Lee, Padmanabhan and Whang on the bullwhip effect
📚SOURCE
💎PERMANENT NOTES
🧵SYNTHESIS
This is the paper that gave the 🧩 Bullwhip effect its name and, more usefully, its four causes. The observation is old: orders to a supplier vary more than sales to the customer, and the gap widens at each step upstream. The contribution is to show that this happens when everyone is rational. Earlier accounts leaned on managers misjudging the pipeline. These three write a model for each cause in which the people are optimising correctly and the distortion appears anyway.
The four:
- Demand signal processing. A retailer sees a rise, updates the forecast, and orders for the rise plus the stock the new forecast implies. Longer lead times make it worse.
- The rationing game. When supply is short and allocated in proportion to orders, everyone orders more than they need, then cancels.
- Order batching. Orders arrive monthly or by the lorryload, so the supplier sees lumps.
- Price variation. Promotions make buyers stock up, then go quiet.
Each cause has its own remedy, and none of the remedies is forecast harder.
✨KEY IDEAS
Variance of orders exceeds variance of sales, and grows upstream, even when every party behaves optimally.
A forecast update is passed on as an order, so the supplier receives the forecast error twice: once as demand, once as safety stock.
Under allocation by order size, an order stops being information about need.
Batching is driven by the fixed cost of ordering and of a full lorry.
Share sell-through data, shorten lead times, allocate on past sales, smooth prices.
📝READING NOTES
Paraphrase, not quotation. Management Science, 1997. The mathematics is in the appendices and I followed about half of it.
Timeline, honestly
Yusuf Demir's e-mail in September gave me the four causes in his own words, and that is what 📝 Issue 48 - Notes from a freight conference was built on in a hotel room in Hamburg. I read the paper itself on 1 and 2 November, three weeks after publishing. It held up. I had the causes right and the argument thinner than it deserved.
What is new compared with Forrester
- 📖 Forrester on amplification in supply chains shows that it happens and blames structure and delay in general.
- This paper splits the structure into four mechanisms you can test for separately.
- It moves the explanation from people misperceive to people respond correctly to what they can see. That is a stronger claim and a kinder one.
Reading it as a freight person
| Cause | Where I see it |
|---|---|
| Signal processing | A forwarder books capacity on last month's volumes plus a margin |
| Rationing | Peak weeks, when space is allocated and everyone asks for double |
| Batching | Weekly sailings, full-container pricing |
| Price variation | Rate changes announced a month ahead |
Where I am not convinced
The models assume a single product and a tidy chain. And the remedy share the data is offered as if the only obstacle were technical. A retailer's sales figures are leverage. In every negotiation I have watched, the party with the downstream numbers kept them. The paper knows what would fix the problem and has little to say about why the fix is against somebody's interest.
Fed 💎 Demand swings grow as they travel away from the customer, which stays at developing until I can show it in our own booking data.
Questions it left me with
- Can I see cause three in a customer's bookings? A weekly lump with flat sell-through behind it would be the test. One for the autumn module's regression tools.
- If carrier scorecards rank by volume handled, do we invite a rationing game of our own?
- Which of the four does an exceptions product make worse? Alerts on a shortage could start the over-ordering early.
🗣️IN SHORT
- In : Information Distortion in a Supply Chain: The Bullwhip Effect, Management Science, 1997
- Lee, Padmanabhan and Whang argue that : order swings amplify upstream for four separate reasons, each a rational response to the situation
- by stating that : forecast updating, shortage gaming, batching and price changes each distort the order signal, and that each needs its own remedy in how information, allocation and prices are set