The tip of the iceberg at Rent the Runway
Rent the Runway is an iceberg of a business: when you look at the website you only see the smallest bit above the surface. Below the surface is the hugely complex array of people and systems that we use to fulfill the promise we make to our customers to deliver a Cinderella experience.
To frame the problem, imagine that you are building Netflix, only instead of shipping out DVDs, you are shipping out thousand dollar dresses. And those dresses have to get to someone in time to be worn to a big event, say, her prom night. Then they come back to us, to be inspected, cleaned, repaired, and sent off to the next Cinderella who needs them. This is the operational challenge that Rent the Runway has had to attack over the past 4 years of its existence. We describe our problem as ""reverse logistics"", far from just pick pack ship, we have to think a lot about the return phase of the order and how we can quickly process things coming in to enable them to go back out again.
I will discuss about how we have gone from a system run manually via Google Docs to one that runs on a network of event-driven services that coordinate the operations of hundreds of staff shipping tens of thousands of orders a week. This is a story about marrying complex technology, logistics and data considerations to physical processes that require large numbers of people making time-sensitive decisions. There are a few interesting key points of change in the story, from manual tracking to semi-automated to redesigning the entire warehouse flow (and entire software stack) to support real time inventory adjustment and same-day processing.
21. Time sensitive full reverse logistics
Weekend-dominated business means peak return
and ship out day is Wednesday
Over 60% of our inventory is received and
reshipped in a single day
36. Last Ship Today
UPS/Fedex
Last Ship NOT
Today, past/on
early ship
Early ship not
started, will be
needed soon
The Complete JIT Reverse Logistics System
(“ResCal”, “Allogator”,“Baywatch”, “Shipmunk”, “Panda”)
- A set of home grown systems powers our complex eco system of reservations, allocation and unit
prioritization, assembly, shipping, and the overall fulfillment.
ALLOCATION MAP
Presort Receiving Pre-Inspection Cleaning
Spotting
Steaming
Repair
QC
RACKED UNITS
Assembly
Bay (JIT op)
Shipping
Group
Assembled,
Completed
UPS Drop off
Since its launch RTR has grown from around 400 units of inventory to over 6x9
10 thousand (and counting)
From a rack at a drycleaners to busting at the seams of 40,000 square feet of warehouse space to 160000sqft wh
We’ve learned a lot!
You can go a long way on scrappy solutions but…
In the long run, it takes a lot of smarts to run a good logistics operation
Rent dresses and accessories
4 or 8 days
Size + backup size
Ship it to you, you ship it back
Launched in late 2009 in a drycleaner where we rented racks
Early challenges:
Realized on the week of launch that we needed to manage these items of inventory…
Oops, we need some sort of warehouse management software!
Forgot to track the backup sizes we were shipping out…
Dresses were referred to by name, not SKU
Barcodes? What barcodes?
Knew that we had 3 sizes, 3 units each size
When exactly should we be shipping things?
When will they come back?
Answer: Quite Far!
The Bible: Our answer to cross-region communication between customer service and the warehouse!
Customer calls with a problem? Record it in the bible!
Order that can’t get shipped? Put it in the bible for CI to deal with
31 columns…
Works until it doesn’t
Communication but no error tracking
Hard to get data out of the bible
Does not scale well
All of these things worked ok at small scales, but as we grew from a floor of a small building to a 40K sqft warehouse, things had to change
Simple Warehouse:
Usually provided by a “3PL”, third party logistics provider
Creating most efficient path to pick, saving human capital, is most important part of procss
Yeah, there are a few business that have done this (Netflix, Chegg), but…
Dry Cleaning!
Accurately model transit time, turnaround time, decay curve, lateness
Now that we know where inventory is, we can pre-allocate based on what is actually in the warehouse!
Downside: This original allocation took hours to run, and could only be run once or twice a day
Alerts on units that are needed for orders shipping soon
Assemby Bay
Enable same-day turnaround of inventory
Old system handled max daily cross docs (in and out same day) of ~400
We now do thousands
No more IUI line
The more incomplete groups staged on a line, the harder it was to find your specific group when the missing unit becomes available. Having to find the staged group, scan the new unit in a workstation and placing the group back on the line was highly inefficient.
Control all shipping label printing out and inbound
Able to set different inbound methods
Smooth the peak
What we have built so far has let us scale to turn around the majority of our inventory in the same day
Typical turnaround window is 8 hours, most in 24
The future is a combination of warehouse status, reservations, inventory, and predicted utilization to drive not only operations but our website itself
Multiple warehouse, global optimization, retail
Inventory: 2X a month to every single weekend with 0 day turnaround
And overwhelming
And it’s worth it to make all of these amazing happy moments happen