AWS re:Invent 2024 - Achieve personalization at scale using Amazon Redshift and Twilio (AIM265)

okay well hello everybody I hope you can hear me okay this is my first time doing something with headphones uh great to see you all and Welcome to our session achieving personalization at scale with Amazon red shift and twillo we're incredibly glad uh you're here today especially with all the great content and different options you have to choose from uh we're glad to see you in this session we've got a lot to cover but first let me kick it off with some introductions I'm Robin grel VP of product for twilio Segment and I'm joined by my colleague Asha chakra Bari the VP of product for twilio Builder experiences who you'll hear from in just a bit you want to say hi real quick all right great so here is what we've got in store for you today first we're going to explore the challenges of building personalization at scale something I'm sure many of you have grappled with firsthand we'll break down why it's so challenging and why this challenge hasn't gone way yet despite a ton of effort um then we're going to dive into the core pillars for achieving personalization at scale including how AWS and twio segment fit fit into this picture after that we will bring it all to life with a live demo and finally we'll hear from Matt miritello VP of engineering at the skim and monasi jaganatha head of strategic partnership at AWS and they'll share some realworld examples of how companies are tackling personalization at scale all right so why are we even talking about personalization today why does it deserve a spotlight well it's simple companies that win create personalized experiences think a bit about the businesses that you interact with uh and that you love most they don't just slap on a high Robin uh on an email and leave it at that they strive to make every interaction feel effortless convenient and tailored to you and the numbers back this up according to McKenzie custom companies that excel at personalization generate 40% more Revenue than those that don't so knowing your customer really pays off less than half of companies think they're doing a good job at this and what's worse only 15% of their customers agree so let that sink in despite how valuable personalization is and the investment being poured into it most customers feel that the companies they're interacting with are falling short and it's not for lack of trying uh businesses are pouring billions of dollars into this Pro into this problem 96 billion is poured into CRM tools investments in CRM tools alone so with that investment like what's the holdup what's the problem here and the short answer is it's a really hard problem and that problem all starts with data there's data in every corner of the business it's hidden it's siloed off from other teams colleagues and systems and the various tools and systems that are used across these teams they don't play nice with each other they're not integrated they're not connected with the rest of your Tech stack they're they're just simply disconnected and don't talk to each other and because of this disjointed view of customers it makes it impossible to know which touch points or channels are best to engage with them especially on an individual basis let alone at scale when you're working with thousands or hundreds of thousands of customers and all of this makes it incredibly hard to connect the dots across teams systems and truly know your customers and what's worse your customer facing teams who are interacting uh with your customers they don't have this context either so to solve this problem requires focus on three key pillars contextual data Communications data and trusted AI I'm going to drill into each of these let's start with contextual data when when we say you know contextual data we mean the data that gives you the context about your customers and helps you understand them holistically it's your objective data basic information name address purchase history loyalty points the second bucket is behavioral data what are customers doing on your website or your mobile application how are they browsing are they adding items to their cart are they abandoning items in their cart and the last bucket is conversational data what is being said between you and your customers this can be via phone calls SMS and email changes or even on messaging to tools together all of these data points create a full context Rich picture of your customer and without that complete picture personalization falls short so making sense of all this contextual data is arguably the most important part of providing personalization and because if you don't have a clear sense of your customer you certainly can't personalize for them accurately or effectively and texax today have become so convoluted you know something that that looks like this here um it's so convoluted so complex that it that it's really a Herculean effort to try and connect and reconcile all of this customer data that's out there let alone doing it in a consistent or a real-time manner but the truth is good customer profiles are the heart of personalization and without that complete customer context you'll never be able to deliver the realtime impactful personalized customer customer experiences that the consumers are demanding today so this is what you want your profile to look like a comprehensive view of all of your customers activities their characteristics you've got some AI in here where predictions with with this data you can start to do things like How likely are they to convert what's the next best action which we'll hear more about in just a bit here and so how do you get something that looks more like this well this is one of the key roles of a customer data platform or a CDP twilio segment CDP helps Stitch all these various data points from various systems diverse systems into a single unified identity resolved profile with all of the context about them including the data that may may be housed in a data warehouse like Amazon redshift so with this accurate and reliable customer profile you can more