Supply chain network design and inventory management for cost-efficient, in-time, perfect deliveries

Summary

在麻省理工学院（MIT）的春季系列直播活动中，Paulo Sousa Jr. 和 Miguel Rodriguez Garcia 共同主持了最后一次活动，邀请了 Wayfair 运营分析高级经理 Rafael Grillo Illipronti 作为嘉宾。Rafael 分享了 Wayfair 在网络设计和库存管理方面的策略，特别是如何通过优化配送网络和利用多级分销策略来实现成本效益、准时和完美交付。他讨论了 Wayfair 如何通过模拟和优化工具来权衡成本和速度，以及如何利用 AB 测试来衡量客户对交付速度的敏感性。此外，Rafael 还提到了 Wayfair 如何激励供应商在靠近客户的地点持有库存，以及如何通过激励而非强制的方式来影响供应商行为。最后，他提到了选择供应商时考虑的关键因素，包括产品类别管理和市场需求。整个讨论强调了学术理论与实际业务决策之间的联系，以及如何将复杂的供应链概念应用到现实世界中。

Takeaways

• 🎓 这是MIT供应链管理MicroMasters项目春季系列的第三次也是最后一次直播活动，前两次可以在MIT的Etiology YouTube频道观看。
• 🤝 Paulo Sousa Jr.和Miguel Rodriguez Garcia共同主持了这次活动，他们分别负责SC3x供应链动态和SC1x供应链基础课程。
• 📈 Rafael Grillo Illipronti作为演讲嘉宾，他是Wayfair运营分析部门的高级经理，负责全球履行中心网络设计。
• 🌐 Wayfair是一个连接供应商和客户的平台，专注于家具和装饰品市场，拥有国际化的供应商网络和遍及北美及欧洲的客户群。
• 🚀 Wayfair从纯粹的直运业务模式转变为投资建立物理分销网络，包括履行中心、交叉码头站、拉货点站和最后一公里交付站。
• 📊 Rafael强调了网络设计和库存管理在成本效率、及时交付和完美交付中的重要性，并分享了他在运营分析团队中的具体做法。
• 🔍 Wayfair使用模拟方法和经典优化方法的组合来分析成本与速度性能之间的各种场景和权衡。
• 📈 Rafael讨论了如何通过多级分销策略和安全库存管理来平衡库存水平和交付速度。
• 📝 他提到了如何使用AB测试来衡量客户对交付速度的敏感性，并将速度转化为销售价值。
• 🤔 在选择供应商时，Wayfair考虑的关键因素包括客户期望、供应商的地理位置、产品组合和市场特定需求。
• 📚 Rafael提到，MicroMasters项目的证书不仅有助于申请MIT的供应链管理混合硕士项目，也适用于全球其他大学。
• 📱 Rafael鼓励观众通过LinkedIn与他联系，以获取更多信息或提问。

Q & A

Wayfair的商业模式是什么？

Wayfair是一个平台，连接供应商和消费者。供应商主要负责生产Wayfair网站上销售的家具和装饰品，而消费者则是购买这些商品的人。Wayfair不拥有库存，除了特定情况外，他们的目标是保持资产轻量化。最初，Wayfair是一个纯直运业务，完全依赖第三方物流，如FedEx和UPS进行配送。后来，Wayfair开始投资建立自己的物理分销网络，包括配送中心、交叉装卸站、拉式站点和最后一公里配送站点。

Wayfair如何确定产品在配送中心的存放位置？

Wayfair使用运营分析团队开发的模型来确定产品在配送中心的存放位置。这些模型每六个月更新一次，以确保当前的决策与市场条件保持一致。他们考虑的因素包括运输成本、租金成本等长期趋势，而不是短期波动。

如何衡量配送速度对客户的重要性？

Wayfair通过与竞争对手的比较和AB测试来衡量配送速度的重要性。AB测试涉及选择一些产品，对不同客户群体显示不同的预计交付时间，并测量这些变化对销售转化率的影响，从而了解客户对配送速度的敏感度。

Wayfair如何管理其全球供应链网络中的库存？

Wayfair采用多级分销策略，将库存保留在区域中心（hubs），并根据靠近客户的配送中心（FCs）的消耗情况，从中心向这些配送中心运送产品。此外，Wayfair通过算法优化网站排名，激励供应商在更多地点持有库存，以提高产品的可用性和配送速度。

Wayfair如何选择供应商？

供应商的选择更多是由Wayfair的品类管理部门负责，他们根据目标市场客户的预期和需求来选择供应商。这涉及到为不同地区的客户提供符合他们预期的产品组合。

Wayfair如何平衡成本和配送速度？

Wayfair通过模拟不同的场景来平衡成本和配送速度。他们不追求单一的最优解，而是评估多个可行方案，并根据成本效益和速度进行权衡。他们还使用AB测试来了解不同客户群体对配送速度变化的敏感度，从而做出更有针对性的决策。

