# Achieving Cost Savings in 2024: Advanced Purchasing Strategies

Advanced Purchasing Dynamics recently surveyed purchasing leaders concerning the challenges they would be facing in 2024. A top concern of the purchasing leaders is that their companies are expecting them to deliver an unachievable amount of savings in 2024.

While savings are always a purchasing priority, in the past few years, Covid shortages coupled with across-the-board inflation shifted Purchasing Team’s focus to ensuring the availability of materials and cost avoidance and mitigation of supplier cost increase requests.

With retreating commodity pricing and a leveling off/reduction in inflation the priority once again is on year over year cost reduction. It seems that many companies are expecting purchasing teams to play “catch-up” and recover economic increases and or offset savings shortfalls from previous years.

In discussions with Purchasing leaders, they identified skills and knowledge as two of the most important reasons they will have difficulty in meeting savings expectations.

However, some buyers are improving their skill sets and achieving savings at the same time. Let’s explore some of the processes these buyers are using:

**Building Cost Catalogs**

Cost catalogs are commodity specific data tables containing SKUs for the commodity and associated pricing. But going beyond pricing Cost Catalogs also include commercial, design and service attributes that should be determining or driving pricing.

Here is a simple example:

Buyers create Cost Catalogs by engaging engineering, operations, and suppliers in discussions about what attributes are driving costs. They then complete a table with the information.

Our experience is that there should be between 12 and 20 cost drivers for most commodities. There are diminishing returns from too many and the possibility of missing key drivers with too few. Not all the SKUs need to be included. Buyers fill their tables from highest spend to lowest and stop when the SKUs start seeming insignificant. For many commodities, 20% of the SKUs represent 80% of the spend and buyers stop there.

Buyers who build Cost Catalogs gain knowledge about the cost drivers on their commodities while identifying savings opportunities. Here are some examples:

- A buyer of O-rings identified 2 parts with the same physical and commercial attributes except one was specified to be blue in color and the other black. The blue one had a 20% pricing premium. Since the O-rings were used internally in the buyer’s company products replacing blue O-rings with black ones was an easy decision that resulted in a $260,000 annual savings.
- A Chinese buyer enrolled in Cost Management Certification developed a Cost Catalog for contract software programmers for his certification project.

In the columns, he placed attributes that would drive the hourly rate for each including items like region, years of experience, programming languages, contracting firm, mark-ups. Building the catalog got him to ask the contracting firms about variances rates per hour, mark-ups, tax rates in different regions and quantity discounts provided by some but not all firms. In the end the buyer saved over $250,000 through negotiations, consolidating spend and moving to regions with lower taxes.

*Creating and maintaining Cost Catalogs serves the following purposes:*

- They can be used for quick costing. Often the “what will it cost?” question can be answered by sorting the table using the attributes of the part or service requiring estimation.
- They provide a good method to identify pricing that may be out of line. If three out of four parts with similar attributes have similar pricing, what is going on with the one part priced higher?
- The data contained in cost tables can be used to develop regression-based cost models.
*(see linear and multivariable regression models below).*

**Creating Linear Price Models**

Linear regression for price analysis is the process of using one of the cost drivers from a Cost Catalog to analyze current pricing and evaluate or estimate new pricing. For example: linear analysis can be used to analyze castings or injection molded components on a price per pound basis or wiring, tubes or welding on a price per inch basis.

The following is a graphical example of a linear analysis which results in a model price of $.46 per pound.

All the individual part prices and weights are represented with blue dots. The orange trend line represents the pricing formula: weight* $.46. The closer the price points are to the trend line, the better the formula is at predicting price. This is a very good predictive formula.

An initial analysis rarely, if ever, looks this good. Example: Analysis of the data created a model where weight * $.59 = price. However, the resulting chart revealed a very high percentage of actual prices away from the trend line in the model.

Looking at the data table the buyer noted several instances of parts with similar weights but different prices and different volumes. Here is an example:

So, the data set was divided, and separate analyses were run for parts with volumes of less than 100,000 and parts greater than 100,000. The following predictive models were the result:

In this example, the buyer was able to create two linear models to cover the commodity. In practice, developing predictive linear models often requires looking at many factors on a trial-and-error basis. An example set of models for die casting is summarized in the following table:

Buyers can build linear cost models easily once they have a cost catalog in place. When they do, they often identify parts that are not priced like the others for negotiation and/or market test. Here is an example:

- A buyer responsible for outside painting of stampings was being inundated with supplier cost increase requests related to surging paint pricing.
- Creating a simple linear model of paint cost per square inch showed that suppliers’ pricing before the requested increases was inconsistent and got even worse with the requested economics.
- Market testing the parts with opportunity lead to a significant cost avoidance.

*Creating and maintaining Linear Pricing Cost Estimation Models serves the following purposes:*

- They can be used for quick costing. Often the “what will it cost?” question can be answered by selecting the right model and completing the calculation.
- Creating Linear Price Models will identify parts that are not priced like others and can be investigated for cost reduction.
- They can easily be constructed to include the impact of raw material cost variation.

**Creating Multivariable Regression Models**

Developing multiple linear formulas to cover a commodity can be a time-consuming task.

By using multivariable regression in Excel, we can look at many attributes, determine which are predictive of cost and develop a costing formula.

The costing formula will include an intercept value and coefficients. In a graph of the equation, the intercept value would be the location where the line intersects the vertical axis. The coefficients are values that are multiplied by the part attribute values in the equation.

**A costing formula would be structured as follows:**

Should-be Estimate =

intercept value +

(net part weight * net part weight coefficient) +

(number of machining operations * number of machining operations coefficient) +

(number of times washed * number of times washed coefficient) +

(raw material index * raw material index coefficient) +

(volume * volume coefficient)

Following is an example where part attribute values are being multiplied by the resultant regression coefficients to estimate pricing:

Buyers can build multivariable cost models easily once they have a cost catalog in place. When they do they often identify parts that are not priced like the others for negotiation and/or market test. Here are some examples:

- A buyer of tubes analyzed their 3 suppliers’ prices by performing separate MV analyses for each supplier. The analyses resulted in a predictive model for only one supplier and non-predictive models for the other two signifying that there might be savings opportunities in their part portfolios. The buyer then applied the one supplier’s predictive model to the other suppliers’ parts to see what savings could result if that supplier produced the parts. The result was over $300,000.
- A Mexican buyer of packaging could not get a predictive multivariable cost model for her corrugated box spend. The analysis and discussions with suppliers about the model results led her to the conclusion that a market test was warranted. The market test resulted in an 8% savings and a highly predive (.95 R Square) model she could use to provide her sales team with quick costing for new programs.

**Creating and maintaining regression estimation models serves the following purposes:**

- They can be used for quick costing.
- Creating regression models will identify parts that are not priced like others and can be investigated for cost reduction.

**Some points to consider:**

- It can be quicker to create a single, multi variable model than multiple linear models.
- Fluctuating commodity prices can be accounted for by including raw material pricing indices.

**Conclusions**

Achieving savings and building buyer skills and knowledge should not be looked at as separate events or workflows. By employing the correct techniques buyers can achieve both simultaneously.

**Learn about APD’s certified flagship training program “Cost Management Certification”**

**Watch an insightful webinar on Negotiating and Costing in the Market’s New Reality**