March 5, 2019
As competition continues to heat up and giants like Amazon gobble up market share, the pressure is mounting on retail executives to find new means of growth. For those who haven’t already adopted it, dynamic pricing may be this year’s solution to the growth conundrum.
Dynamic pricing is a variable price strategy aimed at fulfilling a particular business goal like maximizing profits or sales volume. Simply put, prices are adjusted whenever there is an expectation of business benefit.
Dynamic pricing does not mean rapid or frequent price changes—prices are only adjusted when they will benefit the business’ bottom or top line. Furthermore dynamic pricing does not require significant manual labor.
Modern day dynamic pricing is a result of the marriage of big data with machine learning. In order to optimize margins or revenues, learning algorithms change prices based on data measuring things like supply and demand, seasonality and competitor pricing.
Many algorithms also take the individual customer’s data into account with the goal of calculating the customer’s willingness to pay. Prices can therefore vary from customer to customer, by time of day, by season, by store, and so forth.
It is a common misconception that dynamic pricing is meant only for online retail. Modern day retail is omnichannel and pricing must follow suit. Brick and mortar stores have just as much to gain from dynamic pricing as retail websites.
Another common misconception is that dynamic pricing means rapid or frequent price changes—prices are only adjusted when they will benefit the business’ bottom or top line. Furthermore dynamic pricing does not require significant manual labor—the algorithm takes care of all price recommendations.
Amazon’s dynamic pricing strategy
Amazon sets the bar for success in dynamic pricing. Besides considerations of supply and demand, seasonality and the like, Amazon’s algorithm aims to both undercut competitor pricing and maximize margins.
That may seem contradictory, but it is standard operating procedure for modern dynamic pricing algorithms. In short, the algorithm identifies the most price sensitive (usually the most popular) products and prices them to undercut competitors. Meanwhile the algorithm also identifies less price sensitive (usually less popular) products and prices them to maximize margins.
In other words, a good dynamic pricing algorithm is constantly updating and exploiting calculations of price elasticity for each product. This is a very difficult job that requires large data sets and advanced machine learning capabilities.
The result of a quality dynamic pricing engine, according to a study by McKinsey, is sales growth of 2%-5% and 5%-10% increases in margins, along with higher levels of customer satisfaction through improved price perception on the most competitive items.
Human input into dynamic pricing
As we saw with Amazon’s 24 million dollar book about flies in 2011, algorithms are not perfect. Though pricing algorithms today are light-years ahead of where they were seven years ago, human input and oversight is still priceless.
Oftentimes, the people involved in setting prices have no idea how the algorithm works or why it arrives at seemingly bizarre price recommendations. This should not be the case.
At the very least, category managers and pricing managers should set minimum and maximum prices per product. Ideally, they should be intimately involved in developing the dynamic pricing algorithms so they have a good understanding of why the algorithm comes up with particular price change recommendations.
How Evo helps
Most retailers can’t, and should not, invest in developing dynamic pricing algorithms in house. Rather they should look to companies that have years of experience perfecting these algorithms.
At Evo, we forecast sales using a vast amount of data including seasonality, store size, store location, past revenues, changes in product mix, price changes, discounting, market trends, competition and weather. Our fine-tuned sales forecasts attach a probability of sale to each particular item depending on where and when it is sold.
We use three methods to formulate price recommendations. The first is based on SKU trends. Our engine responds to any rapid changes in sales, especially of the best -or worst- performing products. Strong performers can sustain a price increase while poor performers get a price reduction or promotion.
Second, we simulate sales at new prices versus baseline forecasts of old prices. This allows our algorithm to make price-elasticity-driven recommendations when opportunities to increase overall margins are identified.
Third, our machine learning algorithm takes the results of past price changes into account. Generally speaking, our engine learns from each price change to better forecast optimal pricing.
Instead of developing a one-size-fits-all solution, we adjust our algorithms to fit the specific circumstances of the retail client. We insist upon brand input: clients should remain in frequent contact with our staff so we can continue to optimize their specific dynamic pricing engine.
As the sea of available data continues to expand and the quality of machine learning continues to improve, dynamic pricing will only get better. Retailers that stick with static pricing will ultimately be destroyed by the likes of Amazon.
About the author
Will Freeman is a content expert at Evo.
He is a former economic journalist and part-time entrepreneur.
His interests include economic development, China, India, cryptocurrency and blockchain, and financial technology in general.