Wayfair的供应链网络设计有哪些关键考虑因素？

Wayfair在设计供应链网络时，考虑的关键因素包括成本效率、配送速度、市场需求的预测准确性、安全库存的需求、产品的边际利润以及客户对快速配送的期望。

Wayfair如何决定在哪些地点建立配送中心？

Wayfair通过分析长期趋势和市场条件，使用优化和模拟方法来决定在哪些地点建立配送中心。他们每六个月更新一次模型，以确认或调整之前的决策。

Wayfair的供应商主要位于哪些地区？

Wayfair的供应商主要位于东南亚、中国，以及北美、加拿大和欧洲的部分地区。这些供应商通常是市场特定的，即北美的供应商通常只为北美市场服务。

Wayfair如何利用高级IT分析来辅助供应商选择过程？

虽然选择供应商主要是品类管理部门的工作，但高级IT分析可以帮助他们更好地理解市场趋势、消费者行为和竞争对手的动态，从而做出更明智的供应商选择。

Wayfair的供应链管理教育课程有哪些？

Wayfair的供应链管理教育课程包括MITx MicroMasters program中的五个课程，涵盖了供应链基础、供应链动态、运输与物流中心的课程等。

Wayfair的全球供应链网络有多大的规模？

Wayfair的全球供应链网络服务于超过2000万客户，主要在北美、德国、英国和爱尔兰。他们的供应链网络包括超过1700万平方英尺的仓库空间，包括配送中心、交叉装卸站和配送站点。

Outlines

00:00

🎉 春季系列直播活动介绍

Paulo Sousa Jr. 和 Miguel Rodriguez Garcia 共同主持了春季系列的第三次也是最后一次直播活动。他们邀请观众查看前两期节目，并介绍了MIT的供应链管理教育YouTube频道。Paulo是MITx MicroMasters供应链管理项目中的SC3x供应链动态的领导，而Miguel是SC1x供应链基础课程的领导。他们讨论了供应链管理中的两个关键话题：网络设计和库存管理，并介绍了来自全球最大电子商务零售商之一的演讲嘉宾Rafael Grillo Illipronti。

05:04

🌟 Wayfair业务模式及供应链概述

Rafael Grillo Illipronti 介绍了自己，并概述了Wayfair的业务模式，它是一个连接供应商和客户的平台，专注于家居和装饰市场。Wayfair原先是一个纯粹的直运业务，但自2014年起开始投资建立物理分销网络，包括配送中心、交叉码头站、拉货点站和最后一英里配送站。Wayfair服务于北美、德国、英国和爱尔兰的2000多万客户，其供应商主要位于东南亚、中国以及一些在东欧、美国和加拿大的供应商。

10:05

📈 Wayfair的供应链结构和物流

Rafael 描述了Wayfair供应链的结构，包括国际和国内供应商的不同物流路径。对于国内供应商，产品可以直接运送到配送中心（FCs），有时也可以直接运送到交叉码头位置或基于资产的配送站。对于国际供应商，产品通常通过集装箱运输，可能需要30至40天的海运时间。Wayfair的物流决策包括是否将产品放置在靠近客户的位置或保留以利用安全库存池。

15:11

🔍 网络设计和库存管理的决策过程

Rafael 解释了Wayfair如何结合库存管理和网络设计来决定产品的定位。这涉及到决定是推迟还是提前产品的定位。Wayfair使用内部工具来评估不同的场景，并选择成本最优的解决方案。团队使用经典优化和模拟方法来分析成本和速度性能之间的各种场景和权衡。

20:14

🛠️ 应对日常挑战的策略

Rafael 讨论了Wayfair如何使用模拟方法来分析不同的场景和权衡，以应对日常挑战。他们生成多种“如果”场景，然后比较并展示给利益相关者，以决定最佳的网络设计。这种方法允许Wayfair在保持战略一致性的同时，根据市场条件的演变每六个月更新一次模型的结论。

25:23

📦 库存管理：产品定位的挑战

Rafael 强调了将产品放置在靠近客户的位置的重要性，以及这样做的挑战。他解释了如何通过多级分销策略来控制库存水平，并推迟产品定位决策。此外，他还讨论了如何根据产品的利润率、需求变化和客户对速度的期望来决定在特定地区的产品放置。

30:24

⚖️ 定位的易度与价值评估

Rafael 描述了如何评估产品定位的易度和价值。他提到了需求变化、利润率、客户期望、交货时间的变异性和交货时间本身等因素如何影响产品定位的难易程度。他还讨论了如何通过多级分销策略来平衡成本和速度，以及如何避免进入“危险区域”，即客户不关心且难以定位的产品区域。

35:27

🕒 模型更新周期和决策过程

Rafael 解释了Wayfair如何每六个月更新一次模型，以确保之前的决策仍然符合市场条件。他还讨论了如何通过模拟不同的场景来回答利益相关者的问题，并确认当前的路径。此外，他还提到了如何通过与竞争对手的比较和AB测试来衡量速度对客户的重要性。

40:32

🤝 供应商库存管理和选择策略

Rafael 讨论了Wayfair如何处理供应商的库存管理，包括如何激励供应商在更多地点持有库存，以及如何通过算法选择更靠近客户的库存。他还提到了供应商选择过程，这更多是由Wayfair的品类管理部门负责，他们根据客户期望和市场需求来选择供应商。

45:33

📝 总结与问答环节

在问答环节中，Paulo 和 Miguel 提出了几个问题，包括关于模型更新周期、速度对客户群体的重要性、供应商库存管理和选择策略等。Rafael 详细回答了这些问题，并提供了关于Wayfair如何操作和优化其供应链的见解。最后，他们感谢了参与者，并提醒了即将开放的SCx课程。

Keywords

💡供应链管理

供应链管理是指对产品从供应商到最终客户的整个流程进行规划、执行、控制和优化的过程。在视频中，这是整个讨论的核心主题，涉及到网络设计和库存管理两个关键话题。例如，MIT的供应链管理MicroMasters课程就是围绕这一领域设计的。

