. The purchase rate is the # of customers divided by the # of potential customers. The purchase rate times the price gives the score for that period.

Here are some fragmented thoughts expressed in Python:

bp = 4.95   # base price
inc = 1.00  # increment amount
dec = 1.00  # decrement amount
period = 100        # number of visitors before analyzing

periods = []        # list of periods

class Period:
    """
    attributes:
        date  - the date the period started
        score - the important number. how well this period did
        rate  - rate of customers who saw pricing information and purchased
        parameters - dictionary of parameters at the time such as bp, inc, dec, etc.
    """
    pass

if len(periods)>1:
    prev = period[-1]
    beforePrev = period[-2]
    if prev.score>beforePrev.score:
        newPrice = price + inc
    elif prev.score<beforePrev.score:
        newPrice = price - dec

By giving all potential customers the same price for a given period of time, customers are treated fairly (contrasted with the original Amazon approach). The period price adjustments are no different than manual adjustments made by other companies for their products. What's special is that the adjustments are automatic and precise.

What's also special is that the price adjustments are made whenever needed. The underlying factors could be all kinds of things:

• change in product quality

• change in economy

• appearance or disappearance of a competitor

• presentation of product

• and much, much more

But the computation doesn't need to know the factors and will therefore, never miss any, because the score is the bottom line.

Additional ideas to refine this approach are:

• Have a minimum number of days for the current price point, at least 1. This would avoid customers experiencing changing prices in a single day, which could make them feel uncomfortable.

• Likewise, have a maximum number of days in order to put a limit to the timewise duration of a pricing period.

• Refrain from price adjustments if the score difference was less than some minimum percentage such as 1%.

• Adjust prices dynamically according to the difference in scores. Here's a thought:

  factor = 0.5
  scoreRelDiff = (period[-1].score - period[-2].score) / period[-2].score
  price += price*scoreRelDiff*factor

• Another approach could use multiple moving averages. For example, if the average score of the past 5 days is higher than that of the past 10 days, increase the price. This "moving average" approach is used by some stock market investors.

• Some sites would need to be cautious of the cost of goods sold (COGS) in order to avoid selling product at a loss.

Summary

The advantages of Dynamic Pricing are:

• maximizing profits

• adjusting price whenever needed

• automating price adjustment

• treating customers in a given period equally

There's a huge number of approaches that could be taken and lots of different topics to be discussed. Feel free to add to this Wiki page.

-- ChuckEsterbrook - 17 Mar 2002

---

This is brilliant -- thanks Chuck! JamesBecker