Forbes.com

An Interview With Ed Macri, CMO Of Wayfair

Wayfair's Chief Product and Marketing Officer, Ed Macri, discusses the new rules of product development and marketing in a challenging e-commerce landscape.

著者：Cesar Brea and Ed Macri

First published on 7月 26, 2018
min read


This article originally appeared on Forbes.com. As retailers struggle to get e-commerce right, one company is breaking records. Wayfair is now the largest online furniture retailer in the U.S., with $6 billion in revenues. Like all online retailers, much of Wayfair’s success rests on its ability to make new rules in the critical areas of product development and marketing. We asked Chief Product and Marketing Officer Ed Macri to discuss how his team tackled that challenge. Bain: One of the fundamental difficulties in the crowded e-commerce landscape is distinguishing your company from competitors who are just as easily available to shoppers. What choices did you make to position your company differently? Ed Macri: We’re doing something new—we’re a “long-tail retailer in highly visual/emotional categories”—home and fashion are the big two. In some ways, we’re less “Amazon” with a very functional discovery process, and more “Zappos” with a need to enable discovery in ways beyond search—but across a much wider spectrum. We have category management, but it’s not a traditional “buying” role of making bets on specific items with inventory commitments with constrained shelf space. We want to offer the entire selection, merchandise based on customer preference with rich imagery, and partner with manufacturers to forecast demand for them, inform product development to fill gaps in our selection, and watch customer behavior to help us optimize how we present things. Bain: A lot of that depends on advanced analytics. How do you find a different breed of marketer with the right skills for the task? Macri: Yes, this leads us to look for different people than a traditional retailer—even an e-retailer—would. Historically, our biggest source of recruits was college grads with strong quantitative coursework. We liked this profile because they didn’t have preconceived notions about how our business should work. Now we also recruit directly from top MBA programs and consulting firms, as well as from start-ups, where people have demonstrated not just ability but intellectual flexibility in framing and solving challenges. We deliberately look for smart, motivated generalists. On campus, we’re competing for supersmart, analytical, quantitatively oriented people. Very often these are not folks who are thinking about going into marketing per se. Rather, they’re looking for GM roles, or product management roles at tech companies. It’s forced us to rethink how we conceive of our marketing organization internally, and how we project it externally (for recruiting). Inside the company, for convenience we call ourselves “Marketing.” On campus, we’ve rebranded as “Customer Growth.” And, we don’t call our new hires “marketers,” we call them “analysts.” We’re over 200 people in marketing now, excluding creatives, plus an engineering and product management team of over 100 people. We build all of our own ad tech, and the analysts drive the adtech product roadmap. So we also need them to be measurably technical; for example, they need to be comfortable talking with data scientists about models and the data underneath them. We screen the would-be analysts we recruit for quantitative skills pretty rigorously. It’s a case-driven process, including one that tests math explicitly. For example, we might ask someone to do a breakeven analysis on the fly: “OK, you want to spend more on search, and expect to pay a higher cost per click. How much would conversion rates have to improve to justify this?” We’re not testing whether you remember the textbook, but we are looking for comfort and facility with numbers. This may seem a bit extreme, but the reality is that it’s easy to underestimate how quickly technology is evolving at the edge of our category, where we compete. Even creative execution is going technical. For example, Google now designs display ads with AI, governed within design standards, grids, for example. This allows many more to be built, and tested. This is the kind of thing we need to be able to do to compete. Bain: Hiring people with quantitative or technical skills is one thing. But they need to be able work effectively in your culture—or be shaped to work effectively in your culture. Macri: Once we get them here, we emphasize learning by doing, and we have a high tolerance for risk. We can have this high-risk tolerance because it’s possible in our business to detect performance quickly, correct it and learn from it. It makes for a much faster learning curve. But at the same time our approach to development is pretty formal. We recruit “classes” on a summer schedule. We have a nine-month training program with coursework, but it’s really more of a co-op program where you work even as you complete the coursework. We see it as the graduate degree in marketing that you didn’t necessarily get, and can’t really get, at school, because so much of this is about doing it. Bain: As the talent pool shifts, you need different approaches to managing marketing, too. Macri: The objective for anyone joining is to operate like they are running a P&L. So, we don’t tell teams they have x to spend on search, display, social, whatever. We give them ROI targets, where ROI is calculated with income, and not revenue. They can ask for, and get, as much money as they want as long as they are hitting those ROI targets. It’s really motivating, and it also helps us avoid fights over a limited pool. Here’s one way that empowerment manifests itself. Marketers build their own forecasts, rather than getting them from others and then negotiating. We have what we call a 50-50 sales forecast model. We set our forecast to the point where the odds are even that we’ll make it or miss it. This expectation encourages risk-taking and helps us avoid sandbagging. We started this a couple of years ago and feel good about it. One reason is that we have really good feedback across the funnel about how our investments are influencing behavior. For example, we look at the ROI of “monthly cohorts” of newly acquired customers. In other