8 :author: Paul Tremblay
\r
9 :organization: Zappos
\r
13 This report sums changes to ZFC not found in the Amazon
\r
16 ========================
\r
17 Changes in Productivity
\r
18 ========================
\r
24 ZFC experienced a decline in productivity in 2011.
\r
27 .. figure:: charts/outbound_rate_2011.png
\r
30 :alt: Outbound Rate, 2011
\r
32 Rate for units processed for outbound. The sharp decline in the shaded
\r
33 area, at week 38, occurred during the cutover. _`outbound-graph`
\r
35 Looking at last year helps us better understand the decline.
\r
37 .. figure:: charts/outbound_rate_2010_2011.png
\r
40 :alt: Outbound Rate for 2010 and 2011
\r
42 Rates for outbound processes for 2010 and 2011.
\r
44 There is a 8.32 unit difference, +- 2.58 error (95% confidence
\r
45 interval). That translates to a -25% change, +- 7.9%. The graph shows
\r
46 that the decline started around week 8, labelled “migration” on the
\r
47 first chart. From weeks 6 to 11, Zappos moved all of its merchandise
\r
48 from warehouse 1 to warehouse 2 in order to make warehouse 1 ready for
\r
49 Amazon products. Not only did this migration require extra labor, but
\r
50 ZFC processes non-footwear at a slower rate than footwear. The section
\r
51 on static picking outlines the exact differences for static picking,
\r
52 and also shows that static picking accounts for most of the decline
\r
53 for the outbound process. If WH 2 had not received so much apparel,
\r
54 then most likely the rate would have remained at 35 units/ man hour.
\r
56 The next big decline occurs in mid June, labelled “Testing start.”
\r
57 Testing closed the warehouse at 1:00 AM in the morning, 5 hours before
\r
58 the brief time the warehouse is usually closed between shifts. The
\r
59 early closing time caused a backlog. Not only does a backlog by itself
\r
60 mean goods are processed slower, but it also required ZFC to hire
\r
61 temporary workers. Testing lasted until the week of cutover.
\r
64 Other warehouse disruption also hurt the outbound ate. For example, in
\r
65 early August ZFC tested SLAM shipping stations, and closed down PANDA
\r
66 lines in order to install the SLAM counterpart. With the PANDA line
\r
67 down, and the SLAM not yet functioning, workers had to manually
\r
68 prepare boxes for shipping. A week before cutover, ZFC tore out two
\r
69 more PANDA stations, again temporarily replacing them with manual
\r
70 stations. Graph shipping-graph_ shows the results on this disruption.
\r
71 Note the decline in week 33, and again in week 37, a week before
\r
72 cutover. Other processes show the sharp decline occurring about a week
\r
75 Last, the trends for both years run more or less parallel to each other. The
\r
76 sharp declines, starting at the end of September, results because of Zappos
\r
77 business model. Peak season for Zappos starts at the end of September, when
\r
78 ZFC receives a huge influx of goods, until the end of December, when Zappos
\r
79 ships a large amount of product. In order to handle this extra volume, Zappos
\r
80 hires more workers than it needs, making sure customer's orders never have to
\r
81 wait and Zappos reputation and future profitability suffer. The hiring trend
\r
82 was magnified in 2011, when ZFC hired even more worker than the previous year,
\r
83 because the previous year Zappos had not hired enough help and experienced
\r
88 Productivity by Process for Outbound
\r
89 ====================================
\r
91 As expected from the overall decline of outbound rates, the rates for
\r
92 individual processes also declined. The following graphs show the rates before
\r
93 and after the cutover.
\r
95 .. figure:: charts/outbound_rate_2011_all.png
\r
98 :alt: Outbound Rate for Several Processes, 2010
\r
104 Shipping can handle a much larger volume of goods than other stations, and
\r
105 there are fewer shipping stations. Hence the rate is much higer.
