Note: As a standard practice, I will include a plagiarism detection link in all the articles on my blog site. You can use the online tool below to evaluate any article published on the web for plagiarized content. All you have to do is to paste the url of the article there and it will go through every sentence in the article and flag any plagiarism. I believe as a follower of “writer’s integrity” I should include this tool in all my articles:
Introduction: Bullwhip is back in limelight
We have been hearing the word “Bullwhip” a lot recently, haven’t we ?
When the pandemic made the demand for certain products skyrocket, we started hearing about Bullwhip effect a lot. In early March, I even posted a rant video on why I think these mega events are not exactly true drivers of Bullwhip effect. I still chose to stick to that belief. Mega trends like these cause demand spikes so unprecedented that when that effect travels upstream, planners will be easily able to identify them. They will not be easy to get embedded in data.
All that apart, in this post, we will cover two key aspects:
- How can we quantify the magnitude of Bullwhip effect in our Supply Chains ?
- How can we leverage Digital and process management to mitigate Bullwhip effect.
I. They Do Exist !
We will use an extremely simple Supply Chain structure to illustrate the measurement methodology for Bullwhip effect. The Supply Chains, as shown below is a retailer, fulfilled by a manufacturer, that has only one raw materials supplier.
Now let us lay down some demand planning and ordering rules for our simple Supply Chain.
- The demand is relatively stable
- Ad hoc forecast updating and ordering policy at each level in the chain
- At the end of a period t, place an order for 2Dt-It. Dt is the actual demand in period t and It is the ending inventory in period t, computed after filling as much of Dt from inventory as possible
- Current demand is used as the forecast of future demand
- Replenishment lead time is one period (an order placed at end of one period arrives at the begining of next period) or zero
- Safety stock is one period worth of inventory
Now let us assume that the consumer sales data at retail locations for 20 periods is as shown below:
Now let us try to simulate the orders this above demand will generate across our simple Supply Chain. Based on the rules defined above, the following is the order quantity for manufacturer and its supplier, for the 20 periods, is indicated above.
Now, let us look at the order quantities for the consumer, manufacturer and the supplier on a graph. The Graph is shown below.
As you can see in the graph above, this example demonstrates clear Bullwhip effect, with manufacturing orders demonstrating twice the volatility of customer sales and Supplier orders demonstrating much more variability than manufacturer’s. So we know that Bullwhip exists but duh…we already knew that. So let us move to the key questions:
II. How do we measure the magnitude of Bullwhip effect ?
Now there can be various ways and in my perception, you an create some unique metrics based on your unique Supply Chain to measure “True” bullwhip but to keep it simple and general, one straightforward way is :
Taking the ratio of the standard deviation of orders placed at each stage to the satndard deviation of consumer sales reported.
So, going back to out above example, we calculate the Bullwhip rations as shown below:
In the example and the subsequent measurement of Bullwhip, we can see that even without including various “real world” Supply Chain complexities in the demand stream, you can see substantial Bullwhip effect in action.
Even simplest of complexities makes it worse
Now this is an experiment that I leave up to you to calculate but if you introduce a complexity as simple as changing the replenishment lead time to two periods from one, the ratios that you see above change to the following:
- Manufacturer’s Bullwhip ratio: 3.85
- Supplier’s Bullwhip ratio: 8.44
That simple change nearly doubled the Bullwhip ratios.
III. Illustrating the impact of Mitigation strategies
Let us try few simple approaches to mitigate the Bullwhip effect in this example.
First, we will use a base stock ordering policy at both the manufacturer and the supplier. This base stock policy will be an “updated” base stock policy where the policy will be updated periodically, in this case, every five periods.
Manufacturer bases its base stock policy on a forecast of demand, with a forecasted mean and standard deviation of per period of:
- 65 and 10 in periods 1-5 and 11-15
- 55 and 10 in 6-10 and 16-20
We will assume that the supplier uses a forecast of manufacturer’s order to compute its base stock in first five periods, with mean as 60 and standard deviation of 20. In the next five periods , supplier uses actual orders from manufacturer to compute the mean and standard deviation of demand.
At both stages of the Supply Chain, the base stock policy is computed as Si = Mean + z x Standard Deviation x Sqrt (2)
The results are shown in the illustration below. As you can see, this approach reduces the Bullwhip significantly, however, there are fill rate issues in some periods, as would be in the real world.
If we update our example to a Point of Sales (POS) data and Collaborative Planning and Forecasting (CPFR) scenario , where we assume the same base stock policy at the manufacturere as mentioned above, based on the forecast of rising and falling consumer demand. At the supplier, however, we will assume that the initial base stock level is computed directly from the manufacturer’s forecast so that mean =65 and standard deviation =10. In the succeeding fiver periods blocks, the supplier updates its base stock policy levels using actual consumer sales data. The resulting ratios and fill rate issues are shown in the illustration below.
IV. How can Analytics and Digital help ?
Ponder on this for a second:
Based on the example, and the subsequent mitigation example- what can make the most significant impact on reducing Bullwhip effect ? (I say reducing since in the real world, some level of Bullwhip will always exist and is not worth chasing).
The answer is, exactly what every Supply Chain textbook out there screams – Sharing information within the chain.
As you can see in the above example, using the same information (consumer demand) for planning purposes by all entities in Supply Chain is the most significant aspect. Decades ago, CPFR was introduced for this single purpose but admit it or not, it did not work wonders. What was the most significant reason ?
Lack of trust- with every entity in the chain thinking that they need to look out for themselves, ignoring the interest of others.
The classic and universal problem of Supply Chain. One that is also the reason for struggling Sales and Operations planning processes in companies but we don’t want to go there in this article. So the question is, how can we leverage Digital to build that trust so that all the entities start using the same number ?
A key challenge in my mind with CPFR was that higher-level management processes like CPFR were not designed to deal with what goes on day by day, item by item, and store by store. It’s essential to get good forecast information from retail customers to develop demand and supply plans. When the trust in number do not exist, entities in Supply Chain start covering for themselves. It becomes “Us vs Them”.
Build trust by sharing Manufacturing data- A proposed architecture
Now I will explain the simple Digital architecture that can address the challenges but to make it simple, the solution addresses the following two challenges:
- Lack of trust in forecast data
- Lack of integration of inter entity systems
The solution focuses on one aspect:
Even though forecast data and POS data is shared, the trust issue is that the entity downstream may not be actually using the numbers they shared. So if we can focus on sharing the Manufacturing data, at regular intervals, to show progress against the forecast, across the entity, it will alleviate many concerns.
Now let us review the proposed architecture below:
The primary concern is that the forecast data being provided by entity downstream is not what they are actually using for operations planning (like manufacturing planning). So the central aspect of the above architecture is to exchange information with the partners that shows progress of manufacturing plans based on this forecast.
Best in class processes and technologies, in order to be productive and effective, need best in class teams as well. And a critical aspect of these teams in inter team collaboration. The same can be extended to Bullwhip effect. No matter how tightghly integrated information flows you create, unless those flows are designed so that the resulting information exchange helps build mutual trust, they are are not effective enough.
Views my own.