Gap, Inc. consists of three retail apparel chains: Old Navy, which offers relativley low-priced apparel; Gap, which offers moderately-priced apparel; and Banana Republic, which specializes in a higher-priced line of apparel. There are more than 3,000 stores of the Gap family in North America.
Most of the labor at a Gap distribution center is in its piece-pick operations, which replenish retail stores. A piece-pick module is shown in Figure 1, where the three levels of one module can be seen. Pieces are picked from each such modules and then flow by conveyor through a sortation system and then to packing/shipping.
Figures 2 and 3 show the layout of one level of the module, with a conveyor running down the middle, from which shallow aisles emanate. Workers pick from paper pick-lists, following the conveyor down one side and back up the other. When an order is complete, the picker pushes the box onto the powered conveyor, which takes the box to sortation.
The Gap changed to bucket brigades that move along the conveyor, with order-pickers detouring into side aisles as necessary. Each order is picked into a cardboard box, which is pushed along the passive conveyor; when complete, the order is pushed onto the powered conveyor, which travels to the sorter.
Management of The Gap Southern Distribution Center in Galatin, Tennessee (USA), where these photos were taken, reported a 25% increase in throughput after changing to bucket brigades.
The orders at the gap were very variable in quantity and location. Some orders could be completed within the first few aisles, while others required a complete traversal of the the pick module. Accordingly, the pickers were instructed that anyone of them who completed an order was to push that order onto the powered take-away conveyor and walk back to get more work, as required by the bucket brigade protocol. (In other words, the walk-backs could be triggered by any worker who completed an order, not only by the last worker.)
To further improve productivity, the Gap tries to store the most popular items on the end-caps to reduce travel into the aisles.
Similarly, the variability in orders meant that some orders required picking within the first few aisles but others had no picks until the last few aisles. This could mean that a slower worker with an order that started near the end could catch up to a faster worker with an order that required work early. To reduce this inefficiency, engineers at the distribution center cached the paper pick-lists in the following way. Immediately before beginning a pick wave, the paper pick-lists were sorted into 4 piles: One was for all orders the first pick of which lay within the first quarter of the aisles; another was for all orders the first pick of which lay within the second quarter of the aisles; and so on until the fourth pile was for all orders the first pick of which lay within the last quarter of the aisles. Each worker was instructed that, when walking back to get more work, they were to take the first work they encountered, except that here the first work could be either the order carried by their predecessor; or else a paper pick-list cached in one of the four piles.
Figure 4: Pick lists are cached in one of two piles, whichever is nearer their first pick, as indicated by the blue order and the green order. The red order would be treated as usual.
This scheme helped reduce the variability in location of work that was inherent in the orders; and it also helped reduce the walking, because workers did not have to walk all the way back to the start to get an order and then have to walk forward to its first pick. Of course this sacrificed any FIFO sequence of orders, but this was unimportant as the orders all flowed through downstream sortation.