Privacy-Centric Identity Resolution for Analytics and Activation

Summary

A single customer often appears as five different records across fragmented data sources such as email addresses, loyalty IDs, and device identifiers, making unified customer analytics unreliable without identity resolution.
Databricks Clean Rooms enable organizations to resolve customer identities across partner datasets without exposing raw PII, creating a governed and privacy-safe unified customer view at scale.
A joint LiveRamp and Databricks implementation demonstrates how Clean Room-based identity resolution can standardize fragmented data and increase audience reach—with one insurance company case study showing an 18% improvement.

Privacy-Centric Identity Resolution for Analytics and Activation

Identity resolution sits at the core of modern data and AI. As signals disappear and privacy constraints tighten, the ability to unify fragmented customer records across partners directly impacts data quality, analytics accuracy, and model performance. In this video, we show how Databricks Clean Rooms enable organizations to build a trusted view of customer data by resolving identities across datasets without exposing raw PII. Join us for a technical deep dive with LiveRamp as we demonstrate how a durable identity backbone supports accurate, privacy-safe matching across organizations. You will learn how to standardize fragmented data into interoperable models, improve match rates, and unlock higher-fidelity analytics through governed collaboration using Clean Rooms. We will also cover practical implementation patterns, including enforcing privacy controls and scaling compliant collaboration across multiple partners. The result is a foundation for identity-first analytics that improves your marketing outcomes while maintaining strong privacy standards.
Talk By: Lisa Cramer, Director of Product, LiveRamp ; Nikhil Gaekwad, Product Lead, Databricks ;

Chapters

FAQs

What is identity resolution and why does it matter for analytics?

Identity resolution is the process of linking fragmented customer signals—such as email addresses, loyalty IDs, and device identifiers—into a single unified customer profile. Without it, a single loyal customer can appear as five different records in your data, leading to inaccurate targeting, flawed analytics, and poor model performance.

How do Databricks Clean Rooms protect customer privacy during identity matching?

Databricks Clean Rooms allow organizations to run identity resolution logic across partner datasets without either party exposing raw PII. The ID provider brings matching logic into the clean room environment without revealing underlying data, while output governance controls such as aggregation-only columns and row limits protect each party's information.

What role does LiveRamp play in identity resolution on Databricks?

LiveRamp provides a durable identity backbone that standardizes fragmented signals into interoperable identity models within the Databricks Clean Room framework. The integration enables organizations to improve match rates and unlock higher-fidelity audience analytics through governed cross-organization collaboration.

What business outcomes can identity resolution on Clean Rooms deliver?

This video presents a case study in which an insurance company achieved an 18% increase in audience reach by resolving identities across partner datasets using Databricks Clean Rooms. More broadly, improved identity resolution leads to better-targeted marketing campaigns, more accurate analytics, and higher-performance machine learning models.

