Why your supplier orders swing harder than your sales

The effect has a name

In 1997, Hau Lee, V. Padmanabhan and Seungjin Whang published "Information Distortion in a Supply Chain: The Bullwhip Effect" in *Management Science*. Their central observation: the variance of orders is larger than the variance of sales, and the distortion increases as you move upstream.

A retailer's demand wobbles a little. Their orders to the wholesaler wobble more. The wholesaler's orders to the manufacturer wobble more still. Nobody is behaving irrationally; the amplification is a property of how information travels as orders rather than as demand.

The four causes

The paper's lasting contribution is the diagnosis. There are four distinct mechanisms, and they call for different responses:

1. Demand signal processing. Each party treats an order from downstream as information about future demand and adjusts its own forecast, so a small blip gets read as a trend and amplified at every step.
2. The rationing game. When supply looks short, buyers over-order to secure allocation. The inflated orders are then cancelled or returned, leaving the supplier with a demand signal that was never real.
3. Order batching. Nobody orders continuously. They order in batches — weekly, monthly, by the container — so a steady trickle of sales reaches the supplier as a lumpy sequence of large orders and silences.
4. Price variation. Promotions and deal pricing cause forward buying. The customer's purchase pattern stops reflecting their consumption pattern, and the supplier cannot tell the two apart.

A freely readable copy of the paper is available as a course PDF if the journal is behind a paywall for you. Cite the journal — *Management Science* 43(4), 546–558 — and link the PDF as a courtesy.

Order batching is the local one

Of the four, order batching is the mechanism most visible in Ethiopian trade, and it is largely structural rather than behavioural.

If a supplier sells by the container, or the economics of a letter of credit only work above a certain order size, or the difference between importing monthly and importing quarterly is a meaningful freight saving, then you will order in large lumps. Your sales are smooth and your purchasing is spiky, because that is the only shape the supply arrangement permits.

That has a consequence worth being explicit about: your own order history is a bad forecast input. If you forecast from what you ordered rather than from what you sold, you are forecasting the batching pattern, not the demand. The paper's own remedy is to share the real demand signal, and for a single business that means forecasting from sell-through, full stop.

The remedy is information, not better guessing

The authors' conclusion was that the fix is sharing demand and order information across the chain, rather than each party forecasting harder in isolation. Most Ethiopian retailers cannot restructure their suppliers' visibility. But they can do the version of this that sits inside their own four walls:

• Compute reorder points from sales, never from past purchase orders.
• Record actual supplier lead times as observed, per supplier, rather than as promised. The variance in lead time matters as much as its length.
• Keep promotional periods identifiable in the sales history, so a forward-buying spike does not get read as a new baseline.
• When a large batch order is unavoidable, treat it as a financing decision with a known cost rather than a saving — the holding cost arithmetic is the other half of the unit-price discount.