💡网络设计

网络设计是供应链管理中的一个环节，涉及到如何规划和布局产品的生产、仓储和分销网络以提高效率和降低成本。视频中提到，Wayfair通过建立物理分销网络，包括履行中心、交叉对接站和最后一英里配送站来优化其网络设计。

💡库存管理

库存管理是确保企业有适当数量的库存以满足客户需求同时又避免过剩的一系列活动。视频中讨论了Wayfair如何通过多级分销策略和安全库存来管理库存，以及如何决定在每个履行中心存放哪些类型的商品。

💡双 sided platform

双 sided platform（双面平台）是指一个平台服务于两个互相关联但又不同的客户群体，比如Wayfair连接供应商和消费者。在视频中，Wayfair作为一个专业的家居装饰市场平台，其供应商也是其客户，因为他们需要通过平台销售产品。

💡资产轻模式

资产轻模式是指企业尽量减少自身资产的持有，以降低固定成本和提高灵活性。Wayfair在视频中提到，除了特定情况外，他们不拥有库存，而是依赖第三方物流服务提供商来完成配送，体现了资产轻模式的特点。

💡CastleGate

CastleGate是Wayfair建立的物理分销网络，它包括履行中心、交叉对接站等。这个网络设计使Wayfair能够更好地控制库存和配送，提高服务质量。在视频中，CastleGate网络设计是实现成本效益、及时和完美交付的关键。

💡运营分析

运营分析是指使用数据分析工具和技术来优化企业运营的过程。在视频中，Rafael Grillo Illipronti作为运营分析高级经理，他的工作涉及到使用复杂的决策模型和模拟方法来优化Wayfair的网络设计和库存管理。

💡模拟方法

模拟方法是通过模拟现实世界中的不同场景来分析和预测系统行为的方法。在视频中，Wayfair使用模拟方法来评估不同的供应链场景，以确定成本和速度之间的最佳权衡。这种方法允许他们探索多种可能性并选择最优的供应链配置。

💡AB测试

AB测试是一种市场研究方法，用来比较两个或多个版本（如A和B）的优劣，以确定哪个更有效。在视频中，Wayfair通过AB测试来衡量顾客对不同交付速度的敏感度，从而评估快速交付对销售的影响。

💡供应商选择

供应商选择是确定哪些供应商最适合企业需求的过程。在视频中，Wayfair的供应商选择过程不仅考虑成本和质量，还考虑供应商的产品组合是否符合特定市场顾客的期望。这涉及到全球供应链网络中的品类管理，是确保供应链效率和响应市场变化的关键环节。

Highlights

MIT的春季系列直播活动已经举办了三次，本次为最后一场。错过前两场的观众可以在MIT的Etiology YouTube频道观看。

Paulo Sousa Jr.和Miguel Rodriguez Garcia共同主持了本次活动，他们分别负责MITx MicroMasters供应链管理课程的不同部分。

活动邀请了Wayfair运营分析高级经理Rafael Grillo Illipronti作为演讲嘉宾，他分享了关于供应链网络设计和库存管理的见解。

Rafael Grillo Illipronti拥有丰富的供应链管理经验，曾在巴西咨询公司ILUS工作，并持有圣保罗大学海军工程学士学位和麻省理工学院供应链管理硕士学位。

Wayfair作为一个连接供应商和客户的平台，专注于家居装饰市场，并且是一个典型的双边平台。

Wayfair原先是一个纯直运业务模式，但自2014年起开始投资建立物理分销网络，包括配送中心、交叉码头站、拉货点站和最后一公里配送站。

Wayfair服务于北美超过2000万客户，并在德国、英国和爱尔兰也有业务，供应商主要位于东南亚、中国以及部分在东欧和北美。

Wayfair的供应链管理是一个全球性的挑战，需要从亚洲运输商品到北美和欧洲，并确保客户能够快速收到产品。

Rafael讨论了如何使用运营分析团队的工具来确定成本效益、及时性和完美交付的网络设计和库存管理策略。

Wayfair使用模拟方法和经典优化方法的组合来分析成本与速度性能之间的各种场景和权衡。

Rafael强调了在实施网络设计决策时，需要考虑的不仅仅是理论最优解，还有实施的可行性和对供应商的解释。

Wayfair通过AB测试来衡量顾客对交付速度的敏感度，以此来评估速度对销售的实际影响。

Wayfair采用了多点分销策略，通过在区域中心持有库存，然后根据需要将产品从中心运送到靠近客户的配送中心。

Rafael解释了如何根据商品的利润率、需求变化和客户期望来决定商品的定位难度和价值。

Wayfair的供应商选择过程更多是由品类管理部门负责，他们根据客户需求和市场特点来选择供应商。

Rafael鼓励观众在LinkedIn上联系他以获取更多信息或提问，表现出了开放和互动的态度。

MITx MicroMasters供应链管理课程的五个课程中有些目前开放报名，鼓励观众查看并加入。

Transcripts

(bright music)

- Hi, everyone.

Welcome again to our spring live series.

Time flies.

This is already our third and last live event

for the spring series.

If you missed the first two,

you can watch them on MIT's Etiology YouTube channel.

By the way, there you will find many other great videos

on supply chain management education,

so feel free to explore the channel.

Thank you for joining us today.

I am Paulo Sousa Jr.

Of course lead for SC3x supply chain dynamics,

which is part of the MITx MicroMasters

in Supply Chain Management program from MIT.

Once more, I'm happy to be co-hosting this live event today

with my colleague Miguel Rodriguez Garcia,

course lead for SC1x Supply Chain Fundamentals.

Hello, Miguel.

- Hi, Paolo.