effectively Target the right customers and personalize your Communications with them from any of your business tools from marketing use cases customer service eCommerce or any use case so now for the second pillar I'd like to hand it over to Asha to talk about Communications oops thank you Robin um so for our second pillar why is communication so vital to achieving personalization at scale well it's because every opportunity that your business has to actually interact with your consumers matters irrespective of where you're at in your customer acquisition life cycle whether it's whether your customers are considering your brand whether you are trying to convert prospects to customers uh whether you're trying to build relationships for retention or ultimately just cultivating Advocates uh for your business so it's not just about what you say though it's about reaching people on their preferred Channel using the right systems and the right technology while staying in budget and at the same time ensuring that your communication and your messages are getting delivered in fact 91% of global consumers actually expect businesses to reach them on their preferred channels but when we say preferred channels it's kind of a loaded term and that's because preferred channels it varies on on a number of different dimensions such as region your use case uh the generation you're dealing with as well as consumer specific sentiment which can be anything from perceived level of urgency to something as complex as brand loyalty and meeting these preferences is hard most of the time providers often cannot handle the global scale through throughput and compliance needs leaving your consumers without receiving your messages siloed teams and channels are disconnected leading to communication preferences also being disjointed to add to the complexity consumer preferences change in real time and lastly how do you know that you're actually interacting with the real human being on the other end in in and and not wasting a budget on fraud or spam when as businesses when we miss the Preferred channel to engage with our consumers we miss the mark and Lead that leads to opt outs as well as overlooked messages but what if what if you could deliver one seamless integrated experience across channels across teams how would that look like so I'll use an example that all of us are going through right now um it is the hotel experience we're all here in Vegas um let's say you you're going to check into your hotel you walk up to the front line you're waiting to check in you you go you exchange some information you get your room key and you go up to your room now the best best case scenario for personalization is that you see your name on the television hi Asha but what if instead of having to wait in line you get a text message on your phone letting you know that your room is ready and that you can check in with one click what if you get up into your room and you get a push notification from the hotel asking you to confirm if you'd like to order this dinner and a smoothie because they remembered that's what you ordered last time and you just click confirm and you get your food what if your hotel knows that you're here for business and sets up a fast and easy way to book your ride in the morning for your first meeting and they take it a step further they know that you're a VIP who stays at this hotel pretty often and they give you complimentary late checkout and to top it all off at the end of your stay you get an email saying thank you for your stay here's 10% discount on the beach resort that you normally go to with your family now that is personalization however personalization does not matter if there's no person to actually engage with on the other inside of that Communication channel this is why verification is essential to the online relationship with your user by verifying that a user is actually who they say they are you can reduce spam and fraud on your site and also ensure the user security and protection so this ideal scenario can be a reality and this is something that Tulio works with many customers on by by relying on tulo's Global suler Network you can ensure that your recipients aren't left waiting for your communication you can too provides a bread of communication channels for you from SMS messaging email so that you can communicate on uh your consumer preferred Channels with Rich centralized insights and with consideration from message failover delivery ensures that you're optimizing for reach while also attaining global scale and and above all establishing trust with verified users and also ensuring that you your ver you're you're verifying trust with branded senders and by doing that you're you're ensuring that you're engaging with real people by authenticating users and you're also letting consumers know that they're engaging with legitimate businesses through branded features such as RCS and branded calling and so this brings us to our final pillar that Robin alluded to before which is trusted AI so I'm sure most people here are not surprised to learn that AI is one of the core pillars of any company's technology strategy today um and for good reason uh almost 88% of companies actually say that they will be using some sort of AIML Tool uh within the next year however only 15% of Those comp companies actually are able to realize any sort of benefit or contribution to their bottom line so let that sink in for a minute close to 90% of companies say that they will be incorporating some sort of AI capability or AIML tool in the next year but only 15% are actually seeing any impact so far so why are so many businesses actually experiencing this trough disillusionment this discrepancy between the what could be and what it is today well that's because AI isn't a strategy it is an enabler of your strategy it is this is why it is the third pillar in achieving personalization at scale because without foundational data to understand your customer and without the communication tools to reach them AI is just will be left waiting to see the results and so the most important