4. Collaboration: Data Sharing and Planning Alignment
Teaching rewrite — still the Data step of the framework, now looking beyond your own walls. Original wording; numbers are illustrative.
Introduction
Lesson 3 showed how to recover demand inside your own business. But most companies don’t sell to end consumers directly — they sell to retailers, distributors, or other manufacturers. Every link in between reshapes the demand signal before it reaches you, and the further upstream you sit, the more distorted it gets.
This lesson covers:
- How multi-echelon supply chains distort demand information
- The bullwhip effect and its five causes
- Internal collaboration: end-to-end visibility and central planning
- External collaboration: leaner logistics, four stages of information sharing, consignment, VMI, and CPFR
- Collaborating with suppliers
4.1 How supply chains distort demand
Take a two-echelon chain: manufacturer → retailer → end consumers.
flowchart RL
C[End consumers<br/>unconstrained demand] --> R[Retailer<br/>constrained sales]
R -->|orders shaped by inventory policy| M[Manufacturer<br/>forecasts from deliveries]
| What each party sees | Distorted by |
|---|---|
| End-consumer demand | — (this is the signal the whole chain should forecast) |
| Retailer sales | Shortages at the retailer |
| Retailer orders to manufacturer | Retailer’s inventory policy: pallet/truck multiples, order calendar, stock targets |
| Manufacturer deliveries (what most manufacturers forecast from) | Shortages at the manufacturer |
Example: the retailer changes its policy from holding 2 weeks to 3 weeks of stock. It places one big order — yet end-consumer demand hasn’t moved at all.
Each extra echelon is like another player in the children’s game telephone: the message drifts until it is unrecognisable.
If you can forecast end-customer demand, everything else is planning.
4.2 The bullwhip effect
The bullwhip effect: the further upstream a node is, the more variable the orders it receives — even when end-consumer demand is steady. A small flick at the handle (consumers) becomes a big swing at the tip (raw-material suppliers).
flowchart RL
A[Consumers<br/>steady] --> B[Store<br/>small swings] --> C[Distribution center<br/>bigger swings] --> D[Factory<br/>large swings] --> E[Supplier<br/>largest swings]
Jay Forrester (MIT) described the dynamics in the 1960s. The name “bullwhip” came in the 1990s from Procter & Gamble, which saw wildly variable factory orders for diapers even though babies consume them at a very steady rate. The beer game — a classroom simulation of a four-stage beer supply chain — lets players experience it first-hand.
Five causes
Lee, Padmanabhan, and Whang (1997) identified the first four; lead-time variation is a useful fifth. They reinforce one another.
| Cause | Mechanism | Remedy |
|---|---|---|
| 1. Order forecasting | Each node forecasts its incoming orders, not end demand — guessing the guess — and its inventory policy amplifies changes | Share end-customer demand (POS, sell-out) |
| 2. Order batching | Upstream nodes order less often, in bigger lots (truckload discounts, monthly cycles) → lumpy, delayed signal | Smaller, more frequent orders; pricing that doesn’t reward big lots |
| 3. Price fluctuations & promotions | Promotions pull demand forward; upstream sees swings without knowing why | Share promo calendars; stable pricing (e.g. everyday low price) |
| 4. Shortage gaming | Fearing shortages, nodes over-order to secure supply; others run short — often a self-fulfilling prophecy | Fair-share allocation; central control; transparency |
| 5. Lead-time variation | Longer quoted lead times → bigger safety stocks → bigger orders → more congestion → even longer lead times | Share capacity information; avoid panic re-planning |
Worked example: how an inventory policy amplifies a forecast change
A retailer sells a steady 100 units/week, keeps a target of 4 weeks of stock (400 units), and orders ~100/week.
It now believes demand has dropped 10% to 90/week:
| Item | Calculation | Units |
|---|---|---|
| New stock target | 90 × 4 | 360 |
| Current stock | 400 | |
| Excess stock | 400 − 360 | 40 |
| Next order | forecast − excess = 90 − 40 | 50 |
A 10% drop in perceived demand becomes a 50% drop in the order to the manufacturer. If the manufacturer applies the same logic to its own suppliers, the swing grows again.
Reverse it: if the retailer thinks demand rose to 110, target = 440, shortfall = 40, order = 110 + 40 = 150 — a 50% jump.
Case snapshot: grabbing the hot product
A distributor had one central warehouse and many independent stores. Whenever a store suspected a top seller might run short, it pulled as much as it could from the central warehouse. That store ended up overstocked while neighbours ran out — lower total sales and plenty of resentment between teams.
The lead-time spiral
flowchart LR
A[Longer lead times] --> B[More safety stock wanted]
B --> C[Bigger orders]
C --> D[More pressure on suppliers & ports]
D --> A
If a supplier’s quoted lead time goes from 21 to 28 days, your safety stock must cover a longer risk horizon, so you order more. When thousands of buyers react the same way to congested ports, the extra orders lengthen lead times further. The reverse happens when lead times shrink: everyone cuts orders and is left with overstock.
