Supercharge Your Marketing with Cloud (Cloud Next '19)

[MUSIC PLAYING] DAMIAN LAWLOR: We have lots to cover this morning We’ve also got some really, really great stories from customers who are going to join us up onstage So I’m going to speak very quickly, not just because I’m Irish, but because I want to get off the stage as fast as I can and welcome some more of my colleagues and some of those great customers up to share their experiences with you So a little bit about myself I’ll just give you a little kind of personal journey I’m 11 years in Google I spent two years running the Publisher side in Europe, then I moved across after the DoubleClick acquisition and ran DoubleClick for six years in Europem and I’ve been in Cloud for about three and 1/2 years at this stage And I was really attracted to the transformation that was taking place in people moving from an on-premise world into a cloud world and what that could mean for helping companies with their own digital transformation But I was particularly interested, coming out of the ad tech world, about what cloud computing could do to really help people with their marketing and how they could interact with their own customers So I saw it as a compelling transformation story as the consumer shift to mobile, which we all saw, or the shift to programmatic advertising So if you think of those two big shifts, another really, really large shift that we’re seeing, and I love this quote from Gartner, is, this transformation of how many marketers see that they’ll be “competing mostly or completely on the basis of customer experience in two years’ time.” The quote’s from 2017 That means that it should actually be happening now So just a quick question to the audience The marketers who are in the room, are you now mostly or completely competing on the basis of customer experience or not quite yet? So yes? OK That’s pretty interesting And I think one of the reasons why– nothing’s really changed in marketing Marketers are still looking to acquire, retain, and grow customers And they want to put the customer in the center and focus on customer engagement, and really want to focus on what we call creating delightful customer experiences And that’s going to help you to predict how much people will spend in the future, predict is your customer, is she likely to spend again, how do we find others like her at scale, et cetera So you’ve got all of these touchpoints from your customers, all of this fantastic data you’re collecting, but a lot of that data is sitting in either technological silos or in organizational silos in your organization And so that makes it very hard to get a complete view of your customer So your email marketing team may have fantastic insights in your customers, but is not sharing that with your search team And you may not be getting feedback back fast enough to your product development or your customer service teams And you may also be working with– I know there’s a stat out there that on average people use 13 different marketing technologies But in fact, even somebody today told me that, actually, the different tools that collect data that sit in silos numbers in the thousands in their organization, and how difficult it is to actually pull all that data together So this is what really interested me So when I started off in Cloud, I did a global round trip, just going and meeting as many CMOs and C-suite CEOs, and pretty much asking them the same question, which was, how satisfied are you with your own ability to understand your own data on your own customers? And I always tell the joke that I have 100% record of people saying they are very dissatisfied And the joke is that there was actually one CMO who told me that they were very satisfied with the amount of the value and insight they got from their own data So that kind of broke my 100% record But then afterwards, they came back up to me and said, I only said that because I was in front of my team, and I didn’t want to embarrass them In fact, we’re really unhappy with how much value we can get out of our own customer data So what that actually means is that the statistic behind me, where 13% of organizations say they’re making the most of their available customer data is exactly 13% more than I’d met in actual discussions with all of you And what I find really interesting is that the more sophisticated the enterprise that we work with get, the less satisfied they become with the amount of value that they can get out of their own data, because it feeds their requirement If they do start seeing how fantastic insights they can pull out of their own data, then they want to get more and more insights They want to start using machine model, learning models They want to make it faster to get to those insights They want to be able to use smarter technologies And they want to make it easier for all of the people in their organizations to be able to get to those insights quickly So this is where Cloud for Marketing comes in Our overall vision is to take what is typically,