words, we look at the 2-, 3-, 7-, and 14-day behavior of folks we spent money to attract in the first place. The deeper we see these folks in our funnel, the greater their conversion propensity—we’ve modeled it out. What this allows us to do is tweak how we execute across the funnel, given the relative strength of the performance we’re seeing. For example, we can adjust our bidding on search keywords to terms that are more specific, and thus closer to conversion, based on where our targets are. In short, our estimates of customers’ lifetime value are much more dynamic based on this early experience. Bain: How do you evaluate team members’ performance? Macri: Our hires will typically rotate to a new channel assignment every 18-24 months. During that time, we’re looking for results of course, adjusted for degrees of difficulty in market conditions. But we’re also looking for the quality of their thinking—how analytically bulletproof are their recommendations? Are we seeing real innovation? What are they contributing that’s new? Bain: You also need to determine how to support them with analytics. Macri: I started the analytics group at Wayfair back in 2007. But the model was “self-serve.” Our job was to make data sets available to analysts, half of whom at that time knew SQL and could query this data to get what they needed to support their decisions. We’re not quite that extreme now, as our tool sets have gotten more powerful and accessible over time, but we still provide classes on SQL and other tools, and expect our people can query our big data stores for what they need to manage their operations. This frees our analytics folks to spend more time on insight generation, rather than support routine analytic tasks for non-technical, non-quantitative managers. Once you get the data extract you need, you have to analyze it—for example, build a model. A simple rule of thumb is that if it’s doable in Excel, we expect our analysts to handle it. If it takes R or Python, either because of the size and granularity of the data set, or the flexibility, we need to build models or algorithms, it’s something for the data scientists. An example would be modeling expected performance of long-tail items, where the data’s thin and it’s a little harder to distinguish signal from noise. Bain: How extensively have you needed to adapt your technology infrastructure? Macri: Pretty much every technology we have for supporting marketing is homegrown, tailored to our particular needs. Why? Historically, as a smaller firm, we never had access to marketing technology budgets [to go out and license stuff], but being where we are [in Boston], and with very technical cofounders, we did have access to great engineers. So it was always easier to build versus buy. This opportunity makes Wayfair more actively requirements-driven, rather than passively demo-driven. In other words, Wayfair gets what it needs, and doesn’t pay for stuff that it doesn’t use. Further, the firm shares some of its code and supports open source development. Not every firm will practically be able to be this engineering-driven, but the example is instructive. One benefit we derive from being this strong from an engineering perspective is that our product teams are tight with their counterparts at places like Pinterest, Facebook and Google. So we spend less time trying to figure out what they’re up to, and more time adapting to their evolution and trading ideas for how to make the overall experience better. An example of how being strong in engineering helps us is the augmented reality (AR) feature, View in Room 3D, that we’ve incorporated into our mobile app. It’s a tool that allows you to visualize how things in our product range would fit into a space in your home. We used the iOS and Android SDKs (ARKit and ARCore) to develop it. Bain: As everything moves online and your internal data skills become so important, how have you changed how you approach “old media” and the role of agencies—if there is a role for them? Macri: In 2012, we started advertising on TV. After a while, we observed that doing this through intermediaries wasn’t consistent with how we wanted to manage marketing. There was more of an emphasis on just hitting targets versus exceeding them. So in 2015 we brought it in-house. All of it, from media buying—we now go direct to HGTV, for example—to production. Our buyers are more motivated, and we get a better yield. The production team can do a spot for a fraction of what an agency might charge. This allows us to do all of our testing in-market, nationally side by side. To evaluate the effectiveness of what we’re doing with TV, we started by simply correlating the change in mobile traffic to our home page with our ad flights. Bain: Can you give us an example of how your analytic orientation and control over infrastructure and process have been applied in a specific situation? Macri: We’re relaunching our rewards program. Like everyone else who has one, we’re concerned about just giving margin away without driving incremental lifetime value. Last time we evaluated it, we studied it deeply for six weeks, looking at all the numbers and talking with a few hundred customers. After that, we killed the program completely. Because our culture values experimentation and objectivity, there was none of the political gnashing of teeth you might expect with something like this. We just moved on. Now, we feel we’ve got better insights, and we’re ready to try again. Bain: What’s your best advice for other marketers? Macri: In other settings we see line marketers and even analysts who declare “I’m not technical, we’ll have to get someone from IT to handle that.” In our view, this line gets drawn way too conservatively. Wayfair’s culture of experimentation [like its “new mover” program for people moving homes], its willingness to break quickly when change is called for, its expectations for numeracy if not technical fluency, and for flexibility and curiosity, and finally its treatment of marke—um, analysts as businesspeople and not just project managers, all seem ideas with universal applicability, not just within a two-mile radius around MIT and Harvard. This interview was conducted by Cesar Brea, a Bain & Company partner in Boston and member of the firm’s Advanced Analytics and Marketing practices.

