Building a Data-Driven Go-To-Market Engine in Retail - with Michael Tambe of Amazon

Foreign.

0:00

Welcome everyone to the AI and Business podcast. I'm Matthew DeMello, senior editor here at Emerge Technology Research. Today's guest is Michael Tambe, head of Scaled Insight Science for Amazon Advertising. Mike has led data science efforts in sales and marketing at leading edge companies like Amazon ads and LinkedIn. Through these experiences, he's become an advocate of enterprises building a quote, data driven go to market engine. He joins us on today's podcast to talk about what that means along with the challenges and possibilities of new emerging AI capabilities. A quick note that the views Mike expresses on today's show are his and his alone. And now, without further ado, here's our conversation. You described your work as building a data driven go to market engine. I think it would help to have a strong enterprise definition on that for our listeners. How do you define that within Amazon?

0:07

Yeah, so the way I think about this is having a data driven go to market engineer. You can even argue nowadays you can call it an AI driven go to market engine, but it would be basically having your sales, your marketing and all of your in product messaging really triggering and based off of a data driven platform that is continuously improving to, to deliver the right message to the right person at the right time to help them get the best use of your product. Right. Right.

1:10

Now, given that backdrop and now that the listeners have a, have a really strong sense of your expertise, what do you see as some of the major challenges in the retail and e commerce space right now?

1:36

Yeah, so I think, well I'll talk about the major challenges overall, then I'll kind of dive into the retail e commerce space. And so I think, you know, this concept isn't limited to retail and e commerce. Lots of companies outside can apply this concept. And I think the biggest thing would be how do you sort of break down some of the, how do you connect the dots and break down some of the organizational silos that exist here? So let's take B2B sales for example. Kind of I'll talk about it in a general way, then I'll tie it back into retail. So you basically have two big problems. One is how do you plan, how do you set up your salesforce and then how do you actually help the salesforce succeed? So the first one is all about okay, which accounts am I going to go sell into? You know, how many salespeople do I need to hire? Is it like 1 for 10 accounts? 1 per 100 accounts? What's their target? And this is a huge optimization problem that has massive implications to the top and bottom Line, right? If you're not selling into the right accounts or if you're over, you're understaffing your salespeople, you're just leaving opportunity on the table, right? If you're selling into bad accounts or you're overstaffing your salesforce, you're just wasting their cycles and burning cash, right? If you set your targets too high, you're going to lose great people. If you set because they're going to miss their quota and they're not going to get paid. If you set your targets too low, you're going to be paying out huge bonuses. And so it's this huge optimization problem. But for so many organizations it sits in like some like operations or finance and it's one of these things where survival is success, right? If you can somehow manage to get through the year where the sales people and the finance and the operation people agree and everybody has, has a book of business coming in, you've succeeded. So the amount of investment of the optimization has been I'd say a lot less than it could be. So that's like one problem. Then you kind of take on the other side. You have kind of within marketing or sometimes you have like sales productivity teams, teams to think about, okay, the day to day activities. And so a lot of this and outside the retail space comes in like demand generation or you know, outbound sell. You want to acquire new logos. In retail it's a little different because your customers are your sellers. And so it's a lot of like upsell motion versus acquisition motion. But you know, like a lot of it comes into okay, how do I create leads for these people, how do I give them the opportunity to network in? And then for existing customer sales, which is where I tend to think of a lot more, which is heavily relevant for retail. It's you know, how do I deliver the right insight to add value? And you know, typically it's just a lot of tableau dashboards, right? People go compile this stuff across the way and the tools that can help these people, salespeople in their day to day lives. It's the power can be immense, the impact can be immense. But the impact can be somewhat harder to articulate. Like particularly if you're not just generating leads. Like I hear a team saying well we saved 400,000 hours or whatever, however many salespeople hours. But it's like how does that translate to revenue? What are they doing with it otherwise? And so it's like imagine if you can combine these and say okay, we're going to plan our sales team around a bottoms up view of what they actually do. We're going to say, okay, managing account is like an upfront conversation at the beginning of the year to set goals, then quarterly check ins. On this and monthly check ins, you can kind of come up with how much time that takes and you can really kind of plan it there so that you have a clear way of knowing all those planning questions that are so important in getting the right answer. And you can connect all the efforts you're doing with the salespeople to the planning there. So if you create a great tool that saves salespeople time, you can then bake it in that you can have more accounts for salespeople. They can just do more next year and it can kind of continuously improve on a quarterly or an annual basis, however you're doing it. Right.