Thank you so much for the introduction and hi, everyone,

I'm really happy to be here with you.

I'm excited to bring the industry perspective

for our MicroMasters learners.

Once again, as you guys know, this is our core,

that connection between academia and the industry.

So we are really happy

to be hosting one of these events again.

So today we are gonna be discussing two key topics

in supply chain management.

One is gonna be network design

and the other one is gonna be inventory management.

And for that we have an amazing speaker joining us

from one of the largest e-commerce retailers in the world.

So stay connected

'cause it is gonna be a really, really great talk.

And as always, we are gonna follow the same agenda.

First, our guys speaker will give us a presentation

that will last around 25 minutes

and after that we'll have some time at the end

to answer some questions from you guys from the audience.

So that will be probably around 15 minutes.

And so the total length of the life event will be 45 minutes

and we encourage you to participate

by using the Q&A feature in Zoom.

Please try to avoid the chat for questions

'cause it is gonna be really hard for Paulo

and I to keep track of them.

So when you wanna ask any questions to our guest speaker,

use the Q&A feature in Zoom.

And at the end as we said,

Paulo and I will be channeling those questions

to our guest speaker.

And remember, Paulo also mentioned

that the event is part of the MITx MicroMasters program

in supply chain management,

a program that we develop here at the Center

for Transportation and Logistics at MIT,

and as well as supply chain fundamentals

and supply chain dynamics,

the MicroMasters program includes five courses in total

and some of them are currently open for enrollment,

so don't hesitate to check them out.

We'll be posting the link in the chat group

in case you guys are interested.

And now with this back to you, Paolo,

so you can introduce our guest speaker for today.

- Thank you so much, Miguel.

Today we are honored

to have Rafael Grillo Illipronti directly

from Berlin, Germany.

Rafael is in Operation Analytics senior manager at Wayfair.

He lets global fulfillment center network design

within the operations analytics team.

Before Wayfair, he worked

for a Brazil based consulting firm,

ILUS, enabling growth and competitive advantage

through supply chain management strategies

for e-com, retail, and CPG companies.

Rafael holds a bachelor's degree in NAVO engineering

from University of Sao Paulo

and a Master's in supply chain management from MIT.

He's also a MicroMasters alum,

which means he passed all courses

from the MicroMasters program

like many of you are doing right now.

As some of you may know, one among many other benefits

from earning the MicroMasters program credential

is that you become eligible to apply

to the MIT Supply Chain Management blended Master's program

at MIT, just like Rafael did, and also to other universities

around the world.

All right, so welcome back to the MicroMasters program,

Rafael the floor is yours.

- Cool, it's awesome to be here, Paulo.

Thanks everyone for being here as well

and to discuss two super interesting topics.

I remember maybe four

or five years ago attending these live events

and seeing like what folks were doing in the industry

and looking forward to be doing that as well.

And it's great to be here again now as a speaker, (laughs)

it's a bit different,

but, well, let me get started with the presentation

that we have for today.

Lemme know if you can see my screen, but hopefully you can.

- Yes.

- Awesome, thanks guys.

So the topic for today is network design

and inventory management for cost efficient, in time

and perfect deliveries.

So network design

and inventory management are rather broad topics,

but I wanna explore a little bit of how we are doing it

at Wayfair in more specifically how I am doing

within the operations analytics team,

of course supported by many other teams at Wayfair.

So quick intro about myself,

Paulo already mentioned a bunch.

I'm an MIT SCM alum from the class of 2021.

I have a bachelor's in naval engineering and spent six years

in consulting prior to Wayfair

with heavy focus on network design

and inventory management,

which is probably the reason I ended up doing

what I'm doing today. (laughs)

I'm currently based in Berlin, Germany,

so I've added two pictures that I took here.

The one in the bottom right is the Wayfair office.

So we are right in into the central part of Berlin

in Alexanderplatz overlooking the TV Tower,

which is quite nice view during the summer.

This was about last week in the end of the day,

but why don't we get started with what you guys are here for

and not my silly pictures.

So let me briefly try to capture Wayfair's business model

in a slide.

Basically Wayfair is a platform

that connects suppliers and customers.

So suppliers are essentially the ones that manufacture

the furniture and decor that we sell in our website.

And customers, of course are the people

that won't wanna go there and buy it.

Now so we are essentially a classic dual-sided platform

like many other platforms like Huber

or other marketplaces like Amazon.

The difference is we are specialized in a specific market,

which is home and decor.

Now, the fact that we are a dual-sided platform

makes suppliers also customers in a way

because our platform needs to be something that drives value

for them and therefore they need to feel compelled

to ship products to us or sell stuff in our website.

So in that sense, they're also customers for us.

So here really supplier relationship management is key

to doing what we do.

Our key important topic,

we don't own inventory apart from very specific cases

where we are mandated to do so,

but we aim to be asset light in that sense.

Originally Wayfair was a pure drop ship business

and what does that mean?

Essentially drop ship business means

we don't have any delivery infrastructure,

we rely fully on third party.

So we relied a lot on FedEx, UPS to do the deliveries.

So deliveries were happening from suppliers directly

from their warehouses in North America

and Europe to customers.

Now around 2014, we shift things around a bit

and we started heavily investing in building

a physical distribution network.

So that entails not only adding fulfillment centers,

but also adding cross docking stations,

pull point stations and delivery stations

for last mile deliveries.

We don't aim to own trucks despite that,

there are sometimes that we paint trucks

with the Wayfair logo, but those trucks are not ours.