for a lot of people, this linear model of the data journey– collecting your data, transforming it, analyzing it, visualizing it, and activating it– and offering faster, simpler, and smarter solutions for you to create these delightful customer experiences from your data and from your marketing technology, and ultimately, with the power of cloud computing, to turn that more linear model into a virtuous circle Because, obviously, as you activate these campaigns, as you activate these delightful customer experiences, then customers react and supply more data about what their preferences are and what they want from you And 63% of consumers say they want brands to understand them better and use the data from their purchasing history to provide them with a more personalized solution And what we’re here to do is to bring technology to bear on that to enable you to do that more and more with your customers So to go into more detail about how we’re looking to turn that vision into reality, I’d like to welcome my colleague Biren Kalaria up onstage, who’s the global strategy lead for Cloud marketing Thank you, Biren BIREN KALARIA: Cheers, Damian Thank you Good morning, everyone So I’m going to talk to you about three things before bringing up three amazing customers who are going to come and share their experiences So the first thing I’m going to cover is how you can get started with us Secondly, why you should work with us And thirdly, going to give you a bit of a product preview to show how we’re making that process that you saw earlier from Damian even more faster, simpler, and smarter So the first thing that we’ve done– and I hope a few of you were here with us last year as well, but we’ve developed a set of solutions based on high-value use cases that we’ve heard in conversations with enterprises over the past two years These are eight use cases that are packaged to be as turnkey and high-impact as possible And each of the use cases is designed to help you make more of your data in service of both long- and short-term goals whilst minimizing the initial financial investment required to implement, and also delivering you a fast time to insight So we’ve got an email address that you can contact the team on at the end of the presentation And if you’re interested in getting a little bit more detail on these solutions, please do drop us a note, and also make sure you pop over to our booth down in the main expo to speak more to the team The second area that we’ve put a lot of work into over the last 12 months in terms of putting in things to help you succeed have been around our partners So you know that partners are an incredibly important part of the overall cloud and Google Cloud ecosystem We’ve done a few things there I’ll just touch on a few things So we have 12 marketing analytics specialized service partners available today And this list is growing all the time In fact, we’ll be launching this specialization more formally during the course of this week Specializations, as you know, are awarded to GCP partners with industry-leading competencies And these partners are there to help you from the very earliest point in your journey to Google Cloud So those are the folks on the left-hand side On the right-hand side, you have a number of data integration partners And these broadly fall into three buckets The first group, the kind of ETL and data partners, through these companies, you can access over 300 connectors to data sources that you really, really, really care about And with these connectors, you can be sure that the data from the marketing channels that you want and the sources that you really care about is always up to date and accessible in BigQuery The connectors themselves very quickly scale to your needs whilst promoting kind of agile workflows that require less planning and fewer resources to maintain The second group there are the analytics partners So working with our BI partners, we’ve created business dashboards with preconfigured templates which provide robust analytics to help you understand and analyze your marketing data And the third group are marketing technology partners themselves So these are companies that are inherently built on GCP, and can provide you with that kind of SaaS layer that you need, or the technology layer that you need, to actually make more use of any data that you have already sitting on Google Cloud Platform So we’ve put a lot of work into these areas Lots of the different keynotes this week will touch on various folks in these ecosystems as well Secondly, I just want to touch on a little bit why we’re different And this is one of the first breakouts I hope that you’ll be attending today and through the course of the week But just want to touch on a few points here As you know, GCP is a leader in smart analytics Forrester rated Google as the only leader

in their Insight’s platforms-as-a-service report last year Secondly, we have solutions designed to meet market needs So we’re going to touch on some of those in a little bit Like the use cases we just walked through and the console that I’ll be showing you in a moment, we make it easier for marketers to deploy workloads on Google Cloud Platform and generate time to insight very, very quickly And of course, we do a lot of work and we are a leader in the machine learning space with technology that is truly different than any other in the market We have over 7,500 projects powered by AI, including Google’s consumer products from Search through YouTube through Maps Secondly, to the differentiators in the marketing analytics space, I want to touch on points that you may have heard in the keynote earlier this morning about why our cloud is different from other clouds Firstly, security is at its core Google infrastructure was born in the cloud and built from the ground up with security in mind Secondly, our fully managed serverless offerings eliminate that operational complexity that is required when moving to cloud, and making cloud easy to adopt And thirdly, openness is in Google’s DNA From our commitment to open source software in areas like TensorFlow and Kubernetes to the way in which we build our products to give you flexibility, our open approach means that you can develop applications that are easy to run and fast to deploy And finally, we’re bringing you the best of Google for your marketing needs Google is inherently deeply entrenched in the marketing space through all the work