\r
108 .. figure:: charts/shipping_rate_2011.png
\r
111 :alt: Shipping Rate, 2010
\r
113 Units shipped per hour over weeks, with a trend line. The first
\r
114 decline occurred during week 31, at the start of August, when ZFC
\r
115 had to close down PANDA stations to install SLAM lines. The second
\r
116 decline, in the shaded area, occurred during the cutover. _`shipping-graph`
\r
118 Here are the other processes, compared with the previous year, included for
\r
121 .. figure:: charts/picking_rate_2010_2011.png
\r
124 :alt: Picking, 2010, 2011
\r
127 There is a 31.8 unit difference, +- 7.9 units, from 2010, or a
\r
128 -34% change, +- 8.5%. Static and carousel picking are
\r
129 combined for convenience. Although carousel picking has a much higher
\r
130 rate, its volume is low enough that it does not change the overall
\r
133 .. figure:: charts/multis_rate_2010_2011.png
\r
136 :alt: Multis Rate for 2010 and 2011
\r
139 There is a 10 unit difference in multis, +- 5.5 units. That translates
\r
142 .. figure:: charts/singles_rate_2010_2011.png
\r
145 :alt: Singles Rate for 2010 and 2011
\r
149 There is a 10.3 unit difference, +- 6.8 units. This translates to a
\r
150 -17% change, +- 9%.
\r
152 .. figure:: charts/shipping_rate_2010_2011.png
\r
155 :alt: Shipping Rate for 2010 and 2011
\r
158 There is no significant difference for shipping between the two years.
\r
163 Static picking accounts for most of the decline in rates for outbound,
\r
166 .. figure:: charts/static_picking_2010_2011.png
\r
167 :alt: Static Picking, 2010, 2011
\r
171 .. figure:: charts/static_vs_other.png
\r
172 :alt: Static Picking Compared to Other Processes
\r
178 Note how the rates for static picking mirror those for the outbound overall.
\r
180 From Justin Williams’ email
\r
183 I can definitely shed some light on reduction in efficiency last year in
\r
184 static picking. From a direct rate perspective there were three main reasons
\r
185 for declining performance. Firstly, we started the year with only footwear In
\r
186 W2. Starting in late February we transferred all of the non-footwear from W1
\r
187 into W2. This ended in early April. This transition of inventory resulted in
\r
188 a change to our picking procedure and our picking teams had to be trained on
\r
189 how to pick to tote. Also, Picking non-footwear to tote is less efficient
\r
190 than picking footwear to conveyor. Secondly, as we gained inventory from the
\r
191 move and in preparation for peak, we had to use pick mods that didn't have
\r
192 conveyance yet. This required pickers to travel several hundred feet and
\r
193 through gates to drop product onto takeaway conveyors. Finally, as we
\r
194 received non-conveyable product from W1, we had to start up a new department,
\r
195 Singulate, to take full totes of Multis and individually induct product into
\r
196 our Multi sorter to Multi lanes. I believe this department labor was shown
\r
197 under picking for several months until it was split out separately.
\r
199 ______________________________
\r
200 Footwear Vs Nonfootwear Rates
\r
201 ______________________________
\r
203 The decline in footwear resulted largely because it takes longer to
\r
204 process non-footwear than it does footwear. The following graph
\r
207 .. figure:: charts/foot_vs_non_foot_rate_2011.png
\r
208 :alt: Rates for Picking, 2011, Footwear and Non Footwear
\r
212 We can estimate the affect of apparel on the overall static picking
\r
213 rate by determining how much longer it takes to pick to tote. We can
\r
214 achieve this estimate in two ways:
\r
216 1. Estimate the difference from the figures that generated the above
\r
217 chart. Doing so we find that it takes 1.6 as long.
\r
219 2. Use figures from tests done at ZFC. These tests show that it takes
\r
220 1.7 as long to pick non footwear to a tote.
\r
222 These two figures can generate two more adjusted lines. Normally, the
\r
223 rate is determined by the units/hours. For static picking, the rate
\r
228 (text(units-footwear) + text(units-nonfootwear))/(text(hours-footwear) +
\r
229 text(hours-nonfootwear))
\r
231 We adjust the rate by taking out the extra time for the non-footwear
\r
232 picking. If it takes 1.6 as long to pick to a tote, then we divide the
\r
233 hours by the same amount. Doing so tells us the rate if all units
\r
234 picked were shoes, instead of apparel and non conveyor items. The
\r
235 equation for the lines then become:
\r
239 (text(units-footwear) + text(units-nonfootwear))/(text(hours-footwear) +
\r
240 text(hours-nonfootwear)/1.6)
\r
245 (text(units-footwear) + text(units-nonfootwear))/(text(hours-footwear) +
\r
246 text(hours-nonfootwear)/1.7)
\r
248 We can generate a third line to test the accuracy of the other two.
\r
249 Starting at week 8, we see a sharp decline in the static picking rate.
\r
250 Nothing else explains this drop except the massive amount of apparel
\r
251 moved to the warehouse. The drop showed that it took 1.3 times as long
\r
252 to process all product. In order to adjust this line, we can multiply
\r
253 the rate from week 8 by 1.3.
