Data driven Cost Estimation Optimizing Categories

How Artificial should Procurement Intelligence be ?

As the COVID-19 Pandemic changes all aspects of the business landscape, the Procurement function is taking center-stage in areas of risk mitigation, cost control, innovation, and ensuring overall business continuity. As companies around the world search for levers to gain a competitive advantage in this challenging environment, supply cost reduction is proving key to protecting and boosting operating margins.

With this in mind, companies need to adopt a more data-driven, strategic methodology to derive savings from these strategic partners. This is when Machine Learning and data driven approaches can renew the function and accelerate efficiency.

Traditional cost reduction measures are becoming less and less effective, as supplier networks become more strategic in nature. Value within the supply chain is no longer solely a function of the lowest product or service provided.

Therefore, it should not come as a surprise to see supplier leverage moving away from the purchasing organization and becoming more neutral (a zero-leverage scenario based on the give and take of vendor/supplier dynamics).  

 

Data empowered Procurement

Three ways Data can add to the function

1- Estimate the cost of a new product or service with near-perfect  accuracy
2- Identify savings opportunities across entire product categories
3- Provide live anonymous benchmarks on specific categories
 

A Fasteners' Category Case

Let’s assume we are comparing two screws within your Fasteners purchase category. 

Both screws have identical raw material, head type, and finishing, and very comparable length, width, heat tolerance, and thread length, and are purchased in similar quantities over the year, however the price of one screw is 25% more expensive than the other.  That 25% difference in price cannot be explained based on its attributes, and therefore becomes the basis for cost reduction at the SKU level. 

This simple example, which most likely exists several times over within your purchase portfolio, describes precisely what easyKost uncovers for its users.  It is a fast, accurate, easy-to-use tool that delivers serious results. 

 

 
 Mr Jason Mallory

KEPLER N-America

The present case study showed an overall savings rate of 8% year-over-year within the Fasteners purchase category vs. 2% in previous years. 

Cost reduction : based on data? A simple ask

With a data-driven approach, the ask is simple. Instead of a blind request of 10% reduction in price across the entire portfolio, you are simply asking your suppliers to deploy a consistent pricing model for the products you purchase. Kepler has found this is much more well-received by suppliers, as it is a focused ask rooted in in-depth analysis of their own pricing structure.

…And a leading U.S vehicle manufacturer

  • Over 8 billions of revenue
  • 13 000 employees
  • Spend : $5 Billions

As a leader in the production of light, medium, and heavy-duty trucks, buses, emergency and military vehicles, this client was looking for ways to optimize their purchase portfolio in accordance with their Procurement Transformation efforts. 

KEPLER and easyKost associating

EasyKost is cost modeling software utilizing machine learning and your own purchase data to identify pricing inconsistencies at the SKU/PN level. With KEPLER, they are working to help the client : 

  • establish cost savings, 
  • consolidate their supplier base 
  • reduce the number of Part Numbers in their overall portfolio.
 

5 step process to rapidly identify and realize savings

  1. Preparation
  2. Data Collection
  3. Data  Modeling / gap analysis
  4.  Validation of opportunities
  5. Execution

Smart Value Initiative

First savings captured after 2 months (negotiations) with a full savings scenario validated under 3 months.

Inputs required to kick off the preliminary analysis

Spend Data Collection

Preliminary Analysis Cost reduction

Category initial questionnaire

Machine Learning Costing KEPLER

Key criteria to assess addressability and opportunity of the category, general questions regarding:

  • Products
  • Data accessibility
  • Market
  • Strategy
 

Introducting the future of Costing

Turbo-charging cost-reduction with data

Kepler Consulting has partnered with easyKost, a leading Costing software Company that utilizes Machine Learning to identify savings opportunities within a company’s existing purchase portfolio.  

Using a Random-Forest based algorithm, the solution compares your own internal purchasing data against the attributes of the products you purchase to identify inconsistencies in your supplier’s pricing.  

 

The discovered inconsistencies are then used as the basis for renegotiating pricing with suppliers.  

This thoughtful, quantitative approach is helping numerous companies around the world turbo-charge their savings achievement, streamline the supplier base and create an overall more efficient supply chain.  

