The Analytics Compass Blog

Twice weekly articles to help SMB companies optimize business performance with data analytics and to improve their analytics expertise.

Subscribe via E-mail

Your email:

Search SimaFore

FREE SMB Survey Report

describe the image
  • Our report (pdf) is a survey of analytics usage and needs of more than 100 SMBs
  • Find out which analytics applications have
    • Highest value
    • Most demand
    • Best ROI

Affordable Analytics for SMB

 

Browse by Tag

Current Articles | RSS Feed RSS Feed

How to improve cost modeling with predictive analytics: part 3 of 3

  
  
  

This is the final of a three part interview series which discusses how one mid-sized cost forecasting predictive analytics 300manufacturing company is using advanced analytics techniques to address very common business problems - cost modeling and cost forecasting.

Recapping part 1, L&L Products wants to answer two key questions using predictive analytics: can a model be built that connects cost inputs (raw materials) to manufacturing costs? Can a model be built that forecasts part costs based on fluctuating input costs? In part 2, we discussed how they put together the data and what tools they used for predictive analytics.

In this final part, we talk with Steve Reagan of L&L Products to understand how their business teams are actually utilizing the model, the ROI from using analytics and future plans for integrating analytics more firmly into their business.

Bala Deshpande: Can you explain how your Purchasing team is employing your model? Do they need support from time to time to update the model or to deploy it?

Steven Reagan: We were able to accomplish two major tasks for our purchasing team. One, we reduced the 40-50 indices and cost metrics they were originally tracking to 4 or 5 most important ones using feature selection tools like principal component analysis. Two, we were also able to give them some predictive capability that was not based on past experience or gut-feel.

BD: Steve, tell us about what analytics techniques you tried on the data and which one did you finally end up using.

SR: We initially tried an artificial neural network model, but because ANNs are black boxes, the business team was not very confident about using it. The final predictive model was a linear regression model that had 3 key predictors which influence the price of L&L’s raw material cost (down from 40-50 known from experience). This was very important for our purchasing team because they are able to understand and relate to how these key variables are affecting the raw material price.

BD: In terms of overall value addition, would it be correct to say that your efforts have led to making part price forecasting cheaper, faster or better? Can you explain how?

SR: As a result of the feature selection analyses, our purchasing manager now needs to track only 3 key indices instead of 50. This was a key moment for the purchasing team. They understand now what are the few things that most directly relate to raw materials cost. They can now focus their attention on a smaller subset of factors to make better estimates of when to purchase raw materials and how much to purchase. Secondly, part cost forecasting has been made more quantitative, less subjective.

BD: Can you tell us a bit about your upcoming cross-functional analytics projects?

SR: We are now leading a task force for developing quoting cost model that sales teams can use with our customers. The purpose would be to generate a data warehouse accumulated from Oracle, individual spreadsheets, and disparate human knowledge bases that can be used for generating a more accurate price quote that is less burdensome to our sales force. It takes 4-5 days to generate a reasonable quote today. We want to bring this down to less than 1 day. With predictive analytics we believe this is possible because we have historical data for similar parts (we can use cluster analysis tools here, for example).

BD: Steve, thank you very much for sharing your innovative approach to problem solving. I think many other SMBs in manufacturing face very similar issues and they can take a cue from your initiative. We, at SimaFore, believe this process – of utilizing cheaply available data and tools, and combining with some specialist support to solve business problems – is vital in democratizing business analytics today. We think this process can be systematically replicated in other business critical functions, for example, in Marketing or Finance, to great benefit for SMBs in the country.

SR: Thank you!

Are you a small or medium business that wants to employ advanced analytics and don't know where to start? Sign up for our analytics portal, visTASC that is targeted at companies like yours. visTASC provides tools to connect business problems with analytics techniques.

vistasc analytics made simple

Comments

Currently, there are no comments. Be the first to post one!
Post Comment
Name
 *
Email
 *
Website (optional)
Comment
 *

Allowed tags: <a> link, <b> bold, <i> italics