\r
255 If our adjustments are correct, we should see all three lines match
\r
256 up. In fact, this is the case.
\r
259 .. figure:: charts/static_picking_2010_2011_adj2.png
\r
260 :alt: Static Picking with Adjusted Lines, 2011
\r
264 We cannot adjust the year past the cutover by determining the amount
\r
265 of footwear and multiplying that by 1.6 or 1.7 because there is no
\r
266 data for non footwear for this time period. The Amazon WMS does not
\r
267 allow us to track footwear vs. non footwear. However, we can apply
\r
268 the adjustment of the third method above, of multiplying the entire
\r
269 rate by 1.3. Doing so yields a sensible adjustment:
\r
271 .. figure:: charts/static_picking_2010_2011_adj4.png
\r
272 :alt: Static Picking with Adjusted Lines, 2011, 2012
\r
277 Note how at the start of the year, static picking was at a rate of 62
\r
278 units. If we adjust the rate, we end at around the same place. For the
\r
279 first 3 weeks of 2012, the rates are 55, 55.5, and 55 units/man hour.
\r
280 If we apply the same adjustments, the rates are approximately 71,
\r
281 higher than last year.
\r
284 The adjustment constant of 1.3 may be somewhat generous. If we take
\r
285 the average of the all the adjustment lines from above, we come up
\r
286 with a 1.2 adjustment line. In that case, the graph looks like this:
\r
288 .. figure:: charts/static_picking_2010_2011_adj5.png
\r
289 :alt: Static Picking with Adjusted Lines, 2011, 2012
\r
293 That puts ZFC at 65 units for the start of 2012 (after adjustment)
\r
294 exactly the rate of the preceding year.
\r
298 Productivity by Process for Inbound
\r
299 ====================================
\r
302 .. figure:: charts/inbound_rate_2011.png
\r
305 :alt: Inbound Rate for 2011
\r
307 Units processed per hour over weeks, with a trend line, averaged
\r
308 across inbound processes, for the period before and after cutover.
\r
311 .. figure:: charts/inbound_rate_2010_2011.png
\r
314 :alt: Inbound Rate for 2010 and 2011
\r
316 There is a 9.1 unit change, +- 3.74 from 2010. That translates to -20%, +-
\r
324 .. figure:: charts/receive_rate_2010_2011.png
\r
327 :alt: Receive Rates for 2010 and 2011.
\r
329 There is a 25.03 unit difference, +- 9.50 from 2010. That translates
\r
330 to -19% change, +- 7%. Unlike other processes, the rate diverges more
\r
331 sharply after the cutover. If we take weeks 1-38 as the first
\r
332 interval, and 42-52 as the second interval, we find a 49.6 unit
\r
333 difference, +- 9.91. That translates to -41% change, +- 8%. The
\r
334 indirect labor accounts for much of this change.
\r
336 .. figure:: charts/receive_rate_indirect_total_2010_2011.png
\r
339 :alt: Indirect to Total Hours, 2010, 2011.
\r
342 The Amazon WMS introduced inefficiency to the receive process in
\r
343 requiring an extra step of converting the PO on the side of the box to
\r
344 an Amazon PO. Whereas formerly a ZFC team member only had to scan the
\r
345 side of the box before putting it on a conveyor belt, now he must
\r
346 apply and scan a case sticker, and then associate the case sticker to
\r
347 an Amazon PO by means of PO wrapper tool provided by Amazon. In some
\r
348 cases this involves scanning a code on a separate sheets of paper.
\r
349 These extra steps have increased the amount of indirect labor.
\r
351 In order to figure out the extra hours, we can look at the indirect
\r
352 hours as a percentage of the direct hours.
\r
354 .. figure:: charts/receive_direct_indirect.png
\r
357 :alt: Indirect vs. Direct Hours, Receive, 2010, 2011.
\r
359 We can adjust the lines in two ways:
\r
361 1. Calculate what the indirect hours should have been based on earlier
\r
362 in the year. Earlier in the year the indirect labor amounted to
\r
363 .455 |multiply| the direct labor (with the error margin added in).