thing to do is to bring your AI closer to your customer data and by doing so not only can you better understand your customers individually and at scale but also you enable more teams to be your teams to be more efficient and you're able to deliver great experiences through automation and so recapping uh what we all talked about here the three pillars to build personalization at scale number one is contextual data and contextual data this is really about understanding your customers holistically with objective data behavioral data Communications data and conversational data together these data points will create this full context Rich picture of your customer without this personalization Falls flat the second pillar is your Communications tools it's not just what you say but also how you say it and where you say it customers expect you to meet them uh on their preferred channels whether that's email SMS Voice or video um but these preferences very widely by region by use Case by Generation by customer sentiment the key really is delivering the right message at the right time and on the right Channel and lastly trusted AI again AI is not your strategy it is an enabler of your strategy and by uh adopting AI uh with paired with great data and Communications tools AI helps you personalize uh and automat your workf flows uh much more seamlessly so at this point you all must be thinking okay Asha this is great but how do you bring this all to life well I'll walk you through just a quick one if you're do-it-yourself person how can you utilize toio and AWS together to actually build take these Lego blocks and build a personalization at scale architecture so if you look at everything on the left hand side here uh these are all your different data sources as well as Channel sources such as email SMS messaging data warehouses Cloud applications you pump all your data into toio segment where by using Tulio segment you're ensuring uh the privacy integrity and compliance of your customer data twio segment supports uh activating this data to over 450 plus destinations similarly uh in inclusive of our our AWS uh AWS Services as well so whether you use Sage maker um for predictions or personalize for recommendations or you know good old S3 for cloud storage so you can further process that data in cascading uh systems alternatively if you don't want to if you don't want to use the individual Lego blocks and build an architecture across twilio and uh AWS twilio AI twio has built a set of capabilities uh that allow you uh to essentially uh not worry about the undifferentiated heavy lifting of putting that architecture together we provide AI powered instrumentation with co-pilots AI powered uh identity resolution um advanced predictions and recommendations as well as automation of your workflows and there are a number of associated AWS products that we build on top of that allows this as well but now more importantly let's see it in action and so I'll call sudir to the stage uh to provide a a uh a demo for us thank you all right uh all right thank you Asha so my name is sud Chaka I'm a Solutions architect director so let's walk through a a quick uh customers uh retail journey and see how you can Elevate their customer experience through personalization and bringing the right right contextual context here so here um you our customer Shan is actually navigating a retail website here and based on the session here on the left side as you can see you know it's an anonymous data right now because that means he's visiting this website for the first time so it's a you know it's an anonymous session we don't have much details about Shane other than that he's browsing through all these websites and while he's doing that we are able to collect all this information like you know what are the products he's viewing you know things of that nature and let's say he likes this particular shoe he comes in here unfortunately this shoe is out of stock no worries what we're going to do is you know Shane is going to sign up for a notification here and while he's doing that he's going to provide us some information including this identifying information email so when he submits this form what we're going to do is we actually going to segment and see if we've built this relationship with Shane in the past turns out we did right so so the segment's uh identity resolution what it does is it takes the existing profile information that's because they've done business at a physical store in the past but he's coming to the website for the first time so we were able to merge both the profiles the anonymous profile and then the uh the existing profile be able to you know build all of that using our identity resolution so what that means is when um Shane starts a a web conversation a web chat the virtual agent now has some information about Shan right it's both it's the current information of what they've browsed what they've done so far and also their past history their habits their you know style information or style preferences things of that nature so so I'm going to just zoom in a little bit here and so let's say you know Shane asked for hey give me some U recommendations on my shopping experience so the virtual agent has access to all this information so it's able to provide you very you know fine-tuned personalized recommendations at this point in real time so let's say you know Shane clicks on one of these products he you know he looks at this Allstar High shoes and he says hey you know what I need to talk to someone um about these particular shoes so the virtual agent what it does is it takes all this information and immediately transfers uh Shen to a live agent so let's switch gears a little bit here and I'm going to go into uh the V the live agent view here so as you can see the Shane's request came into the live agent and when the live agent selects Shan it's going to present a dashboard of Shane here and this is all powered by 12o segment all the information about you know their customer preferences their customer preferences their past history uh you know uh some of the the