They're third party, they just have the Wayfair logo.

So my people

from the US might have seen them delivering stuff

near your neighborhood.

But, yeah, essentially that's what Wayfair does

in a nutshell.

Now we serve more than 20 million customers

in North America focusing on Canada

in the US, US is our largest market.

We also serve customers in Germany,

and in the UK, and Ireland.

Our suppliers are mainly in Southeast Asia, China,

a few of them in Eastern Europe, and US, and Canada.

So it's really a global supply chain

that we're managing here.

Essentially needing to transport goods from Asia

to North America and Europe,

and really managing inventory availability to a point

where we can ensure that the customers are gonna have

the products that they want near them

to enable faster deliveries.

Last year we recorded revenues just above 12 billion

and just to give

a sense of the size of the footprint network

that we have today,

we're talking about more than 17 million square feet

of warehousing space

is adding up not only fulfillment centers

but also the delivery stations and cross docking stations

that we have across North American and Europe.

Of course the majority of them is fulfillment centers

because they are the largest buildings.

But the other things are there as well accounted for.

Now, lemme try to give you a summarized view

of what the Wayfair supply chain looks like.

Essentially we have two types of suppliers.

We have international suppliers,

which are those located in Southeast Asia and China.

We have domestic suppliers,

which are suppliers essentially like located

in North America, Canada, and Europe.

They are market specific.

So it's not common that a North America supplier

is gonna serve European customers.

We actually don't want that happening

because like product portfolio is different.

So we try to keep them market specific.

So that's why we are calling them domestic.

And really the international suppliers,

the way to get the product

from the suppliers into our CastleGate network,

which is essentially the physical distribution network

that we have built since 2014.

We have a few ways of doing that.

For domestic suppliers,

we can ship product directly from the supplier to the FCs,

so the fulfillment centers.

Sometimes we decide to skip the fulfillment centers

if, for example, the supplier

has a warehouse nearby one of our fulfillment centers,

there isn't really much of a reason to add a touch there.

So we skip the fulfillment center

and go directly to a cross docking location

where we then can inject into a parcel carrier like FedEx

or UPS, or we can ship it

to one of our asset based delivery stations

and then make the delivery to the customer.

Another option is sending product to what we call

a domestic break bulk facility.

That's something that we started to do this year.

We didn't do much before.

So still the idea

is that this domestic break bulk facility will be able

to break the bulk coming from the suppliers.

So essentially it's similar to a cross liking location

but operating on the inbound side.

And then for the international suppliers,

we also have a few options.

Our international suppliers, when we ship product,

we are talking about shipping via container normally.

So that means product is gonna be on water

for about 30 to 40 days

depending on which coast we are sending product to.

If we're sending to the west coast in the US

that's about 30 days.

If we are sending to the east coast, that's about 40.

Depends a lot.

And essentially the way we do that is we can ship product

directly from the supplier to the FCs.

And we would do that

for a supplier that has a lot of volume,

so enough volume to fill up at least one or two containers.

If the supplier doesn't have enough volume to fill up one

or two containers,

what we're gonna do is we are gonna send the product

from that supplier

to what we called international consolidation centers.

In those consolidation centers,

that volume is gonna wait for volume from other suppliers

to build up container.

And then we're gonna put that container

from multiple suppliers into a ship

and that ship is gonna travel from Southeast Asia

or China to the US, Canada or Europe.

Depends where we are sending that product to.

Then when it gets to the FCs,

we make decisions whether we are gonna position

that product near the customers

or if we're gonna hold it back

to leverage on essentially safety stock pooling.

That is why this whole presentation

has a bit of inventory management built into discussion

because that decision to either postpone

or anticipate the positioning,

it's essentially an inventory management decision

but mixed with network design.

And then we can do the same path that we do

with domestic suppliers.

We can ship from the FCs to the cross stocking

when a sale's made, and then we can move product

from the cross stocking stations

to the parcel carriers or to the delivery stations.

And really what happens here is we have multiple nodes

and we can do any combination of nodes

that it's cost optimal in a way.

So it gets very complex very fast

and it requires a lot of scenario modeling to determine

what's the optimal solution

to get to depending on the supplier

and depending on the type of product

that we are talking about.

So here this slide is just what we use internally at Wayfair

to define what operations analytics

the team I'm in does.

So essentially what we are trying to do here is

we are operating at a longtime horizon for decision making

and highly complex decision making.

So, I mean, it's the best of both words in a way

because it's complex, it's interesting,

but it's also challenging to do so.

So it requires us to leverage

a lot of our analytical ability

and modeling capacity to get to a good answer

that's gonna be robust and that's gonna give us

the results that we're looking for

in terms of cost abnormality and speed.

So really the types of questions

that we are trying to answer here,

specifically talking about network design

and inventory management

is where to locate fulfillment centers,

where to locate cross-talking centers,

where to locate delivery stations.

And then once we have defined

where to locate those facilities,

what types of items or products do we wanna hold

in each facility

and really what infrastructure do we need to build

for each facility such that the facility can hold

the type of inventory that we expect it to.

I am more focused

on the fulfillment center network design piece.

So question number one over here

and the what types of items we wanna hold in each FC,

the cross docking piece and the delivery centers piece

is done by another sub-team within operations analytics,

but I'm gonna try to capture that

in this presentation as well.

So if anyone in away is seeing this,

I'm gonna try to do my best. (chuckles)

But our approach to network design here

is really to have an end-to-end view

where we can trade off cost and speed fully.