This article originally appeared on Forbes.com.

As retailers struggle to get e-commerce right, one company is breaking records. Wayfair is now the largest online furniture retailer in the U.S., with $6 billion in revenues. Like all online retailers, much of Wayfair’s success rests on its ability to make new rules in the critical areas of product development and marketing. We asked Chief Product and Marketing Officer Ed Macri to discuss how his team tackled that challenge.

Bain: One of the fundamental difficulties in the crowded e-commerce landscape is distinguishing your company from competitors who are just as easily available to shoppers. What choices did you make to position your company differently?

Ed Macri: We’re doing something new—we’re a “long-tail retailer in highly visual/emotional categories”—home and fashion are the big two. In some ways, we’re less “Amazon” with a very functional discovery process, and more “Zappos” with a need to enable discovery in ways beyond search—but across a much wider spectrum. We have category management, but it’s not a traditional “buying” role of making bets on specific items with inventory commitments with constrained shelf space. We want to offer the entire selection, merchandise based on customer preference with rich imagery, and partner with manufacturers to forecast demand for them, inform product development to fill gaps in our selection, and watch customer behavior to help us optimize how we present things.

Bain: A lot of that depends on advanced analytics. How do you find a different breed of marketer with the right skills for the task?

Macri: Yes, this leads us to look for different people than a traditional retailer—even an e-retailer—would. Historically, our biggest source of recruits was college grads with strong quantitative coursework. We liked this profile because they didn’t have preconceived notions about how our business should work. Now we also recruit directly from top MBA programs and consulting firms, as well as from start-ups, where people have demonstrated not just ability but intellectual flexibility in framing and solving challenges. We deliberately look for smart, motivated generalists.