that’s done on the ad side of our business And we have integrations with other Google products, like the Google Marketing Platform, Google Ads, Maps, Play, and YouTube, that help you get more from these products and from Google Cloud And we’ve also looked to bring you the best of Google for marketing, bringing technology developed at Google, like the Vision API, to help solve for marketing use cases And Google has teams with decades of experience working with enterprises of every size in every industry and in every nationality to help solve your marketing challenges with our technology So we talked about how we’re helping enterprises today through the use cases that we’ve packaged up with the cloud, kind of hexagons that are available to everyone And I just want to give you a sneak peek into what we’re building to make that data journey that you see on the screen right now even faster, simpler, and smarter So the Cloud for Marketing console is our first step in helping customers solve marketing use cases faster, simpler, and smarter on GCP And it’s in alpha now and available in every region If you’re interested to see more than what I’m going to show you in the next few slides, please do head over to our booth in the main hall to speak to our PMs and engineering teams So firstly, when a user opens the console, their dashboard will show the following things– templated queries, links to templated visualizations, starting with Data Studio, predictive models that are available at the click of a button, and activation points as well And I’ll touch on these four things now very briefly Firstly, we’ve made it faster to collect data We’ve simplified the manual upload process for getting data into BigQuery by providing the tools and step by step instructions you need directly in the console, allowing you to get your data in faster Secondly, managing all your marketing-related data is simpler with built-in tagging and access controls We’ve created a universal schema for web analytics, which allows customers to execute queries against the same schema and combine data from Google Analytics and Adobe for more comprehensive analysis and reporting And you can maintain data privacy and security by managing access controls right down to the column level for sensitive NPII data as well Super cool This is the last area The console provides smarter insights about your customers and products, leveraging templated queries and the same predictive models that power Google’s other products No coding needed Firstly, templating queries, they answer common e-commerce analytics questions and output to a UI with the results Secondly, customers can use our models to predict lifetime value and forecast demand using a simple UI Again, no coding required The demand forecasting model is using Google products to forecast ad demand and search trends, and now we’re making that available to our customers through the console And the last piece here are the activation points

So once you have done your analysis and you run some predictive models, you can, of course, get a raw output of audience segments that you can then use to push into any of the marketing tools of your choice We also have an integration with SendGrid that allows you to push those lists directly into that platform if you’re using it for email marketing purposes So that’s a little bit about the console and a little bit about what we’ve been up to to date And I hope you’re excited about the opportunity to run your marketing analytics on Google Cloud, and the work that we’re doing in this area Like I said, don’t forget to come and see us in the booth And today, we’re actually incredibly lucky to be joined by leaders from Ulta Beauty, American Eagle, and Colgate to come and share how Google Cloud helps them supercharge their marketing data First up, we have Ben Pham, senior architect, and Faris Alqadah, senior director, Data Science Innovation, from Ulta Beauty Welcome onstage, guys [APPLAUSE] FARIS ALQADAH: Thank you Hello, everyone So I’m Faris This is Ben Like he said, senior director of Data Science Innovation at Ulta And we’re happy to share our story with you of what we’re currently building using Google Cloud Platform Just a little bit about Ulta Beauty to give context So we are the largest US-based beauty retailer with just over 1,000 locations in all 50 states now So a lot of things differentiate Ulta Beauty, but one main thing that sticks out to me, and learning since joining, is that the beauty retail landscape has traditionally been segmented across price point and brand So at Ulta Beauty, we break that segmentation We actually carry across all price points and across 500 brands So we carry about 25,000 SKUs Also, each of our locations offers a full salon service with skin, brow, and hair services So it’s really a very different beauty experience when a customer walks into Ulta Beauty than what they’re traditionally accustomed to Now, this gives us great differentiation, and it also presents its own challenges in marketing, because we have such a diverse guest profile So our vision for what we’re trying to achieve in terms of marketing and customer experience, I think, is no different than what most retailers are trying to do What we really want to do is enable seamless and personalized experiences across all touchpoints So for a guest that walks into our salon regularly and has a stylist that knows her preferences and can give her great tips, we want her to have that exact same experience if she logs onto the Ulta app And so that’s quite a challenge with today’s marketing landscape when we have all these different disparate systems with different data stuck in all those channels But that’s our vision And furthermore, we want to go