Price Features

Why do we need different price features?

• As shown in the above image, raw price data changes can happen due to periodic price changes as well as short term fluctuations between two periodic price changes.
• Hence, we need to intelligently seperate out these phenomenons in the raw price and model them separately to capture accurate impacts and have realistic interpretations on revenue.

Different Price Features

• The first step is to identify when periodic price changes happened in the raw price. This is done using ruptures library where the price within each stable region is considered to be the baseline price.
• The next step is the creation of features. We have 5 different price features currently.
  1. Discount: This captures the ratio of raw price to baseline price within each stable region and is indicative of the overall price elasticity as shown below.
  2. Short Term Deviation: This captures the ratio of price within each stable region with respect to the smoothed price within the same stable region and is indicative of short term price elasiticity as shown below.
  3. PINC Increase: This captures the ratio of two baseline prices as learnt by the ruptures library. We model the actual PINC increase as well as the leads and lags of PINC increase because the wholesalers are made aware of any such increases in advance and hence they tend to stock up more beforehand and it takes time for that stock to come back to normalcy as shown below.
  4. Price Ratio Across Brands: This captures how the fluctuations of the average price of other brands sold by the wholesaler affect the sales of the brand into consideration sold by the same wholesaler and is measured as the ratio of the price of the brand into consideration with respect to the average price of the other brands sold by the same wholesaler. It might happen that even though the raw price of a brand sold by the wholesaler increases over time, still the price of that brand is cheaper compared to other brands sold by the wholesaler and hence this ratio wilk decrease over time resulting in an increased impact on revenue.
  5. Inflation Adjusted Price: This captures how the price of a brand compares with the national inflation and is measured as a ratio of the two. It might happen that a highly priced brand over time seems cheaper when compared to the national iinflation rates and hence can affect the revenue positively.

Modelling Summary

• The impacts of discount and short term deviations are modelled using the price elasticity of demand curve.
• The impacts of PINC increase, price ratio across brands and inflation adjusted price are modelled as having some of direct or inverse linear relationship on revenue using the monotonic unbounded positive curve.
• All the price features have a multiplicative influence on the baseline revenue as well as on the marketing investments.