On campus, we’re competing for supersmart, analytical, quantitatively oriented people. Very often these are not folks who are thinking about going into marketing per se. Rather, they’re looking for GM roles, or product management roles at tech companies. It’s forced us to rethink how we conceive of our marketing organization internally, and how we project it externally (for recruiting). Inside the company, for convenience we call ourselves “Marketing.” On campus, we’ve rebranded as “Customer Growth.” And, we don’t call our new hires “marketers,” we call them “analysts.”

We’re over 200 people in marketing now, excluding creatives, plus an engineering and product management team of over 100 people. We build all of our own ad tech, and the analysts drive the adtech product roadmap. So we also need them to be measurably technical; for example, they need to be comfortable talking with data scientists about models and the data underneath them.

We screen the would-be analysts we recruit for quantitative skills pretty rigorously. It’s a case-driven process, including one that tests math explicitly. For example, we might ask someone to do a breakeven analysis on the fly: “OK, you want to spend more on search, and expect to pay a higher cost per click. How much would conversion rates have to improve to justify this?” We’re not testing whether you remember the textbook, but we are looking for comfort and facility with numbers.

This may seem a bit extreme, but the reality is that it’s easy to underestimate how quickly technology is evolving at the edge of our category, where we compete. Even creative execution is going technical. For example, Google now designs display ads with AI, governed within design standards, grids, for example. This allows many more to be built, and tested. This is the kind of thing we need to be able to do to compete.

Bain: Hiring people with quantitative or technical skills is one thing. But they need to be able work effectively in your culture—or be shaped to work effectively in your culture.

Macri: Once we get them here, we emphasize learning by doing, and we have a high tolerance for risk. We can have this high-risk tolerance because it’s possible in our business to detect performance quickly, correct it and learn from it. It makes for a much faster learning curve.

But at the same time our approach to development is pretty formal. We recruit “classes” on a summer schedule. We have a nine-month training program with coursework, but it’s really more of a co-op program where you work even as you complete the coursework. We see it as the graduate degree in marketing that you didn’t necessarily get, and can’t really get, at school, because so much of this is about doing it.

Bain: As the talent pool shifts, you need different approaches to managing marketing, too.

Macri: The objective for anyone joining is to operate like they are running a P&L. So, we don’t tell teams they have x to spend on search, display, social, whatever. We give them ROI targets, where ROI is calculated with income, and not revenue. They can ask for, and get, as much money as they want as long as they are hitting those ROI targets. It’s really motivating, and it also helps us avoid fights over a limited pool.

Here’s one way that empowerment manifests itself. Marketers build their own forecasts, rather than getting them from others and then negotiating. We have what we call a 50-50 sales forecast model. We set our forecast to the point where the odds are even that we’ll make it or miss it. This expectation encourages risk-taking and helps us avoid sandbagging. We started this a couple of years ago and feel good about it. One reason is that we have really good feedback across the funnel about how our investments are influencing behavior. For example, we look at the ROI of “monthly cohorts” of newly acquired customers. In other words, we look at the 2-, 3-, 7-, and 14-day behavior of folks we spent money to attract in the first place. The deeper we see these folks in our funnel, the greater their conversion propensity—we’ve modeled it out. What this allows us to do is tweak how we execute across the funnel, given the relative strength of the performance we’re seeing. For example, we can adjust our bidding on search keywords to terms that are more specific, and thus closer to conversion, based on where our targets are. In short, our estimates of customers’ lifetime value are much more dynamic based on this early experience.

Bain: How do you evaluate team members’ performance?

Macri: Our hires will typically rotate to a new channel assignment every 18-24 months. During that time, we’re looking for results of course, adjusted for degrees of difficulty in market conditions. But we’re also looking for the quality of their thinking—how analytically bulletproof are their recommendations? Are we seeing real innovation? What are they contributing that’s new?

Bain: You also need to determine how to support them with analytics.