beyond the traditional marketing push mentality Where we’ve traditionally used these channels of marketing to merely push content and push users to perform actions to purchase or to engage with us, we really want to make that more of a two-way conversation with guests, so, one, to make them feel more engaged with Ulta Beauty, but also to collect valuable data and to make their experience better So with this vision, we identified three main core capabilities that we think are essential to enable this vision So, one, we need the ability to collect, link, and apply machine learning to all this disparate data at scale Second, we need the flexibility to take source systems in and out of that data collection process without disrupting the rest of the workflow We also need the flexibility to enable that intelligence that we’re garnering through our machine learning to any of these channels So that flexibility is key, so we don’t go into the cycle of collect in batch, run machine learning, and then market, because for today’s customer, that might be too late And that brings us to the third capability, that we really needed real-time integrations, so that if a customer is on the site, is on mobile, or has just transacted, we identify that immediately, not two or three days later, and we can react to them So by identifying those three main capabilities that were needed and for our vision, we’re actually in the process of building two Cloud-native platforms

So one is conversational commerce, and so a Cloud-native, through Google Platform The second is our personalization engine that will enable these experiences And so Ben is going to talk to us about the first platform, conversational commerce, and some use cases that we’re building through that BEN PHAM: Thanks, Faris So as Faris mentioned, our vision is to make a guest’s every touchpoint personalized and relevant So we built a compressive conversation platform for our guests And this platform will enable us to provide two-way conversation with a guest across these touchpoints and then on multiple avenues and channels So these two demo video that we show here, are what we call a Beautybot and Virtual Beauty Advisors, which were built on this platform The one at the bottom is what we call Virtual Beauty Advisors, which just helps our guests to find their skin care needs and finding their products for their skin care needs and concerns And in this case, Virtual Beauty Advisors provide advice and guidance to find the right products for their skin care needs and which go through the conversation of AI and then which recognize the daily skincare routine, their skin type based on their skin concern, and then make the relevant offers and recommendations And that Beautybot on the top is that the guests they did not receive their missing reward points for their purchase, and they have the receipt to prove their purchase So they start interacting with our Facebook Messenger, and they start initiate a conversation with our Beautybot And within our platform, we have an NOP engine that predicts the intent of the message And in this case, it is the missing reward points And then the Beautybot starts to engage the guest, asking them or looking up their account, get their details about their transactions And then by asking them to take a photo of their receipts, and then add their missing point to their reward points So now, I’m going to hand over to Faris to talk about the personalization platform FARIS ALQADAH: Yeah So as we mentioned, we’re building both these platforms cloud native And both of these platforms rely on each other, use each other Just as we described, our vision is to enable intelligence on any channel So conversational commerce is definitely one channel across social, email, bots, et cetera At the center of all that is what we’re building, called our personalization platform And to give some context, in beauty, personalization is essential The very nature of beauty is very personal to guests It’s also, within this domain, beauty is not a necessity It’s something that’s fun It’s something that our 30 million loyalty members are very passionate about So it’s become clearer and clearer to us that engaging with them and understanding our guest’s preferences and being very targeted with them is something that’s essential So the current state of our marketing, but also customer experience, is pretty good We have very rich data on our customers Like we had 30 million loyalty members 95% of our revenue is generated by those loyalty members So we have a rich data set with pretty strong marketing capabilities in our current state But I think, like most retailers, we’re trying to go beyond that and future-proof it So currently, we’re very segment based, very batch and channel-specific marketing Where we want to get to with our personalization platform is to be truly one-to-one, to be real-time and go beyond segments Furthermore, not be reactive, but be proactive and predictive And so, how are we achieving that? Well, by using Google Cloud Platform to bring in all the data from all the sources, going beyond transactions So if a guest transacts with us, it’s almost too late to market to them But going beyond transactions and looking at all the signals, maybe through conversational platform, maybe through what they’re doing on our sites or other channels, and then feeding all that to real time machine learning and not batch-based, and enabling those capabilities to be shared across every channel using a flexible infrastructure and service-based architecture So Ben will actually show some of what that looks like under the hood BEN PHAM: So now we know what the conversation and personalization platform are, but how do we make it all possible?

