Cost Excellence



Bring real cost reduction using machine learning through detailed insight into procurement spends

Weak economic forecast, wars, Brexit is compounding the issues faced by corporates amidst global uncertainties. Competition, newer technologies, ever increasing regulations adds to turbulence. As uncertainties loom large, the first recourse available is cost-cutting. Corporates often resort to knee jerk reaction with short term view. This includes layoffs, shuttering facilities; reduce employee benefits and the like.

Short term benefits seldom give desirable results. Such short term measures often lead to disaster when the tide turns in favor. It is important, therefore, to reduce cost on sustainable basis. This involves improving efficiency in operations and improving effectiveness.

How do we help?

The cost excellence program has to be well thought and planned with all stakeholders involved. One critical element in cost excellence program is classification and categorization of costs. Even mature companies with good controls and practices fail to capture cost correctly. Below are examples of errors:

  • Incorrect cost center

  • Incorrect GL code used

  • Wrong classification of expenses (for e.g. a sub-contractor’s cost is considered as consulting expenses)

For successful Cost Excellence project, it is imperative that correct data is available. Manually reviewing and correcting millions of records needs humungous effort. Not only it is an expensive proposition but consumes too much time.

Quadrisk can develop an algorithm using Machine Learning to classify expenses. Machine Learning iteratively learns from data. An initial logic is provided in the machine learning script. Based on various combination (Vendor, Amount, GL code etc.), the script can provide correct classification.

How does it work?

  • Take sample data of 200 or any reasonable number of records

  • Manually classify costs for these records

  • Create script and run it against sample records

  • Validate the output from machine learning with manual classification

  • Fine tune scripts until results are satisfactory

  • Run the script against the complete data