So really the tools that we have built

across the years are tools that allows us

to not only see what's the cost optimal answer,

but also see what if scenarios

'cause that's ultimately what we need

to convince our stakeholders

that answer A is better than answer B.

And really the ability to trade off speed

and cost is something

that I've seen prior to Wayfair being done

in many different ways.

Sometimes folks try to capture what's the value of speed,

which is somewhat difficult to do

'cause there is an inherently strategic value to speed

that is hard to put a number on, which is essentially like,

what if my competitors invest a lot on speed

and that becomes a reality in the future

and I don't invest in speed, like measuring

that type of thing is very, very difficult to do.

So the way we do at Wayfair is we try to show

what's the speed outcome and what's the cost outcome,

and then we make inferences based on those two results.

So it's almost like we have a multi objective function

of minimizing cost but also trying to be better at speed.

Let me go to the next slide over here, oops.

Okay, cool.

So the the ways we are tackling these challenges

that we face on a day-to-day

is to use a combination of classical optimization

and simulation methods to analyze

the various scenarios and trade-offs

between cost and speed performance.

So it's what I mentioned earlier where say you generate

and folks from the SC1x are gonna remember

that there is a topic specifically for optimization modeling

where essentially we're trying to get to the optimal answer

for an accurate design configuration,

really that approach it works well

if you are comfortable enough

to say that your optimal answer is much, much better

than suboptimal answers.

In some cases what happens and what we see in a day-to-day

at Wayfair is that an optimal answer is not that much better

than a suboptimal one,

but sometimes a suboptimal answer is much easier

to implement and to explain to a supplier, for example.

So it's an important trade off that we make

that we are able to show what are the differences

between an optimal answer and the various scenarios

that suppliers can propose

and that other stakeholders can propose.

So we've transitioned a little bit away

from a classical optimization approach specifically

for FC network design to more of a simulation based approach

where we basically brute force run

all the scenario combinations

and we choose what scenarios we wanna look at really

from a computational standpoint,

there isn't much of a difference.

The simulation model that we use

is capable of generating the answers

within like 30 minutes running

and a classical optimization model

would be maybe 10 minutes faster, but much harder,

sorry, to give you the suboptimal answers.

So what we are doing is generating multiple

what if scenarios and then comparing them,

and showcasing them to our stakeholders and saying, hey,

if we do path A, it's a dollar per unit cheaper,

but we're gonna have to transform our network entirely

to be able to do that.

So maybe we do something simpler in the beginning

and then we transition to that option.

So I think that's really where the academic methods,

really the rubber meets the road in that sense

where we have to change

a bit of how we would solve this type of problem

in a controlled sealed environment.

And where when we have multiple things changing

at the same time, a huge organization

and people asking for different scenarios

and suppliers wanting explanations

for why are we proposing this instead of that,

we need to be able to give 'em sufficiently clear answers

and giving them an answer like don't worry,

trust this black box that it's right,

it's not something that they're gonna be happy with.

So this is the why behind we are really shifting

the way we do this.

So right after we define the location

for the fulfillment centers, the real challenges,

what types of items do we wanna put

in each fulfillment center?

So that becomes

more of an inventory management type of thing.

And let me actually show you why this is an important thing

for Wayfair, right?

So imagine that,

and that's a trend across all of the e-commerce

is that customers want fast deliveries happening faster

and faster and multiple competitors are pushing

for that same thing to delivery faster and faster.

And to do that there isn't an another way

to do it other than having product close to customers.

So being able to position the types of products that sell

in a specific region near the customers

of that specific region is really

what we are trynna do here.

And it actually allows us to have some sort of control

over the demand as well because we can influence

what people are gonna see at the website

by showing them things that we have nearby

and that kind of helps shape

what the consumer is gonna wanna purchase.

But in a way what we are trying to do is have product

near the customers that's really just about it.

But how we do that is the challenge.

And you saw that also in SC1x, if I'm not mistaken,

I think it is.

But the more you try to be accurate at the regional level,

so assume you break down demand too much

and you try to forecast it,

what's gonna happen is you're gonna be wrong more often

and when you're wrong more often,

that means you need more safety stock to cover for the times

that you're wrong.

And in those cases, what happens is we tell the supplier,

hey, Mr. Supplier,

please send us additional inventory

because we are making more forecast mistakes,

so we need more safety stock to cover for those mistakes.

Of course, for Wayfair,

that's not necessarily a burden

because we don't own the inventory but the supplier does.

So we are cognizant of that and we don't want the supplier

to see their inventory levels increasing like crazy.

Otherwise for them,

it's gonna be too much of a cash investment

in terms of working capital tied to inventory

that they're not gonna wanna work with us.

So we need to figure out a way

of positioning near the customer

but also being able to control the inventory levels.

And this is where we find

that a multi echelon distribution strategy

is really very effective.

And that's something that we are exploring

within the operations analytics team.

We're trying to essentially postpone

the positioning decision and actually hold buckets

of inventory in regional hubs

and then as we deplete the child FCs,

the FCs near the customers,

we ship product from our hubs to these FCs.

So that additional touch costs money,

but 40 items that have really high margin, it's worth it.

40 items where we are able to leverage the speed benefit,

it's worth it.

What's gonna happen is 40 items

that they have low margin

or their demand variability is crazy high,

we're not gonna wanna have them in every single location.

That's still a hypothesis that we are exploring,

but still, if you think about it,

it's something that really does make sense.

So really the trade off here is if we spread too much,

we increase inventory levels too much.

If we spread too little,

we don't leverage the speed

that we wanna serve customers with.