\r
364 The total hours for after the cutover then becomes direct hours +
\r
365 .455 |multiply| indirect hours.
\r
367 2. Calculate what the indirect hours should have been based on last
\r
368 year for the same time period. Last year the indirect labor
\r
369 amounted to .365 |multiply| of the direct labor (adjusted for
\r
370 error). The total hours for after the cutover then becomes direct hours +
\r
371 .365 |multiply| indirect hours.
\r
373 The table below summarizes the results of these adjustments.
\r
376 .. class:: receive-hours
\r
377 .. csv-table:: Receive hours, adjusted
\r
378 :file: tables/receive_rate_adj.csv
\r
381 The adjustments show that the indirect labor amounted to 12,381, or
\r
382 19,222 extra hours. The second number is probably more accurate, since
\r
383 it is based on the previous year. Given that the peak season requires
\r
384 a huge increase in direct labor, we should see the percentage of
\r
385 indirect labor decline, and hence have reason for thinking the
\r
386 indirect labor amounts to only .365 of th direct. Using the second
\r
387 figure, we can find the cost of the Amazon inefficiency:
\r
389 19,222 |multiply| $16.10 = $309,474.00
\r
397 .. figure:: charts/putaway_rate_2010_2011.png
\r
400 :alt: Static and Carousel putaway for 2010 and 2011.
\r
403 There is a -35.6 unit difference from 2011, +- 12.7 units. That
\r
404 amounts to -24% difference +-15%.
\r
412 .. figure:: charts/return_rate_2010_2011.png
\r
415 :alt: Return Rates for 2010 and 2011
\r
417 There is a 7 unit difference, +- 1.97. That translates to -23%, +- 7%.
\r
419 Like the receive process, the returns process shows a sharp divergence
\r
420 from the previous year. The divergence starts at week 25. The change
\r
421 occurred for several reasons. In order to meet increasing demand, ZFC
\r
422 increased the size of its warehouse. During the build, ZFC eliminated
\r
423 the inbound line, the trash line, and the outbound line from the
\r
424 return process. Instead of having return items directly conveyed on a
\r
425 belt to the unpack station, workers had to put them on a palettes, and
\r
426 then move the entire palette. Likewise, team members also had to
\r
427 return items to storage by moving them on palettes. Similarly, trash
\r
428 was moved by a manual process. These changes required more manual
\r
429 labor and hurt rates. Both these changes occurred at the beginning of
\r
430 July, and can by the steep decline in graph returns-graph_.
\r
432 In addition, the returns department moved to its own building, one
\r
433 half mile away. Instead of simply placing items on a conveyor belt to
\r
434 transport them to storage, workers now had to put them on a palette,
\r
435 and have the palette driven to warehouse 2. This extra step also
\r
436 requires more labor, hurting efficiency rates.
\r
439 .. figure:: charts/return_rate_2011_2.png
\r
442 :alt: Return Rates for 2010
\r
444 Units Processed per hour over weeks, for returns, in 2011.
\r
447 The percentage of indirect labor also increased at the same time.
\r
449 .. figure:: charts/return_percent_indirect.png
\r
452 :alt: Percentage of Indirect to Total Labor, Returns
\r
454 This change occurred largely in part because returns categorized many
\r
455 processes as indirect that formerly were categorized as direct.
\r
460 Indirect vs. Direct Hours
\r
461 ==========================
\r
463 Since early 2011, Zappos has tracked direct and indirect hours. ZFC counts a
\r
464 direct hour as labor that directly contributes to a process. For example, when
\r
465 an employee picks an item off a shelf and puts it on a conveyor belt, the time
\r
466 counts towards direct labor. In contrast, the labor required to make sure the
\r
467 scanning guns operate correctly count as indirect labor.
\r
469 Not surprisingly, the percentage of indirect to direct labor increased in
\r
470 2011. ZFC expected this change because of the extra labor needed to move goods
\r
471 to a new warehouse, as well as implement thorough testing of Amazon's
\r
472 software. The chart below shows how as the percentage of indirect to direct
\r
473 hours increased, the outbound rate decreased.
\r
475 .. figure:: charts/indirect_out_vs_outbound_percent.png
\r
478 :alt: Indirect Labor Compared to Outbound Rate
\r
480 Indirect labor and outbound rate over time. Each line represents a
\r
481 percentage based on the first week of 2011. While indirect to direct labor
\r
482 increased to over 160% compared to the fist week, outbound rate decreased
\r
483 to under 60% compared to the first week.
