You are on page 1of 6

Building a predictive model to recommend tips to

reduce rejections and wastage


Insights from Data Exploration

● The highest sum of rejection amount is from the side of Skanski US and in the inventory category
of Electronics
● The largest floor space available is also with Skanski US and also over the past one year it has
delivered the highest number of goods
● The highest sum of rejection has come from the zip code 78015 which is catered by Whiting turn
contracting
Linear regression

● “Y” variable: Amount rejected


● “X” variable: Planner, Warehouse Square feet area, Sub contractors, Approved product categories
● After checking multiple models for linear regression, above mentioned independent variables found to be significantly impacting
dependent variable “amount rejected” having R-squared value of 0.6135
● Keeping all other values constant with one unit increase in approved product categories will decrease rejected amount by 132 times
● Keeping all other values constant with one unit increase in sub contractors will increase rejected amount by 211 times.
● Warehouse area have no significant impact on rejection amount
Logistics regression

● Higher the sub-contractors higher will be the chance that rejection is within acceptable limits
● Lower the approved product categories higher is the chance that rejection is within acceptable limits
● Warehouse space has no significant impact on the rejection
● Looking at the p values of Planners we can see that all the planners affect the final acceptance of
rejection amount
● From the coefficient values of planners choosing Shanski US planner will help to keep rejection within
acceptable limits
Similarities and Differences

● Both utilize a linear equation to arrive at predictions and both are parametric predictions
● However if we see, differences are more between the two types of regression than similarities- here if we
see, we get a continuous equation as a result for linear regression with values possible anywhere between
the number line whereas in logistics regression the result is continuous only between 0 and 1
● Here, according to results obtained from two types of regression, warehouse has a significant impact
based on linear regression but has no significant impact on rejection control based on logistics regression
● Through linear regression we got to know which variables can be included in the parametric equation to
calculate amount rejected but through logistics regression we also got to know the individual impacts of
independent variables on the dependent ones
Recommendation

● Warehouse space has no significance on the rejection, so better to try using space optimally and
not to invest on increasing warehouse space unless needed
● Rather than increasing the product categories to serve larger base of customers, better
concentrate on serving the customers with minimal but quality product categories
● Increase the use of sub-contractors
● Planner Shankski US is more preferable since they can manage the vendors better leading to less
rejection rates among the three planners

You might also like