 

Possible use of cost estimating & data mining softwares

  1. Estimation of the cost of a new product/service with near-perfect  accuracy
  2. Identification of savings opportunities across your product categories
  3. Provision of real time & anonymous benchmarks on specific categories

Random forest : cost estimation based on drivers

The Software analyzes all the correlations between the actual purchase prices and the product or service “cost drivers” in order to identify inconsistencies and optimization opportunities.

Cost is estimated through a thousand decision trees in a random forest. 

The method leads to a 60%+ increased accuracy compared to traditional statistical methods.

Random Forest Cost Modeling

Identifying cost-drivers

Differentiating among the Data sources

Cost Drivers

A technical characteristic: the weight, the length, the width or diameter…

A function: E.g. the function “Screwdriver (yes/no)” of a drill…

A country or region: they inevitably influence the cost of the product.

A supplier: you will not pay the same price for the same product.

A percentage: even if it is preferable to use a value, a factor representing a percentage can be used as cost driver.

Not Cost Drivers

Indicators that vary over time or would not include at least two separate values : 

  • a time-variable factor
  • an exchange rate
  • a unique identifier
  • a material cost
  • a date, an incoterm
  • a currency

Collecting Data

By extracting attributes from Client drawings, our off-shore data team based in Chennai India, creates a database that captures all cost drivers per part number. 

 

  • Cost drivers are defined by family using technical expertise
  • For each family, a database is created capturing cost drivers by part
  • Cost drivers are normalized to improve the model
  • The full catalog of attributes is provided to Client to close the project
 
Pricing supplier Artificial Intelligence

Modeling

Three savings generation strategies are used to maximize results while maintaining  the right amount of work per supplier type. 

  1. Supplier Negotiations:  Strategic Suppliers
  2. Supplier Re-allocation & Consolidation: Transactional suppliers
  3. Tail Spend Optimization & Consolidation: Tail suppliers

Focus on supplier reallocation

Machine Learning estimations predict the best price that key strategic suppliers should be able to offer on existing parts that are currently supplied by transactional suppliers.

Price inconsistency analysis

Costing Modeling Price Inconsistency

Estimate transactional with highest cost reduction opportunity. 

Screen Capture : EasyKost software

Supplier change simulation

Based on Batch Estimations run on the exhaust models, the new state of exhaust would shift heavily away from Suppliers 1 & 2 into Suppliers 3 & 5. Total savings if every part number target was achieved would be $2.04M. This would require 596 parts to be moved to new suppliers. 

Supplier Change Simulation KEPLER AI Procurement
Supplier reallocation

Testing the model and entering the validation phase

E.g. Commodity 12 - Before

Original State : $1.7M

Key supplier A ……………………32 PN

Key supplier B ……………………59 PN

22 Non-preferred suppliers .96 PN

 

E.g. Commodity 12 - After

Optimized State : $1.5M

Key supplier A …………………… 59 PN

  • 8% reduction of incumbent business
  • 7% reduction on $180K new awarded

Key supplier B …………………… 128 PN

  • 19% reduction of incumbent business 
  • 21% reduction on $192K new awarded 
 

Validating the Model

Out of a $93 million spend in scope, the analysis and modeling estimated a savings opportunity of $5M.

Supplier Optimization Reallocation
Supplyer Reallocation Optimization KEPLER

Executing the model

Focus on Supplier A

Focus Supplier Costing AI

Highlighting inconsistencies

After reviewing the entirety of our product database and comparing attributes from Supplier A, we have validated that the pricing model demonstrated the capability of achieving accurate targets: 

Price Inconsistency Costing Supplier Model

After reviewing the entirety of our product database and comparing attributes from Supplier A, we have validated that the pricing model demonstrated the capability of achieving accurate targets: 

Price Inconsistency Costing Supplier Model2

Getting the Results

The Washers Example

Category Optimization Costing KEPLER

Client was challenged to consolidate the Washers product family.  KEPLER and easyKost were able to achieve 183% of the savings target and eliminate more than 70% of vendors.

Few project’s additional numbers

  • Savings Target: $5MM / $93MM
  • Mission ROI: 7.1x
  • Current Direct Spend Addressed: $500MM  (15% total)

Going Further

Permanent Webinar Access

Downloadable Version