So this is the overview of our architecture design that show how do we leverage Google Cloud Platform to make it all possible for us So first, we use the cloud identity and security to manage users, roles, groups, services account in GCP And the goal for us is to maintain a privileged access model that we want to make sure that every team, every members, every individual, are assigned to the right access, public access, required for them to do their job And also, we want to make sure that the GCP identity access management roles and privileges are assigned to the groups, not individual And then we manage these groups and the group membership within our Alta Active Directory And we push these groups to GCP via Google Cloud Directory Sync So yeah, first you can see here the ingest the store layer Then this is the layers where we have both platforms are running on this layer, on the Google Cloud Compute Engine, listening to the real-time event from the external systems, such as the Facebook Messenger, as the web hook, as you’ve seen earlier in the Beautybot video And then that web hook is hosting on the cloud network services sending the event directly to the conversation platform And on the left-hand side, we also see the [INAUDIBLE] system can send the data via data stream or batch into our platform via the Cloud Pub/Sub Within the conversation platform, we have a interplay component, which is orchestrate all the incoming and outgoing traffics in the front, and also communicate with our NLP engine for the message intent or intent message– for example, the missing reward points that we saw earlier And then also, interacting with what we call it the quasi running inside the personalization engine This is the place where we have a complete 360 data view of our guests with real-time intelligent models And then also, it’s playing the key role in the Virtual Beauty Advisor that we saw earlier So now, we have the data coming from the batch, data streaming, real-time event, real-time data integration flowing into that ingest and store How do we store all this data? So the answer, we use the cloud storage, what we specify as unstructured raw data And then the important thing is, we want both platform be able to use and read the data from the cloud storage for our continuous machine learning, machine training, models training, is a switching point for us And now next layer, we have the process and transformation, which we use the Cloud Dataflow to process and enrich the data into our BigQuery as our analytics data warehouse and also the BigTable and DataStore for our in-house application use, which is sitting in the analyze and explore layer And lastly, we also enable our DevOps team to use Stackdriver’s container registry for logging, monitoring, CI/CD pipeline file platforms So now we all know how GCP is playing a key role in our marketing platform, enable us to surround our guests for every touchpoint But at the same time, we want to make sure– it’s important point for us to apply our infrastructures, standards, practices, on top of GCP for the long term So here’s a few key highlights that we set as our GCP foundation For example, backup recovery strategy We need to make sure that we have defined the region zones, disaster recovery We had the big automation for scalability We also need to integrate DevOps, our standard tools for monitoring and operations, governance strategy around the technical support, optimization processes, best practices, and also compute engine and storage standards for the OS patching, licensing, pattern for the machine type and cloud storage And security risk, also important for us to continue to manage the user roles and groups And then lastly, the networking We need to make sure that we apply the shared VPC clients via VPN, and DNS, and all of that Thanks, very much [APPLAUSE] BIREN KALARIA: Thank you, guys Thank you Thank you, guys And next up, we have Chris Stephens, who’s the Vise President of Data and Technology from American Eagle Outfitters And I will not take the clicker with me this time There you go, Chris Thank you, very much CHRIS STEPHENS: Thank you, very much I was just telling my team I didn’t

know I was allowed to have friends up here So all the work I’m about to talk about I didn’t do, by the way Some people out here in the audience Raise your hand if you did the work There we are Alrighty So I’m Chris Stevens I work for American Eagle Outfitters And I’m going to talk a little bit about what’s it like to be your customer You saw when the Google Marketing Platform focuses around customer journey and understanding that journey with your customer And we’ve spent a lot of work partnering with our CX and UX team to actually build out a really comprehensive customer journey that I’m going to talk a little bit about And then I’m going to talk about the work that my team is doing to measure and understand that for each and every customer, every time OK So first, American Eagle Outfitters, if you don’t know who we are, we really are two primary brands– American Eagle brand, maker of the best jeans in the world By the way, the number one women’s jeans brand in America, if you didn’t know that And also the women’s brand, Aerie So American Eagle is all about empowering youth and individuality Aerie is really about celebrating real beauty in young women So I say these pronouns we and us, because the reality is I’m a vice president of something-or-other at a $4 billion company I didn’t do this work, my team did all this work OK So you heard a lot about you’ve got to understand who your customers are, what they want, what are their intentions, how do you use machine learning and technology, and so on, to make sure you’re there when they want you to be there? And I think for us, it comes down to listening, right? You’ve got to be there listening and engaging in a meaningful conversation with your customer, every time, when they’re ready to engage in that conversation So if you don’t know this Ramones song, the next line is, you gotta learn to listen or you’re going to get burned