\r
485 The percentage of indirect to direct labor increased at about 3% a
\r
486 month, starting at approximately 25% and peaking at approximately
\r
487 55% right after the cutover, though without the spike, the percentage
\r
488 would have reached a maximum of approximately 44%. Since the cutover,
\r
489 the percentage has declined. This doesn't indicate so much that ZFC
\r
490 decreased its indirect labor as that it increased its direct labor for
\r
493 At the same time, the increase in indirect labor does not explain the decrease
\r
494 in throughput rate, as the following graph shows.
\r
496 .. figure:: charts/direct_vs_total_outbound2_arrows.png
\r
499 :alt: Direct vs. Total Rate
\r
501 Rate of all processes, with indirect hours, and with indirect hours
\r
504 The overall line shows the number of units processed per man hour. The
\r
505 direct line shows the units processed per direct man hour, or the
\r
506 total hours required minus the indirect hours required. Since rate is
\r
507 determined by units divided by hours, the direct rate will always be
\r
508 higher, since it reduces the size of the denominator. The arrows shows
\r
509 the affect of indirect labor. If the indirect hours are zero, the two
\r
510 lines meet. As the indirect labor increases, the arrows get longer,
\r
511 and the overall rate declines.
\r
513 For the sake of examining the affect of the indirect hours on the
\r
514 overall rate, we want to look at the trends (or slope) of the lines
\r
515 compared to each other. If indirect hours accounted for the decline in
\r
516 rate, we should see an improvement in the direct trend once these
\r
517 indirect hours are removed. Instead of sloping down, the trend for
\r
518 direct rates should slope up, or, at least not slope down as sharply.
\r
519 In fact, the graph shows that the rates with and without indirect
\r
520 hours run parallel to each other. Put another way, indirect hours were
\r
523 We can also compare the indirect to total labor for years 2010 and
\r
524 2011. Note that there is no difference.
\r
526 .. figure:: charts/indirect_out_vs_total_2010_2011.png
\r
527 :alt: Percentage of Indirect to Total Labor
\r
530 Percentage of indirect to direct labor. There is only .7% difference
\r
531 between the two years.
\r
533 Last, we can compare the direct rates of both years, and the total
\r
534 rates of both years. Again, there is no difference
\r
536 .. figure:: charts/direct_vs_total_2010_2011.png
\r
537 :alt: Throughput, 2010, 2011, with indirect removed.
\r
541 The picture for inbound looks much the same, with a slight difference.
\r
543 .. figure:: charts/direct_vs_inbound_total_2010_2011.png
\r
544 :alt: Direct Labor vs. Total
\r
548 Direct rate and total rate are relatively close to each other.
\r
549 However, the throughput (overall) rate diverges from the direct rate,
\r
550 indicating that indirect hours had an impact on the overall rate. In
\r
551 fact, the percentage of indirect to total hours increase drastically
\r
555 .. figure:: charts/indirect_inbound_2010_2011.png
\r
556 :alt: Percentage of Indirect to Total Labor
\r
561 This change occurred in large part because ZFC recategorized the
\r
562 indirect and direct labor. For example, returns counted as direct
\r
563 labor the banding process, but later in the year, categorized this
\r
564 process as indirect.
\r
571 Hours for problems increased for problem solving from 2010.
\r
573 .. figure:: charts/problem_solve_2010_2011.png
\r
574 :alt: Total Hours for Problem Solving, 2010, 2011
\r
579 With 95% significance, there is a 1.5% increase for weeks 38-52 from
\r
580 2010 to 2011. With 95% significance, there is a 1.4% increase from
\r
581 weeks 1-37 to weeks 38-52 for 2011.
\r
583 There is no statistical significant change for weeks 1-37 for years
\r
586 The total hours for 2011 are 2,335,137. Multiply this by .015 to get
\r
587 35027 hours. Multiply 35027 times $16.10 an hour to get a cost of
\r
590 Note: if you take away the peak for weeks 38-41, there is still a .96%
\r
591 increase from 2010 to 2011. That is still a $360,918 increase
\r
594 .. put this earlier
\r
604 The cost of the cutover can be estimate by examining the hours and
\r
605 units as the year progresses and setting the first week in January as
\r
606 our baseline. As ZFC increases the units it processes, it should
\r
607 increase the hours in direct proportion. So if week 10 shows a 25%
\r
608 increase, the hours should also increase to 25%. In fact, this is the
\r
611 .. figure:: charts/units_hours_2010.png
\r
612 :alt: Outbound Units and Hours, 2010
\r
616 Note how the two lines practically lie on top of each other until week
\r
617 38. At week 38, the line for hours starts to diverge, and continues
\r
618 to diverge drastically, reaching a peak at week 50. The divergence
\r
619 occurs because of the ZFC business model of increasing labor beyond
\r
620 base need in order to ensure excellent customer service.