1:48

So it sounds like the conversation is moving from time saved to how, what volume of products can you put in the time saved? It's almost like we're connecting the dots to the next end of the question. You also mentioned the bottoms up view, which, which I find really intriguing here, especially just kind of given in this moment that we're in for lots of different forms of AI. We have generative AI waiting in the wings kind of coming up the rafters. Tell me more about this bottoms up view. And I'm wondering how does it look for companies that are doing it right now versus what they're going to be doing in the next three to five years, Especially as we find in this kind of moment of AI, it's a lot like that early Internet, you know, everybody's standards and in 1998 where you know this website's going to load in 20 seconds and isn't that wonderful? And by 2004, that's, that's preposterous and a terrible problem, you know. So what do things look like now from this bottoms up view? And how might it look, say in the next three to five years?

5:35

Yeah, so this is where I think you see the divide between sales and marketing in the current state and how that's going to start. Really hopefully, if companies do it right, start blending. Because you know, marketers often have pretty good tracking. They know exactly if somebody's seen an ad, when they've seen it, if they clicked it, they have a kind of a bottoms up view of when they've engaged with you. Now there's some complications with tracking and cookies and all that, but let's say email or some channels there. Right. So they can Build kind of a bottoms up view of what touch points they want to have customer throughout the entire year, throughout the entire life cycle, sales often ends up being a black box. And so we know kind of you give sales some stuff, you give them some material, they do something, they log something in the CRM. But for companies to really connect the dots, they're really going to need to close that loop and really get good information about what is sales talking about, which customer and when. Right? And that's not easy to do. Like, so, you know, okay, think about it this way. Suppose you say, okay, I want to, I want to know for every sales rep, what meetings are they having with their customer, what topics they're talking about and what insights were shared. Right. So you know, every CRM has functionality to log meetings. But, you know, are the sales reps going to do it right? In my experience with, with sales, if their compensation isn't dependent on something, they're not going to do it right. So the only parts of a CRM that are accurate are the parts that need to be accurate for the sales reps to get paid. So then you could say, okay, well, I'm going to require them to do it. I'm going to say, like you have to do, I don't know, four, 40 meetings a month or whatever. Well, okay, the log, 40 meetings. But the CRM will reflect what is needed for the paycheck. But so I think the better way is to really think about their workflow and have you add value there. And this ties a little bit to what I'm working on at Amazon right now, which is, okay, I'm a sales rep, I want to prepare for a meeting. I know that in order to engage and add value to my customer, I need insights. Right now I go to a bunch of different places. What if there was one place that I could go and I could say, I'm meeting with this customer, we're talking about this, and I can just compile and pull all the insights I'm going to need to make that a great meeting. So there's two things to the sales rep. It adds immediate value because you've saved time for me. You helped me get the insights I needed in one central place. But for me as the company, there's two things. One, I'm getting that tracking. I'm now having more probably the most trustable source of what you're meeting with, when and what you're talking about. And I'm also giving myself a platform that if I want to start giving recommendations and So I think what you're going to start seeing is a little bit of this moving towards, okay, the best companies are going to start putting in the workflow tools to track all of this stuff to get the touch points and to build the bottoms up view. And so one framework that I think is very, very helpful in the space, though it has some problems is Gartner has this very famous framework of predict, sorry, descriptive analytics to predictive analytics, per prescriptive analytics. So the first thing you're going to do is you're just going to tell the metrics so you can go, I want to talk to my customer. Here's what your spend was, here's how many people are buying. Here's kind of what they look like demographically descriptive analytics. Then you can kind of go to Prescriptive and say, hey, this brand is not doing well in the shoes category right now. Their products in the shoes category aren't doing really well, but they're going to lose sales unless you go talk to them. That's predictive. So then you can have all the insights to support that. Prescriptive is they're going to, you know, they're going to do bad here. Talk about X, Y and Z. So it's very much like almost like a punch list. Right. And I have this joke when, you know, because some people say, well, are sales reps really going to like work off some to do lists? And I'm like, they kind of work off a to do list right now. It's called their inbox, right? In their inbox, they're going to address it. And so, so you might as well have it more value driven rather than noise driven.