So I think, for us, it’s really about listening to our customer and understanding who they are at the moment when they’re signaling intent So this is a Googleized representation of a high level picture of about 10 people’s five years worth of work We have this really comprehensive customer journey map that our team has done that I can’t share with you It’s terribly proprietary But you can understand the idea is, as a customer’s engaging with us, they go through these phases of engagement, whether it’s from influences to consideration into in-store digital browsing, purchasing, usage, returns, and so on And you know, the CX team, like I said, did all of this incredible work mapping out the journey, doing ethnographic research, talking to customers, understanding what it is like to be that customer when they’re engaging with our brands And then you see the little gray call-out box there That’s really the call-out then to the technology team, or my team in particular, in that our job is to understand and measure what’s happening for 20 million people at a time as they move through this customer journey So it’s important You know, we’re a business, so we’ve got to tie this back to what we do as a business, right? So we have OKRs and metrics in place that help us measure this journey and understand what’s important– so things like conversion, revenue per visitor, average order volume Those are things that are important to online retailers– Ulta guys, right? I hope And then we put metrics and goals in place So we want to move conversion by 2%, let’s say We want to increase average order volume to $50 and so on And those are quantifiable things that we can understand as we measure customer interaction, how they shop, what they buy, what they browse, and so on But then the last part, which can be the hardest piece– and I’m going to talk a little bit about what we’re doing there– is how do customers feel about it when they’re interacting with us? How does it feel to be our customer? Not how many of them are there? What do they buy? What did they buy with this other stuff, and how can I make them buy more of that stuff? But how do they feel? What do they say? But not one at a time, how did they feel and what did they say 20 million at a time, all of the time, every time? And how can I be there listening as they’re telling me those things? So two themes I’m going to talk about One is personalization So similar to what the Ulta guys said, we have a pretty comprehensive personalization program under way We’ve got a lot of those capabilities built out on the Google Cloud, thanks to my awesome team again But again, we’re not just doing personalization for personalization And quite frankly, we’re not even just doing personalization to be there for the customer when they want us to be and so on We are running a business, so we want to make sure we’re doing personalization in a way that ties back to the things that our business cares about So calling out four particular examples– when we engage customers in one-to-one experiences, we see significant revenue impact I think the number there is 21%, or something like that, of our customers that don’t interact with our personalization tactics Similarly, on the conversion side, we see like 4x conversion over customers that don’t interact with personalization So this might be a sales pitch for personalization

Increased units and outfitting– so we’re a fashion retailer We sell clothes, so we want to make sure customers are outfitting themselves They’re not just coming for one particular item or another But we want them to understand the outfitting intention of our merchants And so we see– I forget there– it’s 18% or 19% increase in the units the customers are buying when they’re interacting with our personalization tactics And then similarly, for higher revenue and margin per sale So the point is that the team is doing amazing work, and it’s having an impact And it’s technically interesting and invigorating, I think, to come to work every day and work on this stuff Then we can go talk to the board of directors about the impact we’re having in 4x conversion and 28%, or whatever it is, increase in average order volume Those are meaningful numbers for the board of a $4 billion public company So how do we do that? This is just a quick overview Again, my team’s here, if you want to ask them questions But you know, we have the problem that Damian talked about at the beginning We’ve got all this data It doesn’t natively exist in Google So on the left, your left, we’ve got to get all this data in there to personalize people’s experiences I’ve got to have customer data, product data, browse data– tf-idf is an awesome recommendation based on text distance frequency for items in our catalog and our assortment– and all this customer data in there so that we can understand who’s looking for what, what they might be interested in adjacent to what they are browsing at the time, right? So we put it through this pipeline on Google, apply some machine learning– we have this robust recommendation engine that we built– and serve those out in real-time in a recommendation service that we built that runs an app engine that our site can call in real-time I don’t have a similar architecture, but we also have a real-time pricing and offer service that we run similarly So our customer can come to the site and, based on segmentation work that we do, I could make pricing and offer recommendations or promotion recommendations to that customer, based on, you can imagine, similar logic OK so voice of the customer– so this is about, how does it feel to be our customer? And how do I measure that? So a lot of this gets back to sentiment analysis, as you can imagine But again, we’re trying to tie this back to running a business So our vision here is putting content about what customers are saying and how they’re feeling about our products and our experiences into the hands of the people in our business that are responsible for those things So you can imagine our sourcing and manufacturing team really care