The really big operations of each generation each had a breakthrough in information processing. Sears figured out to do fulfillment from a big catalog for a large number of orders without O(N * M) operating cost. If you just have a big warehouse and people running around picking orders (like Safeway Select), the bigger the inventory, the costs climb with the product of the inventory size and number of transactions. Sears in Chicago covered several city blocks with multi-story buildings; they needed a better approach.

The breakthrough was the "schedule system". Orders came in and were recopied onto pick slips for different departments. Each order was given a temporary bin and a 45 minute pick window. Only orders for a given time window were picked in that window, and picking was by people close to that area of merchandise. Picked items were sent to the bin for that order, and at the end of each window, the bins went off to the checking and shipping station. If something was missing from the order, a second try was made in a day or two, and shipped separately. This is O(N * log(M)). Huge win.

This was all done with carbon paper, pneumatic tubes, and other advanced 19th century technology. They had the key idea - move data, not merchandise.

WalMart had a similar edge. They were one of the first retailers to bar-code everything. All the scanners were networked. HQ in Bentonville, Arkansas knew within hours what was selling and what needed to be restocked. They were able to keep unsold inventory down without being sold out. Unsold inventory is the curse of retail. Over half of apparel is eventually discounted or dumped. Sam Walton once commented that their edge wasn't size, but data.

Amazon everyone here knows about, so there's no need to explain them.