customer insights which segment was able to build based on that previous information so all that information including some of these AI recommendations are now available to the live agent in addition to that the live agent also has access to some of the historical information you know what the products they've used or also maybe previous interactions they've had so with this information in place and once they accept the request here the additional piece of information the live agent has is the actual interaction that Shane had with the Virgil agent so that way it's a seamless transition from a virtual agent to a live agent so that way the live agent can just jump in and provide that personalized human touch without Shane having to repeat all that information there so let's let's pretend that uh you know the the customers already answered their questions and they end the conversation so when they're on the conversation um what we can do is we can take this realtime conversation that happened again this is across the channel so right now I'm showing a a web chat interaction the same thing is similar whether they come in through a voice maybe a a text message things of that it's a very similar experience so we were able to generate the the sentiment analysis of this conversation we were able to Su summarize the whole conversation and all this information gets fed back into the same till year segment profile enriching the profile every time they are having the touch Point uh with the retailer so let's complete this conversation here and then we still need to help Shan to notify when the products are back in stock so I'm going to switch over to uh segment here the dashboard so this is uh linked LinkedIn uh linked audiences um essentially this is think of as a a no code interface for your business users for them to be able to create you know uh segment uh segmented users based on all the traits for this customer profile that we've been building in addition to that all the data business data coming from the associated data warehouse so we just uh announced uh uh uh beta of LinkedIn audiences that works with uh you know AWS uh red shift which is the warehouse that we're going to use here so what link audience is powered by is a technology called Data graph essentially what think of this as a semantic layer which is building that relationship between the customer profile that we've been developing from all these touch points with the business entities in uh in you know data warehouse like uh red shift in this case so as you can see the profile is now connected with all the orders the product information the product inventor so anytime the inventory changes it goes all the way to the profile information and finally once the business users are able to you know filter and figure out you know which users need to be notified on what uh conditions you want to Define some sort of activation on this an example of an activation here is get all the right information you know their details maybe the product that they View and all that information pass it to a web hook where uh uh you want to send a notification on their preferred Channel as you can see U if I switch over back here actually uh we had all the preferences available at the at the profile level so now we can decide what channel we want to send this notification to Shan and in this case Shane preferred an SMS here so what we can do is we can immediately uh send a text message to Shen we're using too the RCs power text messaging so that way we're sending a branded message and uh you know the customer you know gets the trust and gets the relevant context here here it's not just that hey it's available but you're providing all the right context at the right time and on the right channel so that's end of this this particular use case but again this is just a beginning of what kind of a you know customer experiences you can build if you're able to get the right context combined with AI across the channels so let me switch over here so what what you saw so far here uh you know what uh you know what Robin Asha discussed the demo that I showed here it's all powered by uh the partnership between too and AWS we've been a partner with AWS like since 2018 we have like 2500 plus uh customers jointly between fio uh segment and AWS I'm going to pass it back to uh Robin great thank you so much for that great demo great to see all that uh personalization at scale for for personalization for Shane and for for all of all of our customers um so we are going to Pivot and have a little fireside chat here we're just missing the fire um but if you want to actually just Advan one for me I just want to uh reintroduce Matt and monesy um thank you both for joining us today we are excited to think about how you all are thinking about personalization at scale at your companies um and to get us started Matt if you could tell us a little bit about the skim and your role sure thank you thank you for having me um so uh I'm VP of engineering at the skin um skim for those that may not know uh we are a 12-year-old startup we like to think of ourselves as we reach 16 million users across email mobile and podcasts and uh our Flagship product actually is uh the da skim which is an inbox read um which Delights millions of users on a daily basis with our take on on Current Events sports um and news in general um so what I do at the skin is I manage the engineering and data teams and I ensure that we have the platforms the tools uh and the ability to um to reach our users and not just reach them but ensure that we're able to Delight them uh and to imbue sort of personalization into the experience end to end um and then collect that data to provide insights that's great we're going to come back to that Delight uh in just a bit here uh Mony how about you um you're obviously with AWS we're all here today we know a lot about AWS uh Cheryl with a little about your role yeah Robin um so I manage a team of Partners uh mainly in the business application space so twio is one of our uh you know partners for the past six seven years like Sadir mentioned