if our jeans are smelly, if they’re bleeding, if they’re tearing out at the seams, if the fabric is pilly, those kinds of things And you can get anecdotal one off here and there A customer said this, customer said that We got 10 returns in the distribution center because of this thing But through the service that we’re building here, we can put dashboards in front of those teams and measure every interaction that our customer has and everything they say every time about how they’re feeling about every one of our products So I think that’s going to be pretty powerful for that team, similarly, for the merchant teams So the job, if you don’t know the job, of retail merchant is to figure out what clothes to put together into an assortment, how those things ought to go together at scale We sell 60,000 different units a year, so that’s a big problem to solve And then anticipate what they think people are going to like, and then try to read that through the season to understand what people are feeling about the products that they put out there in their assortment And sometimes, things move faster and slower than they thought they were going to And we can use sentiment analysis or our voice of the customer engine here to help them understand not just this thing’s moving slow, market down, drive-through the inventory That’s kind of retail 101 But maybe the things are moving slow because of something else And people are saying something about the products, right? And so we ran an example last year where it was around this time of year And our men’s shorts were sort of selling slow And the merchants felt really good about the assortment they had put together, but it was just selling slow And in a traditional retailer, you’d think, well, OK, it’s selling slow We’ve got to move this inventory Mark it down Get it out of here Take a hit on margin and suffer for that, right? But through looking at our voice of the customer, we were able to see that, by and large, customers were feeling pretty good about the assortment They were saying positive things They liked how the assortment looked They liked how the product felt, right? But it just wasn’t selling So we were able to kind of hold the line with some confidence to say, no, people like this And it’s not just, hey, do you like this? It’s hey, everybody, 20 million of you at the same time, how do you feel about this stuff? Do you like it? And the customers are saying, yeah, we kind of like this stuff So we kind of were able to hold the line in terms of markdowns And then by the way, the business came back and did great So it was just kind of one other indicator to help us have some confidence in the assortment that we’d put together OK So similarly, the pipeline, again–

to Damian’s point at the very beginning– the problem is that I have all of this data It doesn’t natively exist on Google, right? I have product reviews I have purchase surveys We have social media data We have data from our call center From the call center, we have chats We have calls We have emails There’s a lot of things that people are saying, right? Returns data– there’s surveys on the returns We have to bring all of that stuff together and amalgamate it into one thing, so I can understand with one taxonomy, what are people saying? And how do they feel? And that’s where this kind of sentiment analysis engine lays on top And then for now, we’re kind of natively building in Tableau If there’s a Tableau person in here, shout-out Kind of natively building a little app interface in Tableau so that our merchant teams and some of the business folks I talked about can interact with that data OK, I wanted to talk about a voice of a real customer You know, Aerie was the first brand to not retouch, to not airbrush our models And we’ve kind of taken that one step further in the past year, and we’ve actually recruited real women to be a part of our marketing and advertising campaigns And if you go to our website and you want to see how we merchandise our products on the website, those are all real women wearing our clothes And so you can see in this particular example– I think there’s a laser here– so this young woman was on our site And she had had a heart transplant So she has one of these big zipper scars down the middle of her chest And we just so happened to have another young woman in a similar condition who was merchant modeling some of our clothes on the website And you can read for yourselves, but you can see that something along the lines of she went to bed feeling a little more comfortable with herself that night, based on what we were doing So you know, it’s not just machine learning, and AI, and reasoning over people’s intent to try to sell them more stuff No I think the reality is that if you do it right, you’re listening to your customers, you’re paying attention to what they say and how they feel, and you can have a real impact on people’s lives BIREN KALARIA: Last up today, we’ve got Tim Booher, who is the Global Analytics Lead and Chief Information Security Officer from Colgate-Palmolive Tim, please come on stage Thank you [APPLAUSE] TIM BOOHER: Great Thanks for the opportunity to tell the story of how, at Colgate, we’re using this technology to really pull together a bunch of different data sources, empower our users, and have a lot of fun For those of you that don’t know Colgate-Palmolive, we have over 3 and 1/2 billion customers for different product lines, home care, pet care, personal care, and oral care, which is the majority of our business Very global country, with the majority of our sales outside the United States Very established in developing countries And you can imagine that makes a big challenge for us to understand, with such a diverse group of customers, what they think about our products, what they want, and how they respond to all of our creative work that we’re doing My background before coming into Colgate was with the US government I had the opportunity to run part of DARPA’s machine learning portfolio Whether it was fighting human trafficking networks, using AI to