\r
622 Year 2011 looks slightly different.
\r
624 .. figure:: charts/units_hours_2011.png
\r
625 :alt: Outbound Units and Hours, 2011
\r
629 Again, the lines *nearly* lie on top of each other until week 38.
\r
630 However, the slight space between the lines represents the affect of
\r
631 the cutover. Starting in June, the hours increased greater than the units.
\r
632 This divergence represents inefficiency, or cost. If the units
\r
633 increased 25%, but the hours increased 30%, then ZFC used an extra 5%
\r
634 in labor. We can calculate the actual number of hours by multiplying
\r
635 the extra percent by the hours in January.
\r
637 In addition, we have to adjust for weeks 38 through 41. These weeks
\r
638 clearly fall within the cutover, but also fall within the peak period.
\r
639 Simply subtracting the two lines for this period will exaggerate the
\r
640 cost, since, given the ZFC business model, hours are expected to be
\r
641 higher. Instead, the previous year serves a baseline. For weeks 38, 39, 40,
\r
642 and 41 in 2010, the hours were higher by 18%, 19%, 39%, and 47%. For
\r
643 the same weeks in 2011, the differences were 241%, 431%, 375%, and
\r
644 298%. Subtracting the weeks in 2010 from 2011 yields a corrected
\r
655 The following table summarizes the extra hours, for weeks 25 through
\r
656 42, the period affected by the cutover.
\r
658 .. class:: cutover-costs
\r
659 .. csv-table:: Extra Hours, Outbound
\r
660 :file: tables/extra_hours_outbound.csv
\r
663 There were 86,583 extra hours. 86,583 |multiply| $16.10 amounts to $1,393,983.
\r
665 The following graph magnifies the critical area to show the extra
\r
666 hours. The gray area represents the extra costs.
\r
668 .. figure:: charts/outbound_costs.png
\r
669 :alt: Costs for Outbound
\r
677 Costs can be calculated the same way for inbound, with a slight change.
\r
678 Since returns suffered inefficiencies not related to the Amazon WMS,
\r
679 these will be taken out of the calculations.
\r
681 .. figure:: charts/units_hours_inbound_2010.png
\r
682 :alt: Inbound Units and Hours, 2010
\r
686 2010 shows that hours did not always increase in direct proportion to
\r
687 the increase in units. However, the average for the weeks before
\r
688 cutover show that if anything, ZFC increased its efficiency (increased
\r
689 processing units more than hours), so can can safely assume the same
\r
690 for 2011, since any mistake will underestimate costs.
\r
692 .. figure:: charts/units_hours_inbound_2011.png
\r
693 :alt: Inbound Units and Hours, 2011
\r
697 The following table summarizes the extra hours, for weeks 25 through
\r
698 42, the period affected by the cutover.
\r
700 .. class:: cutover-costs
\r
701 .. csv-table:: Extra Hours, Outbound
\r
702 :file: tables/extra_hours_inbound.csv
\r
705 The costs are found by the area under the curve. The extra hours are
\r
706 65,900. 71,544 |multiply| $16.10 amounts to $1,060,989.00
\r
709 The following graph magnifies the critical area to show the extra
\r
710 hours. The gray area represents the extra costs.
\r
712 .. figure:: charts/inbound_cost.png
\r
713 :alt: Inbound Units and Hours, 2010
\r
718 The costs for both inbound and outbound is $2,454,972.00
\r
723 Predicted Costs for 2012
\r
724 ========================
\r
726 ZFC conducted a study to determine the effect of some of the
\r
727 processes. The following graph and table summarize these estimates.