Full transcript

[00:08] Well, thank you everyone for joining us. Welcome to day two of the data summit. Hope everyone's having a great summit so far. Uh I'm Nikil God. I lead product management for data ecosystem at data bricks. And in this session, we're going
[00:21] to talk about privacycentric identity resolution for both analytics and application. We have a pretty exciting agenda to cover today. Uh we're going to talk about why is an air resolution
[00:34] important. We're going to look at how it's sort of changing the adtech and market industry. We'll also look at how are partners and customers leveraging clean rooms to be able to do identity resolution in a privacycentric way.
[00:49] We'll also look at some of the best practices and we have a special speaker that's going to join us uh Lisa Kramer uh from Live Ramp to talk about how uh you can use at any resolution through the live ramp ecosystem. And last but
[01:03] not least uh we're going to go through some use cases and also show you a demo of how how this looks like within uh the data bricks ecosystem. All right, before I get started, uh I
[01:15] just wanted to do a quick sort of exercise and everyone just sort of think about this uh as I explain it. So picture your best sort of customer for your industry or your product. Imagine
[01:28] that customer that's loyal, that buys your product across all the channels and that's the customer that you want to have as your core persona for who you want to target to attract more users and
[01:41] customers. So imagine that customer in your head and now with a quick show of hands, how many of you can provide me a unified view of all the customer attributes in data when I ask you the
[01:55] question, talk to me about your your best customer. Quick show of hands. Maybe one or two, maybe one. Okay. Well, that's sort of the uncomfortable truth,
[02:07] right? So when you think about your sort of your best customer and how that sort of comes together within your data ecosystem, a lot of times your best customer actually looks like five
[02:20] different people in your data. And that's a big challenge for organizations because your customer that you're targeting, especially with your marketing campaigns, you want to have a
[02:32] unified profile of what that customer is. But in reality, as you're getting data from various systems, whether it's through an email, a loyalty ID, you know, some sort of device identifier,
[02:44] these are sort of disparit signals that you're getting on the customer itself. And again, to your data when you're analyzing it, these look like five different people. And that's what we're trying to solve today with identity
[02:57] resolution. The goal is that you're able to close this gap and really unify and understand who your customer profiles are. So that way you can have a lot of downstream use cases that you can
[03:09] leverage it for. So with ID resolution, the goal again is you can take all these disparate signals that are sort of disconnected across different systems that you have and
[03:22] bring those together into a single identifier, a single customer profile that you can then use for your downstream use cases. Well, some of you might think, well, Nikil, this is sort of existed for a long time in the
[03:34] industry. Why are we talking about it now? Well, actually there's a lot of things that's sort of happening in the industry that's changing be beneath us. Number one, uh there's definitely sort of third-party signals that are going away that you can you cannot sort of
[03:48] leverage because of things like privacy. Uh we also have organizations that are reluctant to be able to share data because again for regulations. So it's really becoming important for organizations to have a strategy around
[04:01] how can they actually go unify the disparate signals that they're getting for their customers and leverage that you know for the downstream use cases and most importantly there's sort of the three things that are here which is you
[04:14] know data quality is important because that sort of determines the accuracy that you get from your data and now with the agentic world you want to make sure that the models that you're training on is leveraging data that has the greatest
[04:27] source of truth and also has a unified view so you don't have sort of duplications moving forward. And with identity resolution, you're able to have a number of disparate use
[04:39] cases downstream. So once you have that unified customer profile, you can do things like personalization. So if you're personalizing your product for the customers that are again loyal to
[04:51] you that are using the product, you're able to understand more attributes about that individual and then cater your product accordingly. You think about all sort of the flywheel for marketing and
[05:03] adtech. If you're doing things like targeting uh certain types of audiences, having the right identity is really useful. Even if you're doing things like suppression, so you're not targeting that same customer that has already
[05:15] purchased your product, identity resolution becomes really useful. And of course, as you sort of think about being able to train models and uh being able to leverage the latest and greatest that we have with AI, having that data
[05:28] quality with the right customer profiles becomes really important. And as I mentioned, identity resolution is not new. It sort of existed for a number of years. And if you look at how
[05:40] organizations currently do this, and this is sort of the traditional method, is that a lot of organizations ship their data out to the identity partner that's perhaps working with you to
[05:52] provide you the right IDs and that data gets replicated into the identity partners ecosystem. They perform the ID resolution and then ship it back to you. So you essentially have a number of
[06:06] organizations that have datas that are in various silos and it's not happening in in real time and this is where things sort of break down right because data once it leaves your organizational boundaries you're losing control of that
[06:19] and again you're making creating different data silos it's not something that you want to leverage uh when you're kind of doing this in real time and also once you sort of build pipelines in place that allow you to make these data
[06:32] transfers. Again, the traditional way of data sharing. Uh these can be brittle, especially as schemas change behind the scenes. You have to make sure that these are updated and you know, it's hard to maintain. Uh so these are some of the
[06:45] things that can go wrong with sort of the existing workflow for identity resolution. And most importantly, what's sort of come through in the industry more more recently is privacy, right? uh
[06:57] for a lot of organizations they cannot reveal the underlying data that they have sensitive information that they have about their customers. They want to be able to protect that data again
[07:09] respect the the different privacy laws that they have whether it's GDPR or CCPA and ensure that the data remains protected within the organizational boundaries and even for partners that
[07:21] are building identity resolution solutions you'll hear from one very soon it's also important for them to ensure that their customers are protecting their uh first-party data as well they don't want to get access to your
[07:34] sensitive first-party data from your organization. they want to ensure that it's sort of meeting uh the the privacy regulations that you have. That's why I feel like privacy is not just a constraint, but it's also a competitive advantage
[07:46] because you can use that as a way to uh leverage some of the latest technologies to be able to do this in a more privacycentric way. So that way you're not revealing the underlying data and of
[07:58] course making multiple copies and creating data silos. So the next question that comes to mind is great so now we understood what identity resolution is we understood how
[08:10] traditional methods of doing identity resolution happens now it's thinking about okay how do I do this within data bricks and of course let's start with sort of the foundation right all these different customer