infiltrate terrorist networks and find radicalization, or trying to put terrorists behind bars before they could act, all that required us going out and getting data We didn’t look for a tool that provided that We found the smartest people we could We built everything using open source, and we had all the computing power that we wanted I also had hundreds of people and millions a year to do it When you’re in a commercial construct, what’s amazing to me is everything that that I got mostly from organizations like MIT that built everything in C and were able to do entity extraction and all this work, you can get instantly now on Google’s cloud And that’s incredibly exciting to me, because we can get a lot done very quickly In this case, we looked at consumer sentiment as a starting point And it really opened the door for us to do a lot of other exciting things The first experiment we did with Google is we looked at several hundred thousand reviews of our customer’s products online, then wrote a bunch of Python glue code to fit together all the master data and add some structure to the data that was out there, and then used Google’s Natural Language API And this still floors me a little bit because I had hundreds of people building capability that got about half of what this has And we were doing some really awesome stuff with it But here, I can use a smaller team, Cloud infrastructure, scale up what we need, and get a lot of work done more quickly What’s amazing about this is entity extraction used to be one of our hardest, most technical problems It comes out of the box, along with sentiment score, salience of how relevant, and even magnitude

of feeling on different topics that are there Once we can do entity extraction, that’s where it gets really exciting Because even when we start with reviews, depending on the business user– and there are eight different business users that are now using this data, everything from supply chain, to e-commerce, to product development, to marketing– because it’s one BigQuery data set, a number of users can do a lot of different things with it And because we can rearrange blue hexagons pretty quickly, we can be agile and respond to our business needs out of that The first set of metrics we looked at, though, were looking at consumer sentiment across a variety of brands and topics, doing the entity extraction from that, and then understanding both the whitespace in products, but also being able to respond quickly to the different complaints, or the different feedback mechanisms, consumers were providing to us One of the really cool things we could do– in one case, we had a user that had a team that spoke two languages, but needed to operate in 10 different countries It was very easy to add a natural language processing, a translation API, on top of that and then quickly use that team to get their arms around much bigger business We really feel like where we’re going to fight in this space with an army of capability that we can quickly give our teams distributed around the world We can also incorporate new data sets very quickly So a source like Twitter, much more structured API, directly bring it in, correlate it to other data sets that are there When I was in the government, we were doing really cool things with tweets We could geolocate based on text alone And at Colgate, we’re finding a lot of great information’s coming from the correlation of disparate data sets and the ability to bring data out of the tools that we have into a common construct, and through using ML and code where we need to Also, using Tableau or the Google Data Studio has been the most helpful for users that don’t have a coding background, because they’ve been able to rearrange and build their own dashboards They don’t get stuck with a static dashboard that presupposes the question And they have the ability to really explore the data that they’re working with Just to show the example architecture that we used for this sentiment analysis, you see a common repetition of the Google Cloud Storage capability throughout that That’s because storing data raw is an important lesson learned in the past and very relevant in this space, as well When we are going out and acquiring data from a number of sources, if we might start on an e-commerce site, ingest all the reviews But we might then need to go to something like a Kickstarter or an Etsy to see where trends might be starting and to follow them down to their very source For that, we need a lot of flexibility to direct what I tell our leadership is a spotlight to go out and get us the data that we need One of the most important things is storage is cheap, people are expensive, and time is critical With that in mind, we store everything raw because we have to go back to the original source data and extract extra information that we can find there Once it’s there, we use Cloud Pub/Sub, also setting the stage for streaming capabilities, particularly important in a supply chain environment But here, it allows for batch processing for whenever reviews arrive, Pub/Sub kicks off a cloud function that then scores all the metrics, produces all the insight around the data, pushes it back into another cloud storage bucket, and eventually into BigQuery The reason we do this temporary storage along these different stages is because it allows us to store data raw, which our data scientists can go back then and extract extra information from And then finally, once it’s in BigQuery, from a CSO security leader perspective, it’s very helpful It’s an append-only database I can give my teams a lot of freedom And by working closely with Google’s security teams I can distribute this capability across the company really quickly and do so knowing that the security is in a construct that my SOC is monitoring and my teams can use very quickly So I’d love to talk about the details of this more I’d love to hear stories about how others are using this capability to quickly move, understand, and act on all the information that’s out there And thank you for your time [MUSIC PLAYING]