\r
729 .. figure:: charts/total_costs.png
\r
732 :alt: Predicted Costs for Several Processes
\r
734 Predicted costs for several processes.
\r
736 .. class:: projected-summary
\r
737 .. csv-table:: Summary of Projected Costs, 2012
\r
738 :file: tables/projected_summary_costs.csv
\r
741 .. container:: caption
\r
743 Summary of predicted costs for 2012, for some processes.
\r
750 Because ZFC does not know all the details of the FCSW return process, it cannot completely
\r
751 predict the costs. However, it does know the indirect labor costs caused by two additional
\r
752 steps. Under FCSW, unloading cartons from the trailer requires that the tracking number
\r
753 be scanned. In addition, return items must be released to the conveyor, either directly on
\r
754 in a tote, requiring a bar code be scanned.
\r
756 The following two tables estimate the change in labor rate and the change in costs caused
\r
757 by the transition to FCSW.
\r
759 .. figure:\: num 12
\r
762 .. figure:: charts/projected_rate_returns.png
\r
765 :alt: Predicted Rates for Returns
\r
767 Predicted rates for returns.
\r
777 As with returns, FC does not know all the details for the receiving process and cannot
\r
778 completely predict changes. However, FC can predict impact for labor for the additional
\r
779 step of retrieving a valid identification number to associate incoming cartons with an
\r
780 Inbound Shipment Delivery (ISD). FCSW acquires this number in one of two ways:
\r
782 1. FCWS scans the SSCC18 bar code or manual identification of the Amazon PO on
\r
785 2. If a vendor does not send an AISN (or does not include an SSCC18 on the
\r
786 carton); and if the vendor still uses legacy ZZZ-PO codes, the legacy
\r
787 ZZZ-PO must be entered or scanned using an additional tool that translates
\r
788 the original ZZZ-PO into the appropriate Amazon PO. This number is printed
\r
789 to a label, which is then applied to the carton.
\r
791 The following two tables estimate the change in labor rate and the change in costs caused
\r
792 by the transition to FCSW.
\r
794 .. figure:: charts/projected_rate_receive.png
\r
797 :alt: Predicted Receiving Rates
\r
799 Predicted receiving rates per hour for Zappos vs. Zappos FCSW for 2012
\r
807 There will be some efficiency gains in the static picking process due to an
\r
808 increase in pick density, as the current FIFO (First In First Out)
\r
809 requirement does not exist in FCSW. Based upon analysis that was performed in
\r
810 the static area of quad 3 (in W2) that compared pick rates from the end of
\r
811 February (when areas were less dense) with rates at the end of April (when
\r
812 areas were more dense) along with reviewing the current pick rates in SDF6, FC
\r
813 estimates FCSW will improve picking rates by 4%.
\r
815 The following two tables estimate the change in labor rate and the change in costs caused
\r
816 by the transition to FCSW.
\r
818 .. figure:: charts/projected_rate_static_pick.png
\r
821 :alt: Projected Rate for Static Picking, Pre- and Post-Cutover.
\r
823 Predicted static picking rates per hour for Zappos vs. Zappos FCSW for 2012
\r
827 ______________________________
\r
828 Receiving Exception Handling
\r
829 ______________________________
\r
831 There will not be any efficiency gains from the exception process itself, but there will
\r
832 be an overall cost savings due to a reduction in volume of goods moved through the
\r
833 exception handling process. Currently, three types of exceptions go through exception
\r
836 FCSW physically handles only the last of these exceptions, catalog related information
\r
837 defects, while treating the other two as RPIs to be routed to putaway and stored in
\r
838 unsellable locations until auto-resolved by the appropriate buying teams.
\r
840 Below is an illustration of this change in volume as well as the cost associated with this
\r
841 change in the process.
\r
844 .. figure:: charts/projected_exceptions_vol.png
\r
847 :alt: predicted Exceptions Handling Volume
\r
849 Predicted exception handling volume for Zappos vs. Zappos FCSW for 2012
\r
852 _____________________________
\r
853 Remaining Analysis Summary
\r
854 _____________________________
\r
856 Currently, ZFC does not have enough details about FCSW to effectively
\r
857 determine process gains or losses for the following: static putaway; carousel
\r
858 putaway; tote picking;carousel picking; singulate (induction); multi binning;
\r
859 multi packing; and Panda / SLAM, truck load. The physical steps for these
\r
860 processes appear similar as they exist today, but as ZFC better understands
\r
861 these operations under FCSW, and confirms and finalizes procedures, it can
\r
862 conduct the appropriate analysis.
