---
title: "Bullwhip Effect: 4 Causes and Fixes for Small Firms"
description: "Orders swing harder than sales because of forecast updating, batching, price swings and shortage rationing. An invented small-roaster example and a fix table."
url: "https://staging.khoi.co.uk/en/blog/hieu-ung-roi-da-doanh-nghiep-nho"
lang: en-GB
author: "Lê Quốc Khôi"
date: 2026-11-03
updated: 2026-11-03
topics: ["Data & analytics"]
tags: ["Bullwhip effect", "Inventory", "Demand signal", "Supply chain"]
---

# The bullwhip effect: why orders swing harder than sales

## Key takeaways

- The bullwhip effect is the tendency of purchase orders to swing harder than sales, and harder still the further up the supply chain you go (Lee et al., 1997).
- Lee, Padmanabhan and Whang (1997) give four causes: updating forecasts from orders, batching orders, price fluctuations, and rationing in shortages (Lee et al., 1997). According to the same authors' Sloan Management Review article, at each link the buyer is still making a sensible decision.
- The remedies run through data and ordering rules: share retail sales, order in smaller and more frequent lots, keep prices steady, allocate by past sales. A better forecasting model is not enough.
- The examples in the sources I read come from large firms, and the inventory reductions of "up to 25%" are a figure the authors state without a source, not a controlled test. Translating them to a small workshop is my own inference.
- The roaster R and the cafés in this post are invented.

An invented example: roaster R supplies 12 cafés. Each week the cafés sell about the same amount, yet orders reach the roaster at 160 kg one week and 70 kg another. The roaster buys green beans on the strength of those orders, and the importer sees the roaster's orders swing even harder. If you sit in the middle of a chain and your stock is full one month and empty the next, this post gives four places to look, based on the [four causes of the bullwhip effect](https://staging.khoi.co.uk/en/framework/bullwhip-effect-causes).

## What is the bullwhip effect, and who named it?

It is the distortion of demand information as it travels up a supply chain: orders placed with a supplier vary more than the sales made to customers (Lee et al., 1997). In a companion article that the same authors published in the Sloan Management Review in spring 1997 (volume 38, issue 3, pp. 93-102), they describe logistics executives at Procter & Gamble looking at orders for Pampers. Retail sales fluctuated, but not excessively. Distributors' orders varied more, and P&G's orders for materials from suppliers such as 3M varied even more, although babies used nappies at a steady rate. P&G called the pattern the "bullwhip" effect. The article also tells how Hewlett-Packard found that orders from a major reseller of one printer swung harder than the reseller's sales, and that orders from the printer division to the integrated-circuit division swung harder again.

The consequences it lists are excess inventory, poor forecasts, capacity that is too large or too small, poor service from stock-outs, production plans revised again and again, and the cost of fire-fighting through expedited freight and overtime. For a small workshop the list is familiar; only the scale differs.

## What are the four causes, and why is the buyer not wrong?

What makes the original worth reading is that all four causes appear when every buyer decides sensibly with the information to hand, as the companion article stresses; nobody needs to be irrational.

> **Figure:** [Four causes of the bullwhip effect](https://staging.khoi.co.uk/en/framework/bullwhip-effect-causes) (redrawn from the site's library framework)

| Cause | Mechanism, per the authors' companion article | A form it could take in a small workshop (my inference) |
|:--|:--|:--|
| Updating forecasts from orders | Each firm forecasts from the order history of its direct customer; safety stock and long lead times enlarge the swings | Roaster R buys green beans on last week's café orders; the importer forecasts from the roaster's orders |
| Order batching | Ordering weekly, fortnightly or monthly because processing an order is costly and because buyers want full trucks; planning systems that run at the start of the month make many customers order together | Cafés order at the start of the month; the roaster buys beans only by the full container |
| Price fluctuations | Promotions make customers buy heavily when the price is low and then stop until their stock runs down | A 10% discount on orders above 50 kg makes cafés buy ahead and then order almost nothing for two weeks |
| Rationing in shortages | Suppliers allocate in proportion to orders; customers know it, so they order above their need and cancel when supply recovers | In the peak season beans are scarce; the roaster allocates pro rata, so cafés order double what they need |

One detail shows why batching is rational. The companion article quotes P&G's estimate that each invoice it processes costs between USD 35 and USD 75, because of many manual steps. With processing that costly, suppliers too prefer fewer orders and so encourage ordering in lots.

## How would a small workshop see the bullwhip in its own numbers?

Take four weeks at roaster R. The figures below are entirely invented.

| Week | Sold by the 12 cafés (kg) | Orders from the cafés to the roaster (kg) |
|:--|--:|--:|
| 1 | 90 | 160 |
| 2 | 110 | 80 |
| 3 | 95 | 90 |
| 4 | 105 | 70 |
| Average | 100 | 100 |
| Standard deviation | 7.9 | 35.4 |

The totals match, but the orders swing about 4.5 times as hard as sales in standard-deviation terms (population standard deviation). If the roaster bought beans on week 1, it would see demand 60% above the 100 kg average while what the cafés sell had barely moved. This is the thing to check with real data: take 8 to 12 weeks, not 4, and compare over the same period.

