Procurement Beyond ERPs - Velocity

Day 2 : Invent-ories! and concept of "Velocity"

Is there a better way to manage inventories? The methods of calculation of inventory ratios have not underwent major changes since early 2000s. Most of the ERP and the good old excel templates use MRP (Material Resource Planning) framework for managing inventories.

Consider the bullwhip effect where 10% variation in demand forecasting is extrapolated into all of the commodities to be procured. This has direct impact on the working capital.

Even the most sophisticated consulting models reolves around statistical methods. I am not saying it doesn't work, it does but not the way it is supposed to work. The inefficiency comes in due to a key component of supply chain - Forecasting Variation.

Every MRP system works on 3 simple components. It needs following inputs to being with :

A.) What is the opening inventory? (First time only, after that, it is auto calculated)

B.) What is the expected consumption?

C.) What is the desired safety stock and inventory norm level?

Hence, total material to be ordered = C - (A - B)

There are obvious methodologies to calculate safety stock and economic order quantity.

These are the standard KPIs ratios of procurement & supply chain (like cycle times, stock levels, cost to sales ratios, inventory turnover ratios etc).

The problem with MRP model is the variation. It works but only when the forecasting accuracy is high. Otherwise, due to high variation in expected consumption, the calculations change and MRP sends too many revisions to suppliers (and this is how it is being done right now!). In the end, MRP is left only as a automation tool and the planning is left to Buyer's intelligence (or in other words, Risk!)

I have started experimenting with the concept of "Velocity". Although, it is still work in progress but with machine learning, the concept has started to take the shape.

By Velocity, I mean following

1.) Velocity of ordering : At what interval the orders have been placed historically

2.) Quantum per order : How much has been ordered per order historically

3.) Cycle times : Order processing time + Transit time

These factors will help in calculating Consumption Time

By combining these 3 ratios (and many other in subsets), I am attempting to measure how much to buy and when to buy. Let me explain, with help of an example

Conventional MRP Method

Opening Inventory = 1000 MT/month

Expected Consumption = 700 MT/month

Recommended Inventory Level = 800 MT/month

MRP will issue purchase order of = 800 - (1000 - 700) = 500 MT

If expected consumption (700 MT/month) changes beyond the threshold limit, MRP will reissue the Purchase orders. In manufacturing industry, variation in consumption forecast is >20% (in India). Hence, number of cases where the threshold is crossed are too frequent to get any operational benefit of implementing MRP.

Velocity Method (Concept)

1.) Velocity of ordering = 2 Purchase Orders every week (Assumption 5 weeks per month)

2.) Quantum per order = 50 MT per order

3.) Cycle times =

3.a) Order Processing Lead Time = 10 Days

3.b) Transit Time = 20 Days

Consumption Time :

Say , for Period 1

2 x 5 = 10 PO every month with 50 MT per order ( Hence, total Qty 500 MT) to be issued

When to be issued = 3.a) + 3.b) days (i.e 10+20 = 30 Days) before Consumption Day starts.

Similarly, we will calculate for Period 2

Difference between Period 2 and Period 1 will give us indication of Consumption Time (in other words, consumption velocity).

After deducing Consumption velocity and Ordering velocity, I built Commodity wise and period wise dashboard using machine learning algorithm. I have built profile of each commodity according to velocity rates (high velocity means - high rate of ordering, so I have started to combine & consolidate the orders to moderate the velocity rate!).

I will continue to build Velocity concept and share my findings with you in the future.

My motivation :

I love building stuff. I am combining my little procurement and supply chain knowledge with my machine learning skills to go beyond today's procurement practices. I do not believe that going with best out of the 3 procurement style, need any skill. It can be and should be automated. Also, I have also seen the majority of manufacturing companies using ERP systems as transaction registers.

I am trying to create Procurement managers more Data Savvy where they focus on real value and not just transactions.

Contact me on [email protected] for more details. I will be happy to assist.