So there really is a trade off here

that's being brought by the demand, the demand variability,

the margin, the customer expectation really,

the lead time variability and the lead time itself.

So for international suppliers

where the lead time is 30 days

and for international suppliers

where sometimes we are using an unreliable sheeping route

that varies a lot the lead time,

we see that these things combined

to form like what I'm calling here,

the ease of positioning or the difficulty of positioning

for that matter.

So, for example, the higher the demand variability

and the higher the lead time variability,

the more difficult it is to position an item

near a customer.

The lower that is,

the easier it is to position an item near a customer.

So that's what is being represented here by the x-axis.

Now the y-axis is capturing here the value of positioning,

which is essentially a combination

between the customer expectation and the margin of an item.

So items with very high margin

and very high customer expectations

in terms of fast deliveries,

that means they have very high value of positioning.

For items where we have very low margin

and low customer expectations regarding the fast deliveries,

that means we have low value of positioning.

Now what can happen is an item has very high margin

but customers simply don't care

whether they're delivered faster or not.

That would be an item

where we also see low value of positioning

because if customers are ultimately not gonna purchase them

because they're being delivered faster,

really there is not much reason for us to deliver faster.

So what we're trying to do here

is trying to segment customer expectations

and value versus the challenges, and difficulty,

and costs of doing that positioning.

And really what we see here is that for items,

it's very easy to position them and it's very high value

to position them, of course,

we will want to have those items positioned close

to customers, so that's easy.

That's like the best of both worlds.

The challenge is for the zones,

and let me circle them over here,

these ones where we wanna be careful with what we do.

In these zones,

we need to make very specific trade offs regarding the value

and the challenges of positioning an item.

And it's really where the multi echelon strategy

that I just mentioned can be a propeller for facilitating

the positioning of products.

So it's really what we are trynna leverage right now.

And then we have this danger zone here,

which we have no value of positioning

and it's very hard to position those items.

For those items,

it's also easy enough to know what the answer is.

The answer is simply don't position those items

near to customers because customers don't care

and it's very difficult to do so.

So this is really nice.

This is question mark, we are exploring

and this is the danger zone or the the zone

where we don't wanna go (chuckles) essentially.

So with that, I think I spoke a lot about this

and now I wanted to leave a few minutes for you guys

to ask questions regarding what we are doing,

and we can explore this together

instead of you just hearing me talk and talk.

- Well, thank you so much, Rafael, for the presentation.

I think it was great.

We already have a lot of questions so that's amazing.

I think we're gonna start like shooting (chuckles) you

with some of them.

So the first one,

I'm gonna try to also wrap up some of your ideas here

'cause I think it was amazing how you shared

that the trade-offs

between having an optimal solution versus how easy it is

to actually deploy that kind of solution is something

that you don't find when you're only academia.

You're only touching the theoretical part.

And that's because not every cost is included

in an optimization problem.

It's impossible to include

the cost of changing some things

within the network whatsoever.

So there is a question from Amir Usar,

which is really interesting and he asked you

what is the period decision,

like the period of time that you guys use to refresh

the optimization model or the simulation model?

Do you run it every month, every day, every quarter

and why, maybe you run it differently

depending on the kind of decisions that you make, so.

- Yeah, so the these models that we have,

the intent of them is really to drive strategic decisions.

So the timeline for these decisions are like two

to five years.

We are refreshing the conclusions

from that model every six months

to make sure that the decisions that we made six months ago,

they still hold based on evolving market conditions, sorry.

And really the inputs don't change that often.

So for example, transportation costs, shipping costs,

rent costs, those types of things,

the way we model them

is to avoid changing them every six months

because we don't wanna capture short-term volatility

when we are modeling, we wanna capture long-term trends.

So unless markets conditions change dramatically

and we think that those conditions are gonna withstand

the next two to five years,

we are gonna keep the same inputs.

But really the rate of refresh for this is every six months.

What happens, this is kind of a good thing,

is that we tend to confirm the decision

that was made six months ago, every six months.

But that brings us reassurance

that we are in the right path.

So I think it's one of those things where we do that

because we need to keep up answering questions

that stakeholders have.

So what happens really every six months

is that we receive a bunch of questions from stakeholders

and we try to generate scenarios

and show them why we think scenario A

is better than scenario B,

and explain why we are going with that option

and following that path.

So it's really more of a communication thing

than more of a our network is changing every six months,

you know, so it's more of a strategic alignment

than change in direction.

- Yeah, no makes total sense.

Thank you so much for the answer.

Paulo, do you wanna take the next question?

- Yeah, sure.

Thank you so much Rafael, for such a great presentation.

You explored many topics from all of our courses,

so SC1x for sure,

inventory management but also SC2x network design

and we have some learners from SC3x here,

supply chain dynamics and in this course

we discuss about supply chain strategy.

You mentioned that you moved

from like a cost efficient supply chain

to a more responsive supply chain in terms of speed,

and this is great to see here in the discussion

and we have a question related to that.

So Andrea, Andrea Lopez is asking how do you measure

how relevant speed is to a given population if you do,

is it a comparison with competitors

or solely social features field research?

- So there are two ways to doing that essentially.

One is the comparison with competitors,

which is difficult to do because you have to map

the exact same types of product that you are selling

with the items that the competitor is selling.

So that's a fair comparison but we do that

to make sure that we are on par with our competitors

if not better.

Another way of doing that is via AB testing.

So what we do is we select a few products

and we do experiments to see how sensitive

the customers are to variations in delivery.