6:32

Yeah. And just going back to your point about like, you know, if you want a salesperson to do something, you have to tie it to the pay. It sounds like in terms of your better strategy, it's more about give them the tool that they know and they're going to see and they're going to touch. That's going to make their commissions better, that's going to make their sales better and they will follow it because, you know, because of the power of that tool and them actively using it. Do I have that right?

10:23

That's the way to think about it. Right. So you want to create these. I mean, Amazon is kind of famous for these virtuous cycle concepts. These fly. Right. You know, because the Amazon overall flywheel is like, you know, lower price, better selection equals more people. Gives us the ability to do more price, better selection. Right. It's kind of this, of like, if you give them a place, if you give them a tool that helps them, they'll use it more. That gives you the data to make the tool better, et cetera, et cetera, et cetera, et cetera. And so you want to try to figure out how to get that cycle and ultimately building that kind of, that bottoms up view. Because then we could start doing is say, hey look, for example, you can segment customers based on how you engage with them. Right. Like a lot of customers might be like, okay, well we're going to have like, you know, thin serve versus insurance versus, you know, whatever. Or it's like accounts that spend more than $10,000 versus the ones that spend 5 to $10,000. But what you really want to do to make a salesperson very effective is, is to give them a uniform sales motion. You want them, you don't want them having weekly performance check in calls with one customer and then planning some huge product launch with another. You want continuous same type of thing, same customer, just get in the rhythm of it. And so if you build that of like, okay, well these types of customers, we have lots of calls with them and they're on this topic. Whereas those types of customers, you have fewer calls, but they're on those topics. You can even start thinking about real reorganizing and configuring the, the sales force so that your, your people have more specialization and more efficiency in that way.

10:47

Absolutely. And just in terms of, you know, the potential problems that go into building systems like this, a great thing about talking about talking to folks from Amazon is you guys are on the cutting edge and therefore have the scars to show for, for being first into the fire. What are some of the landmines companies will encounter trying to build these systems?

12:13

You know, I talked about this kind of, this, this framework that I think a lot of people are following is kind of descriptive analytics to analytics to prescriptive. And you know, framework like that works great when you're in like a fulfillment center. And so the exact tasks are very clear. But in the sales, I think we're finding that making that move from predictive to prescriptive is very challenging. In fact, that framework might even grow break. But the good news I think is that generative AI might have some opportunities for us to address that and actually have a new way of thinking about it.

12:34

Yes, yes. And we're running out of time a little bit on today's show and I do not want to try to cram everything we might say about generative AI. Into a hot 5 minutes to just try to stay in our typical 20 minutes for our listeners. So we'll encourage everybody to check out the next episode for which we'll have Michael on Michael. Until then, thanks so much for being with us today.

13:07

Thank you so much. Look forward to it.

13:26

Before we close out today's episode, I think a very small note that's worth pulling apart for at least the outro for today's episode is where Mike talked about how B2B pipelines are taking cues from B2C pipelines. And this is something definitely we're seeing, especially in the logistics space, the manufacturing space. It's definitely something where at least the lessons of customer experience and I'd like to say this is totally anecdotal, just purely my observation. I think most of this comes from the financial services industry, but at least the headways that they've made in customer experience and how to bring technology to that space, that's what's trickling out into other sectors. We've talked a lot on this program about the impact and how health care has reflected a lot of these changes in their own way by building patient experience pipelines. I don't think B2B is really any different other than that they will need to put their own spin on it. A really great episode to listen back to. For more about this is with Carlos Quizzeda of Hewlett Packard Enterprises. That episode, and it's on the AI in Business podcast. That episode premiered August 8th of this year. Definitely check that out. It's called AI and B2B Customer Journey Signals. Really, really, really interesting stuff going off the beaten path of retail. And we were very, very glad and honored to have Michael on the show today. Many thanks to him for joining us. On behalf of Daniel and the entire team here at Emerge, thanks so much for joining us and we'll catch catch you next time on the AI in Business podcast.

13:37