Macri: I started the analytics group at Wayfair back in 2007. But the model was “self-serve.” Our job was to make data sets available to analysts, half of whom at that time knew SQL and could query this data to get what they needed to support their decisions. We’re not quite that extreme now, as our tool sets have gotten more powerful and accessible over time, but we still provide classes on SQL and other tools, and expect our people can query our big data stores for what they need to manage their operations. This frees our analytics folks to spend more time on insight generation, rather than support routine analytic tasks for non-technical, non-quantitative managers.

Once you get the data extract you need, you have to analyze it—for example, build a model. A simple rule of thumb is that if it’s doable in Excel, we expect our analysts to handle it. If it takes R or Python, either because of the size and granularity of the data set, or the flexibility, we need to build models or algorithms, it’s something for the data scientists. An example would be modeling expected performance of long-tail items, where the data’s thin and it’s a little harder to distinguish signal from noise.

Bain: How extensively have you needed to adapt your technology infrastructure?

Macri: Pretty much every technology we have for supporting marketing is homegrown, tailored to our particular needs. Why? Historically, as a smaller firm, we never had access to marketing technology budgets [to go out and license stuff], but being where we are [in Boston], and with very technical cofounders, we did have access to great engineers. So it was always easier to build versus buy. This opportunity makes Wayfair more actively requirements-driven, rather than passively demo-driven. In other words, Wayfair gets what it needs, and doesn’t pay for stuff that it doesn’t use. Further, the firm shares some of its code and supports open source development. Not every firm will practically be able to be this engineering-driven, but the example is instructive.

One benefit we derive from being this strong from an engineering perspective is that our product teams are tight with their counterparts at places like Pinterest, Facebook and Google. So we spend less time trying to figure out what they’re up to, and more time adapting to their evolution and trading ideas for how to make the overall experience better.

An example of how being strong in engineering helps us is the augmented reality (AR) feature, View in Room 3D, that we’ve incorporated into our mobile app. It’s a tool that allows you to visualize how things in our product range would fit into a space in your home. We used the iOS and Android SDKs (ARKit and ARCore) to develop it.

Bain: As everything moves online and your internal data skills become so important, how have you changed how you approach “old media” and the role of agencies—if there is a role for them?

Macri: In 2012, we started advertising on TV. After a while, we observed that doing this through intermediaries wasn’t consistent with how we wanted to manage marketing. There was more of an emphasis on just hitting targets versus exceeding them. So in 2015 we brought it in-house. All of it, from media buying—we now go direct to HGTV, for example—to production. Our buyers are more motivated, and we get a better yield. The production team can do a spot for a fraction of what an agency might charge. This allows us to do all of our testing in-market, nationally side by side. To evaluate the effectiveness of what we’re doing with TV, we started by simply correlating the change in mobile traffic to our home page with our ad flights.

Bain: Can you give us an example of how your analytic orientation and control over infrastructure and process have been applied in a specific situation?

Macri: We’re relaunching our rewards program. Like everyone else who has one, we’re concerned about just giving margin away without driving incremental lifetime value. Last time we evaluated it, we studied it deeply for six weeks, looking at all the numbers and talking with a few hundred customers. After that, we killed the program completely. Because our culture values experimentation and objectivity, there was none of the political gnashing of teeth you might expect with something like this. We just moved on. Now, we feel we’ve got better insights, and we’re ready to try again.

Bain: What’s your best advice for other marketers?

Macri: In other settings we see line marketers and even analysts who declare “I’m not technical, we’ll have to get someone from IT to handle that.” In our view, this line gets drawn way too conservatively. Wayfair’s culture of experimentation [like its “new mover” program for people moving homes], its willingness to break quickly when change is called for, its expectations for numeracy if not technical fluency, and for flexibility and curiosity, and finally its treatment of marke—um, analysts as businesspeople and not just project managers, all seem ideas with universal applicability, not just within a two-mile radius around MIT and Harvard.

This interview was conducted by Cesar Brea, a Bain & Company partner in Boston and member of the firm’s Advanced Analytics and Marketing practices.