Analysis of Account Level Product Sales

BI Solutions
October, 2015


Executive Summary


  • Objectives:
    • evaluate the value of available account level Product sales data
    • estimate impact of unit price and number of units on sales
    • determine account level potential growth of sales (if it is possible)
  • Data: account level Product quarterly sales for the last 15 quarters
  • Findings:
    • only one account (Distributors) out of four available has enough data for the analysis
    • Correlation between product sales and number of units sold is very strong: 0.92
    • To maximize sales the unit price should be within the range of 500 – 530 (in Euros) and the number of units should be greater than 15,000
      • Additional decrease in unit price below 530 does not lead to increase in sales
      • Diminishing returns in sales lies between 7 and 8 MM, and it is practically impossible to exceed 8 MM in sales by varying unit price alone
      • Lagged predictors are insignificant for dependent variable Net Sales across different models
      • On average,  increase in unit price by one (in Ireland currency) leads to decrease of number of units sold by 37
      • On average,  simultaneous increase of unit price by one and number of units sold by one lead to product sale increase by 4,810 in Ireland currency
      • Taking into account upward sales trend, increase of unit price by one leads to product sale increase by 25,542 in Ireland currency within the same quarter



Dynamics of Gross Unit Number has positive long-term trend



unit number



Dynamics of Net Price has negative long-term trend



net price



Dynamics of Net sales by quarter: during last 8 quarters sales fluctuates around diminishing returns level of 7.5 MM




diminishing returns



Maximum sales occurs when the number of units is greater than 1500 AND unit price is less than 530




area of equal sales



Number of units has long-term negative trend as a function of unit price




long term negative trend



Best Static Models

  • Linear robust regression model (Generalized Maximum Entropy Estimator) for dependent variable Product Sales:

         Product Sales = 2,603,640 4660.45*Unit Price 149.01*Units Number
               Adjusted R-square: 0.36
               All parameters are highly significant: p-value < 0.001

  • Linear robust regression model (MM Estimator) for dependent variable Unit Number:

                  Unit Number = 32,739.7 -  36.8*Unit Price
                   R-square: 0.12


Best Dynamic Model


  • First order Autoregressive model with a trend:

        Product Sales = -9,290,156 25,542 *Unit Price 302,346 *Quarter

All parameters are highly significant: p-value < 0.05, R-square: 0.78