## How do you treat each cause?

The companion article groups remedies under three mechanisms: information sharing, channel alignment and operational efficiency. The table pairs each cause with what large firms did, then with what a small workshop could do. The last column is my inference and has not been tested on small firms.

| Cause | What large firms did (per the companion article) | Version for a small workshop |
|:--|:--|:--|
| Forecast updating | Use point-of-sale data; vendor-managed inventory. The article says inventory cuts of up to 25% are common in such alliances | Ask each café for a weekly sales sheet and forecast from that, not from orders |
| Batching | Smaller lots, more frequent delivery, mixed-product loads | Move large customers from monthly orders to two a week; count the total delivery cost |
| Price swings | Fewer and shallower discounts; P&G cut list prices by 12% to 24% and trimmed promotions | Swap one-off discounts for a steady price list; if you still promote, announce the calendar in advance |
| Shortage rationing | Allocate by past sales, not orders; share capacity and inventory information | Publish the allocation rule at the start of the peak season |

This suits you if your customers will share sales data and someone on your side will read it each week. It does not suit you if two big customers account for nearly all your orders: then ask them directly for their consumption plans instead of building a process. Allocating by past sales also has a price: new customers with no history lose out, so keep part of the stock aside for them. These steps come from large-firm cases, so try them on one or two customers first.

The order I would try, again my inference, is measure first, fix second. In week one, match the orders you received with what customers sold and compute the ratio of standard deviations. If it is close to 1, you have no bullwhip problem and your money belongs elsewhere. If it is high, ask which cause fits the shape of the data: orders bunched at the start of the month suggest batching, orders that jump after each discount suggest price swings, orders that surge and vanish in a scarce season suggest rationing. Change one cause at a time so you can tell what worked.

To link stock data across channels, see [Omnichannel starts with one inventory number](https://staging.khoi.co.uk/en/blog/omnichannel-ton-kho-thong-nhat). For when to use judgement and when to use a model to forecast, see [the map of demand forecasting methods](https://staging.khoi.co.uk/en/blog/ban-do-phuong-phap-du-bao).

## Going deeper: what did I read, and what did I not?

For the Management Science paper (Lee, Padmanabhan and Whang, 1997) I have only the abstract, not the full text, so I have not checked the mathematical model that shows variance growing at each link. The causes as described here, the P&G and HP examples, the USD 35 to 75 per invoice and the remedies come from the same three authors' companion article, "The bullwhip effect in supply chains", Sloan Management Review, spring 1997, volume 38, issue 3, pp. 93–102, available at https://sloanreview.mit.edu/article/the-bullwhip-effect-in-supply-chains/. That article is not in the site's library, and I read all of it. The examples are the authors' account of large firms, and "up to 25%" is a figure they state without a source.

I have not read the textbook that the framework summary on this site lists as a second source. The "small workshop" column in both tables and the figures in the four-week table are mine; there is no public source on small Vietnamese businesses.

## Frequently asked questions

### Do small businesses suffer from the bullwhip effect?

The sources I read do not measure small businesses; the examples in Lee, Padmanabhan and Whang's Sloan Management Review article (1997) come from large firms such as Procter & Gamble and Hewlett-Packard. The mechanisms do not need scale: you batch orders, you run promotions, you forecast from your customers' orders. A simple check is to set the orders you receive beside what your customers sell over 8 to 12 weeks; if orders swing much harder, you are somewhere along the whip.

### Does better forecasting cure the bullwhip effect?

Only in part. According to the authors' Sloan Management Review article, each link forecasts from its direct customer's orders, so the signal is already distorted at the start; my inference is that even a good model still receives that distorted signal. The site's framework summary names this as the trap: curing it with better forecasts alone, without sharing retail sales data or changing how prices and order lots are set [Four causes of the bullwhip effect](https://staging.khoi.co.uk/en/framework/bullwhip-effect-causes).

### How do I measure the bullwhip effect?

A simple way is to compare the standard deviation of the orders you receive with the standard deviation of what your customers sell, over the same weekly periods. A ratio above 1 means orders are more distorted than sales. This is the simple measure I use in this post; I have not checked the measure in the Management Science paper, so do not compare your figure with numbers in the research.

### Can my promotions cause a bullwhip effect?

They can. Price fluctuations are one of the four causes (Lee et al., 1997); according to the authors' Sloan Management Review article, when the price is low customers buy more than they need and then stop buying until the stock runs down, so purchases no longer match use. Before you run one, ask how much your customers will buy ahead and how far orders will fall two weeks later.

## References

- Lee, H.L., Padmanabhan, V. and Whang, S. (1997) ‘Information distortion in a supply chain: the bullwhip effect’, *Management Science*, 43(4), pp. 546–558. https://doi.org/10.1287/mnsc.43.4.546

---

Canonical HTML version: https://staging.khoi.co.uk/en/blog/hieu-ung-roi-da-doanh-nghiep-nho