Checkpoint: distortion and bullwhip
Checkpoint — sections 4.1 and 4.2
1 / 4What does the bullwhip effect describe?
Pick an answer to see if you are right.
4.3 Collaborative planning
Instead of working in silos — forecasting incoming orders and ordering without regard for other echelons — share information and align plans.
Internal collaboration (inside your own supply chain)
- End-to-end visibility — one view of demand and supply across every inventory node: local shortages, sell-outs, point-of-sale data. Visibility improves forecasts; reports alone don’t create alignment.
- End-to-end planning and control
- Demand: coordinate pricing, promotions, and marketing across channels and markets in line with supply — usually in the S&OP cycle.
- Supply and inventory: a central team sets inventory policies for all nodes and makes sure they are followed. With multi-echelon inventory optimization (MEIO), typical inventory reductions are in the 10–30% range at the same service level. Even without advanced models, central control stops shortage gaming and enforces fair-share allocation.
External collaboration (with customers)
Start with leaner logistics. Set up your prices and costs to encourage small, frequent orders from customers (less batching, fewer delayed orders) — and place small, frequent orders with your own suppliers.
Four stages of information sharing:
| Stage | What you get | Limitation |
|---|---|---|
| 1. No information | Customer orders only | Full bullwhip — you forecast distorted orders |
| 2. Buying information | Sell-out and inventory data bought from third-party providers | Often partial, late, low quality, or at the wrong aggregation level; turning sell-out into sell-in forecasts is hard |
| 3. Sharing information | Customers send sell-out, inventory, possibly their policies — more, fresher, better data | Plans are still not aligned |
| 4. Collaboration | Data sharing plus aligned supply/inventory plans | Requires trust and integration — but unlocks MEIO and route optimization |
Using sell-out data at stages 2–3 (two approaches):
- Simulate the customer’s policy: estimate its usual stock cover from history, forecast its sell-out, and derive its likely orders (sell-in). Many assumptions — not guaranteed to be accurate.
- Machine learning: feed sell-out and customer inventory levels as extra features into the demand model.
Three ways to collaborate:
| Model | Who owns the stock at the customer | Who manages replenishment | Notes |
|---|---|---|---|
| Consignment | Supplier | Supplier | Supplier carries financial risk; useful to place risky/new products |
| Vendor-managed inventory (VMI) | Customer | Supplier | Customer carries financial risk; supplier sees POS and stock |
| CPFR (Collaborative Planning, Forecasting and Replenishment) | Per contract | Jointly | Joint S&OP covering quantities, stock targets, pricing, promotions, campaigns; demanding to implement — common among large retailers and in food |
Worked example: VMI in numbers
A supplier manages a customer DC under VMI. It sees: on-hand 1,200, sell-out forecast 400/week, agreed target cover 3 weeks, lead time 1 week, weekly review.
Target = 3 × 400 = 1,200, to be held after next week’s consumption. Inventory after one week without a delivery = 1,200 − 400 = 800. Replenishment = 1,200 − 800 = 400 — exactly the sell-out. Because the supplier sees real consumption, no amplification occurs.
Collaborating with suppliers
Share your expected supply requirements (not your raw demand forecast — net them of your stock and policies first). Suppliers can plan better and cut costs, which gives you room to negotiate prices; in return you get earlier warnings of shortages and more reliable lead times.
Checkpoint: collaborative planning
Checkpoint — section 4.3
1 / 4Under VMI, who owns the inventory at the customer site?
Pick an answer to see if you are right.
Rank: dampening the bullwhip
Sequence a bullwhip-reduction program
1 / 2Rank the steps a manufacturer could take.
OrderDrag the handle to rearrange — top should be first.
- 1Centralise inventory policy setting and fair-share allocation.
- 2Obtain sell-out and inventory data from key customers.
- 3Adjust pricing and logistics to encourage small, frequent customer orders.
- 4Build end-to-end visibility of demand and stock across its own nodes.
- 5Align plans with key customers through VMI or CPFR.
Drag items into the correct order (or use ↑↓), then check.
Check: collaboration and bullwhip
Collaboration, data sharing, and the bullwhip effect
1 / 6What should the whole supply chain ideally forecast?
Pick an answer to see if you are right.
Takeaways
- Each echelon distorts the demand signal; forecast end-customer demand wherever possible
- Bullwhip causes: order forecasting, batching, promotions/price swings, shortage gaming, lead-time variation
- Inventory policies amplify small forecast changes into large order changes
- Internally: visibility, then central planning (MEIO, fair-share allocation)
- Externally: lean logistics → buy data → share data → collaborate (consignment, VMI, CPFR)
- Share supply requirements with suppliers to get better prices and more reliable lead times
Further Reading
- 3. Capturing unconstrained demand
- Multi-echelon inventory optimization
- Next: 5. Forecasting hierarchies
- Background: N. Vandeput, Demand Forecasting Best Practices (Manning) — study outline; H. Lee, V. Padmanabhan, S. Whang, “Information Distortion in a Supply Chain: The Bullwhip Effect” (1997)