and outside of that right like with six seven other partners uh we sit together and figure out how do we build better products and work backwards from our customer needs for example right like how do we meld like red shift and segment and how do we bring it to life uh as either a co-build motion or right like just selling together uh that's my that's what my team focuses on great yeah awesome thank you um well it sounds like we certainly have the right people to talk about personalization uh here today um so Matt you know you mentioned um you know personalization and delighting your customers you obviously have you shared with us the expansive ecosystem of over 16 million subscribers through newsletters podcasts a mobile application skim lab your in-house creative agency can you share with us how the skim approaches personalization and customer engagement sure so um so we actually have uh embarked or we started embarking on a project about a little over a year ago which really was a top down sort of evaluation of our data warehouse and ecosystem and infrastructure um and as part of that our goal was to understand the data that we have ensure that it's in the right sort of the right form factor for us to be able to actually leverage it um and started really more and more using toio segment as our sort of Hub our centerpiece um for how we were approaching personalization and um ultimately started building out personalization features that we layer into our products we spend a lot of time focusing on our email products um and with email given the nature of the platform there's a lot of work that we need to do upfront um to send out an email to 5 million 6 million 7 million people on a daily basis so any sort of personalization that we put in place we knew that we were limited when it came to email and doing something that's more real time so a lot of data pre-processing Upstream a lot of enhancing our pipelines from a Content production and data inest standpoint to along the way enrich um our data um and really build features to Delight our users speaking of building features I mean can you can you share with uh some of those critical touch points like where do you find in that customer Journey are those delightful experiences materializing where you see personalization having the greatest impact so focusing on our Del skim Flagship product um what we've been able to do is we've been able to take an experience that really was the same for all of our recipients and layer in personalized features based upon like let's say location providing a uh local weather forecast sort of in the header um so that people know and can prepare for their day and sort of tea up um you know what they need to accomplish during the day and understand what the weather is like out uh beyond that actually we've also uh integrated functionality and features that um provide our readers with the ability to enter Journeys and those Journeys actually are sort of an interactive back and forth even though it's email winds up scolling onto the web and between all of those various uh platforms we're able to have users essentially personalize their experience um so they're getting the news that they really want the information that they really want on a daily basis um we've also been able to do things like take engagement data um primarily with an email-based uh with an email-based platform we have you know clicks and opens um so being able to understand not just who is clicking or but really understand what they're clicking on when they're clicking and what we've done is we've taken that and sort of been able to build out things like uh interest graphs that we're able to then feed back into our system and uh provide things like customized links that really are there to Delight they're there to really uh be the most relevant content at the time when it's it's the most important to our readers fantastic time for everyone to subscribe to the skim if you haven't already please um so uh Mony with your role at AWS you obviously get to work with a lots of different customers and you know is what Matt described common um what is the AWS perspective on personalization yeah no great question Robin I think from my vantage point day in and day out talking to customers uh I see three common things that they ask for um one is breaking down data silos because customers use many business applications specifically in the Enterprise business context uh they use uh segment they use uh CRM platform of their choice and then they also have other platforms so they're like how do we connect connect and meld all of this and break down that Silo so what you're doing in the segment context is very very relevant and that's that's one request and then the other request is there's a lot of innovation happening up and down the stack whether it is trinium or whether it is at Sage maker so I think customers are asking us to be able to amalgamate all of that and then give them a solution where they're like hey I'm trying to solve a use case a business use case work backwards from that and help me figure out what are the different different platforms different solutions I can plug in uh to be able to be more effective that's the second piece and then the third piece is uh security right like it's I think there is an element of being overwhelmed by ch right like there are so many options out there they're like hey do I use this do I use that what are the what is the end to end uh option that'll give me a secure option to be able to say hey I'm able to personalize my data I'm able to break down those silos and still make sure that you know I don't have to think about End to End security I think those are the three things that I hear customers trying to solve specifically from a personalization context that's great um maybe a close cousin to security is really privacy uh from a consumer standpoint and so I want to touch on something that I call the personalization and the Privacy Paradox um and so you have on the one hand customers want these delightful experiences um which can only be done through collecting