\r
865 .. class:: appendix
\r
867 ====================================
\r
868 Tables of Projected Rates and Costs
\r
869 ====================================
\r
872 .. class:: projected-costs
\r
873 .. csv-table:: Projected Rates and Costs for Returns
\r
874 :file: tables/projected_rates_returns.csv
\r
877 .. container:: caption
\r
879 Data for Projected Rates and Costs for Returns, 2012.
\r
881 .. class:: projected-costs
\r
882 .. csv-table:: Projected Rates and Costs for Receiving
\r
883 :file: tables/projected_rates_receive.csv
\r
886 .. container:: caption
\r
888 Data for Projected Rates and Costs for Receiving, 2012.
\r
890 .. class:: projected-costs
\r
891 .. csv-table:: Projected Rates and Costs for Static Picking
\r
892 :file: tables/projected_rates_static_picking.csv
\r
895 .. container:: caption
\r
897 Data for Projected Rates and Costs for Static Picking, 2012.
\r
899 .. class:: projected-exceptions
\r
900 .. csv-table:: Projected Volume for Exceptions Handling
\r
901 :file: tables/projected_exceptions_volume.csv
\r
904 .. container:: caption
\r
906 Data for Projected Volume for Exceptions, 2012.
\r
909 .. class:: appendix
\r
911 ====================================================
\r
912 Tables of Rates of Productivity Pre and Post Cutover
\r
913 ====================================================
\r
916 .. class:: long-metrics
\r
917 .. csv-table:: Carousel Picking 2010-2011
\r
918 :file: tables/carousel_picking.csv
\r
921 .. container:: caption
\r
923 Data for Carousel picking, 2010-2011.
\r
925 .. class:: long-metrics
\r
926 .. csv-table:: Static Picking 2010-2011
\r
927 :file: tables/static_picking.csv
\r
930 .. container:: caption
\r
932 Data for Static picking, 2010-2011.
\r
934 .. class:: long-metrics
\r
935 .. csv-table:: Singulate 2010-2011
\r
936 :file: tables/singulate.csv
\r
939 .. container:: caption
\r
941 Data for the singulation process, 2010-2011. The singulation process did
\r
942 not exist before week 24 of 2011.
\r
944 .. class:: long-metrics
\r
945 .. csv-table:: Multis 2010-2011
\r
946 :file: tables/multis.csv
\r
949 .. container:: caption
\r
951 Data for the multis process, 2010-2011.
\r
953 .. class:: long-metrics
\r
954 .. csv-table:: Singles 2010-2011
\r
955 :file: tables/singles.csv
\r
958 .. container:: caption
\r
960 Data for the singles process, 2010-2011.
\r
962 .. class:: long-metrics
\r
963 .. csv-table:: Shipping 2010-2011
\r
964 :file: tables/shipping.csv
\r
967 .. container:: caption
\r
969 Data for the shipping process, 2010-2011.
\r
971 .. class:: long-metrics
\r
972 .. csv-table:: Outbound 2010-2011
\r
973 :file: tables/outbound.csv
\r
976 .. container:: caption
\r
978 Data for the outbound process, 2010-2011.
\r
980 .. class:: long-metrics
\r
981 .. csv-table:: Receive 2010-2011
\r
982 :file: tables/receive.csv
\r
985 .. container:: caption
\r
987 Data for the receive process, 2010-2011.
\r
989 .. class:: long-metrics
\r
990 .. csv-table:: Carousel Putaway 2010-2011
\r
991 :file: tables/carousel_putaway.csv
\r
994 .. container:: caption
\r
996 Data for the carousel putaway process, 2010-2011.
\r
998 .. class:: long-metrics
\r
999 .. csv-table:: Static Putaway 2010-2011
\r
1000 :file: tables/static_putaway.csv
\r
1003 .. container:: caption
\r
1005 Data for the static putaway process, 2010-2011.
\r
1007 .. class:: long-metrics
\r
1008 .. csv-table:: Returns 2010-2011
\r
1009 :file: tables/returns.csv
\r
1012 .. container:: caption
\r
1014 Data for the returns process, 2010-2011.
\r
1016 .. class:: long-metrics
\r
1017 .. csv-table:: Inbound 2010-2011
\r
1018 :file: tables/returns.csv
\r
1021 .. container:: caption
\r
1023 Data for all inbound processes, 2010-2011.
\r
1026 .. |multiply| unicode:: U+02715
\r