Data Flows

Labels: [Sourced] checked against a source listed at the end. [General] standard industry knowledge. [Our view] opinion, labelled as such. [Illustrative] made-up numbers to teach an idea.

1. The story: the satellite in Bentonville

Sam Walton described himself as distrustful of computers, and wrote he put up "a pretty good fight" whenever someone wanted a new system, "but I eventually signed the checks." [Sourced: Walton College page quoting his autobiography] In 1987, Wal-Mart completed what CIO magazine's timeline calls the largest private satellite communication system in the US at that time, linking all operating units and headquarters with two-way voice, data and one-way video. [Sourced: CIO timeline] The point is not the satellite. It is that a retailer decided that information moving between stores, warehouses and head office is as important as the goods. [Our view]

2. The one idea

Every box that moves has a stream of data moving with it: what it is, who ordered it, where it is, what it costs. Goods flow one way, data flows both ways, money flows the opposite way to goods. If the data is wrong, the goods end up wrong. [General]

3. The kitchen-table version

Mailing a gift. The box is the goods. The address label and tracking number are the data. The payment is the money. Without a correct label, a perfect gift goes nowhere.

4. The three flows, with our $12 bottle [General]

Flow	What moves	Example
Goods	The product	Factory to DC to store to you
Data	Orders, shipping notices, stock levels, sales	Purchase order, advance ship notice, daily sales
Money	Payment, invoices	You pay $12; the retailer pays the supplier later (Module 14)

5. The documents everyone sends, in order

Item data (master data): size, weight, case pack, price, identifiers (Module 1).
Purchase order (retailer to supplier). Module 8 covers EDI.
Advance ship notice (ASN): "this truck holds these cartons". GS1 lists ASN (EDI 856) among its data-exchange tools. [Sourced: GS1 US toolkit]
Receiving confirmation at the warehouse.
Invoice, then payment.
Sales and stock reports flowing back to planning (Modules 4-5). Event-level visibility, the "what, when, where, why and how" of products, is the purpose of GS1's EPCIS standard. [Sourced: GS1]

6. Why bad data is expensive

A Cranfield University study of UK retail product data found it inconsistent in over 80% of instances. [Sourced: Cranfield report; I have read the summary only] An academic paper cites a report that 80% of an industry's item master data was inconsistent, costing British grocery retailers and suppliers GBP 47 million a year in corrections and workarounds. [Sourced: ECIS 2011 paper citing an older report] Both are UK and dated; treat the 80% as a sign of scale, not a current number. [Our view]
Typical symptoms: a pallet that doesn't fit the shelf data, a case-of-12 entered as each, a price mismatch causing invoice disputes.

Worked example [Illustrative]. Item master says the case holds 12 bottles; the truck brings cases of 6. The retailer orders 600 bottles, expects 50 cases, receives 50 cases = 300 bottles. The system believes 600 arrived. Stock is overstated by 300 bottles x $5 cost = $1,500 until someone catches it, and the reorder rule (Module 9A) will not fire because it thinks stock is fine.

7. Where AI fits

Cleaning and matching product data across sources. [General]
Spotting anomalies (stock says 600, sales stopped). [General]
Forward thread: AI agents will read product data directly, so how a product is described in data decides whether an agent can find and buy it (Module 21). [Our view]

8. Exercise

List the data you would need to ship one box of bottles to a store. Which fields, if wrong, would stop the truck being received?

9. Quiz

Which way does money flow versus goods? Opposite.
What is an ASN? Advance notice of what a shipment contains.
What is master data? The descriptive record of each item.
Why can wrong case-pack data hide a stockout? The system overcounts stock.