the kind of contextual data you described you know how they're engaging with you know your emails location data um you know but on the other hand uh only 50% 51% of customers trust Brands to keep their personal data secure um and use it responsibly so so consumers want datadriven personalization but don't really want customers to have the data so you know how do you how do how do you grapple with that um I'd love to hear from both of you how you think about data governance um while pursuing uh personalization objectives so I'll start um data governance and security and privacy are always at Forefront for us with any feature that we're building especially the more and more that we sort of become data driven um we think about not only the to security and privacy and ensuring that we're shepher um in all of our platforms uh users data but also we do sort of take that step back given the balance that you called out to ensure that we're thinking about the personalized experiences and how they're going to be perceived by our users how are we ensuring that we're not just making sure that their their data is private and secure but also ensuring that that Comfort level exists to know that we're going to uh be thinking about considerate and being thoughtful about any of the features that we're going to build um so for something like the weather widget you know we give people the option to be able to go and update their location they have you know other options to do things like um to like I was saying before customize their Journeys um so giving users tools and options to be able to communicate back to us and provide feedback is critical yeah um I think uh I'll just add one thing to what uh Matthew said I think industry and vertical context is super critical right like especially regulated Industries customers in healthcare and public sector uh the conversation starter is this data governance right uh so we've been learning a lot from these customers about you know what are the bounds and checks and how to think about this and then applying those learnings across uh other segments so uh that is something that I think in the past two quarters that I've seen uh being emphasized uh that's the only thing that I'll add right like in terms of learning from industries that have been Masters at data governance right like how do we bring in that context into uh day-to-day consumer uh context yeah that's great uh Mony I'd like to get your thoughts a little bit on AI um and the role of AI in personalization so geni is obviously a huge Focus uh for AWS as well as your customers how are you seeing customers integrate AI uh for personalization and customer engagement type use cases yeah great question uh I want to go back to uh you know the slide that Asha shared I think where every company is looking at AI but only 15% see the value um so when we talk to customer I think it is very clear that they're trying to solve business problems right like they have specific use cases that they're trying to solve um so we work backwards from the customer in terms of understanding what is the specific use case you're solving for example right like in a in a business user context if they use a CRM platform uh and they use like a communication platform like Zoom right the context between those two right like how do we create get a SAS data lake so working up and down the stack in terms of understanding hey I use these five business applications uh help me understand how do I create value how do I save time right like I think that's the critical problem that I see people solving right how do I get more efficient and then they do that in multiple ways right like it's either one by reducing cost or two uh by figuring out how to use these various Services either at the chip level or at the inference level Sage maker Bedrock right like we see customers innovating up and down the stack but the the key is understanding what is the business problem they're trying to solve and then plugging in the various Services right like whether it is um you know working with a partner to solve a business use case and then laring on our AWS services like Bedrock or Sage maker I think uh that's how we are solving the problem for customers it's really it's really what Asha said is that AI in it of itself is not a strategy it's an enable eror of the strategy so what are the use cases you're trying to solve and then you all can help you know guide with the right tooling um Matt how are you all approaching AI um as you think about customer engagement how are you incorporating that into your strategy so actually we incorporate in a number of places and uh as masasi was saying we're sort of layering it in to some of our core use cases um that we currently support in that um we're using things like NLP um machine learning um and really using that to better understand our customers to train internal models that um can tell us based upon Trends with engagement and opens and clicks uh make predictions about what people would like uh content recommendations from a gen AI standpoint we're very much sort of in the exploratory phase this point um we want to uh continue to sort of evaluate landscape um one of the tools that we're looking to use is Party Rock which we find that from like an AI sandbox standpoint will be very helpful for us um there's a lot of opportunity there for us obviously we're using AI as I was saying that's done gen AI um in a number of places but we're still sort of formulating our strategy around a geni sort of approach and where that best fits in for us okay that's great I think you're a good company there um well I just want to thank you both for being here this was this was great um want to thank you all for joining us today we'll hang out here the speakers will hang out for just a couple minutes after this but please do come visit us at our booth uh we'll have the demos that you saw some additional demos that we'd love to to show you so hope to see you there and enjoy the rest of the show thank you thank you thank you thank you both thank you thank you

2024-12-11 03:07

Other news