So we are gonna select for example, 100 customers,

that's just a random number,

we're gonna select a number of customers

and those customers are gonna see

that that specific item is gonna be delivered to them

in five days.

Then another set of customers are gonna see

that that specific item is gonna be delivered

for them in four days.

Another set of customer is gonna see three days,

and so on and so forth.

You got the logic and really we measure how much we convert

in sales for each of those groups of customers.

So if the customers that saw five days,

you would imagine that they don't buy

as much as the customers that saw four days,

and the customers that saw four days don't buy

as much as the ones that saw three days.

And so we measure that and we have essentially like a curve

that shows you for every item, for every customer segment,

what is their sensitivity to speed,

so how much more in sales we expect to get

from having faster deliveries.

And that's how really we translate speed

to an actual measure in terms of like value.

- Great, thank you so much, Rafael.

Miguel, do you want to take the next one?

- Yeah, actually the next question I had,

I think you already answered it Rafael,

'cause it was about AB testing.

So it's really interesting what you shared

'cause that's somehow a way you also shaped

the customer demand

'cause if you know that having a two hour delivery,

or same day, or like depending on the time windows

or whatsoever doesn't have an impact,

at some point you don't have to offer that anymore,

it actually reduce costs.

So really, really interested.

Yeah, so we don't have much time

but I'm gonna steal one of your questions Paulo

and then if you still want, you can also ask one more.

So Gustav, gosh, this learner wants to know like

when you are talking about holding inventory

nearby the fulfillment centers as safety stock,

like who holds the inventory from that point

'cause you said, for example,

that you don't have the inventory

so did you actually like force your suppliers

to have the safety stock for you,

'cause that's gonna be a key, I dunno, negotiation

in terms of like asking them to have more stock

for you guys without ownership.

- Yeah, yeah, so what what we do is,

it depends on two things, right?

We have a drop ship business

and we have a CastleGate business, right?

And the two things operate together.

So the supplier owns the inventory, right?

But they own the inventory that sits

in their drop ship warehouse and they also own the inventory

that sits in our CastleGate warehouse.

What's gonna happen is that the ranking algorithm

in the website, it's gonna select the items

that are closer to the customer.

So if a supplier doesn't have the item availability

near the customer is not gonna see their product.

What's gonna happen is they're not gonna buy it.

So because of that,

that's an incentive for the supplier to have inventory

in more locations.

That location can be a CastleGate warehouse

or it can be a drop ship warehouse, for all we care,

it doesn't make much of a difference specifically for that.

It makes a difference for other things

because CastleGate has later cutoff times.

So we can get faster deliveries in that sense.

But in terms of like availability on the website,

people are gonna see

despite being on drop ship or CastleGate.

Now the negotiation with the supplier for increasing

the inventory levels is indeed complex

because if we ask them to hold more inventory

because we wanna be in more locations, they can say no,

we are not gonna obligate them to have inventory.

If they don't see the value

in having more inventory, they're not.

So it's really a matter of setting the right incentives

in the platform and setting the right costs for the supplier

to operate where we want the supplier to operate in.

So it's more of a incentive-based approach

than a you need to have inventory here,

you're gonna be punished if you don't typa thing,

you know, so yes,

the supplier owns the inventory and we try to shape

the supplier behavior to where it's beneficial

for both the supplier and ourselves.

- Really, really interesting.

- Nice. - Yeah, thank you so much.

Paulo, do you wanna take one super quick last question

or do you think we wrap up?

- Yeah, I think we can take one more.

So let me take the last one

because I think it's connected to the topic

we are discussing right now.

So the next question here is what are the key factors

to consider when selecting suppliers

for a global supply chain network?

How can advanced IT analytics aid

in the supplier selection process?

- So the selection of suppliers in that sense,

it's more of a category management thing within Wayfair.

So it is more driven by what things we think

the customers wanna purchase in our website

and what things like, for example,

a German customer values more than a US customer.

The types of items

that are in the two websites are different.

So we are gonna wanna select suppliers for the two regions

that offer different portfolios that are in line

with what each customer expects.

That's not something

that my team specifically is working on,

but it's something that I know other teams

at Wayfair specifically category management are working on,

on selecting the suppliers that have the portfolio

that we know the customers are gonna purchase

in each specific market.

- All right, well,

thank you so much for all your answers, Rafael.

We still have a lot of questions,

but we don't have time to to answer them all,

but we really thank all of our learners

for the engagement for sure.

So again, thank you so much everyone

who decided to join us today.

It's been a super insightful session.

So thank you, Rafael, once more.

Before we say goodbye,

just want to remind everyone a couple things, first,

as you guys know,

this was the last life event of the spring series that Paolo

and I have been co-hosting for the last three months

between SC1x and SC3x.

So it's been a real pleasure

to share the experience with you guys.

Second, do you know some SCx courses are still open

for enrollment, for those completing SC1x and SC3x,

it's important to note that SC2x

and SC4x are gonna be opening really soon

within one month.

And so we encourage you to check them out on our website.

Yeah, just again, thank you everyone, Paul and Rafael,

thank you so much and if you wanna share any final words

with our audience, the floor is yours, guys.

- Thank you so much, Rafael, thank you so much, everyone,

it was a great session and have a great day.

- Thanks so much, guys.

If you wanna reach out to me on LinkedIn

or anything like that to ask questions,

feel free to do that,

happy to take questions offline as well.

Thank you, good to be here.

- All right, thank you everyone.

Have a great week, bye-bye.

- [Rafael] Bye.

- [Paulo] Bye.

(bright music)