[08:23] signals that I talked about whether it's uh transactional data that you're getting in whether it's just your CRM data is now sitting in your lakehouse on data bricks and it's govern so you're
[08:35] using uni unique catalog to govern that data you're leveraging some of the a IML capabilities that data bricks is building so the first step I think most of the people here have complete which is your data already exists in your data
[08:48] bricks lighthouse so the next thing is well now that this data exists how do I actually work with various identity partners to be able to do identity resolution in a privacycentric
[09:02] And for that our answer is data bricks cleaners. A quick show of hands here. How many of you are familiar with data bricks cleaners or leveraging it using it? Excellent. Okay. Wonderful. So clean
[09:15] rooms have been uh GA for uh almost a year and a half now. Uh last data summit we announced GA. And just a quick refresher for those that are not familiar with clean rooms. Clean rooms
[09:27] essentially enable multiple parties to come together and be able to run compute against their data assets without you having to reveal the underlying data.
[09:39] This is a pretty cool technology, right? Where you're able to share data in a governed secure environment where none of the parties have access to it and you're able to run compute that you're mutually agreed. So that way you're
[09:51] protecting your privacy with whether it's your first-party customer data and the collaborator that you're working with is also protecting their data assets and also any sort of logic that
[10:03] they have that you're going to be using uh in the clean room setting. The great thing about clean rooms is that it leverages many of the underlying technologies that we've been innovating within data bricks. So you probably
[10:15] learned about open sharing previously known as delta sharing. So within a clean room, you essentially open share your data assets into the clean room environment. So it stays in place. You're not making multiple copies. And
[10:27] the great thing is is that with data bricks, you're able to connect to data regardless of which cloud or data ecosystem it's sitting in. So if your partner's working on AWS and you're on
[10:39] Snowflake, that's no problem. you're able to leverage database clean rooms to be able to connect your data in place and then run compute on it regardless of which platform you're using. So that's
[10:51] the underlying technology and what we learned when we were working with many of these identity partners is they're actually trying to solve a similar problem, right? They're trying to work with their customers where their
[11:05] customers don't want to no longer want to ship their first-party data to them, but they want to be able to run compute against that data and run logic against it and resolve them into IDs, which are
[11:17] customer profiles that you can then use for downstream use cases. So, we've done a number of innovations, which I'll get to in a moment, for you to be able to do identity resolution directly within data bricks. But before I get to that, let me
[11:31] just quickly walk through what a typical journey looks like for a client and ID partner to do ID resolution uh within the data bricks cleaners ecosystem. So the first thing is that both the ID
[11:44] partner and the client they each both have their data bricks accounts. The first step is one of them will initiate a clean room. So in this example, you'll see that the client invites the ID partner into a clean room setting and
[11:58] they're able to initiate the clean room itself. Once they do that, you use open share to be able to add those data assets or share those data assets into the clean room environment. Again, data
[12:12] stays in place. It's not moving. All you're doing is that you're allowing that clean room environment to be able to run compute against those data assets that you're sharing in. And when you do sharing your data assets, this can
[12:24] either be your tables, views, even a IML models. So, data bricks clean rooms allows you to bring any sort of data assets in. And more recently, we also allowed you to do not just notebooks,
[12:36] but also um jar files, which we'll know about in a second. Cool. Now that the data assets are in the clean room, either partner can uh provide uh some sort of analysis to run against it. In
[12:49] the ID resolution use case, it's usually the ID provider that brings the logic on how they can go from PII to a resolved ID. So the next step is for the client to actually run that particular logic
[13:02] that the ID partner has. So it would sort of start the ID resolution process. And after that's complete, now that the client will receive these resolved ids where they went from PII into these ids
[13:16] without them having to reveal the underlying data. So that's a typical process that you sort of go through end to end to perform identity resolution within data bricks clean rooms.
[13:28] uh when we did launch clean rooms uh into G last year, there's a number of capabilities that our ID partners wanted to sort of streamline this end to end workflow. And what we're announcing today as part of data and AI summit is a
[13:42] number of these capabilities are now in public preview. So I'll walk through each of them in a little bit more detail. Uh the first one is that those that are familiar with clean rooms today, you realize that you can run any
[13:55] sort of analysis in a clean room using notebooks. And what a lot of our partners that use ID resolution or build ID resolution solutions ask for is well I have all my logic in a jar file and I
[14:09] want to be able to leverage that and use that within the cleaner. So what we've done is we've given you two options for you to do that. You can either use a jar task to be able to run that jar file directly within a clean room as a new
[14:23] analysis or you can also delta share or open share uh the underlying jar file into the clean room through a volume and then call that using a UC function. So there's multiple uh options for you to
[14:36] be able to leverage some of your existing proprietary code and logic that you might have within a clean room environment. Great. The second capability is shared output. So those of you that are
[14:50] familiar with clean rooms today, the runner of the clean room whenever they run any sort of compute will receive the output as part of that. What we heard from a lot of our partners is well there
[15:02] are some telemetry events or there are insights that you want to glean as part of the run that both the parties that are in a clean room should be able to receive which is why we built a capability called shared output where
[15:15] you'll actually have a common schema that gets created as part of the clean room setting where both the creator of the clean room and the collaborator will receive it. So in a typical ID resolution use case, you'll have both
[15:28] the ID partner and the client where you would create a shared output that typically has telemetry events that you can leverage uh within your ID resolution process.
[15:41] And another capability that we built was uh a lot of our partners also uh did not want to reveal some of the underlying information about the assets that are adding it that they're adding into a clean room and they also wanted to
[15:54] streamline the endto-end process for how their customer comes in and engages within the clean room itself. Again, those that are familiar with clean rooms, we had what we call an approval clean room model where if you want to
[16:07] run compute against the data that's in a clean room, all the parties had to mutually agree on the clean room itself. Again, this is more of a approval workflow that you needed to go through.
[16:19] But for ID resolution, this is no longer relevant, right? because typically it's a customer that brings their sensitive PII data and you're just running a logic that the ID partner provides and you're
[16:32] receiving an output and then you're doing a lot of downstream use cases with that. So, uh what we built is what we call package cleaner mode. So think about this like almost like an app that you can build without you actually having to go build an app that you can
[16:45] enable for your ID resolution use case and it provides a disparate experience for both the creator and the consumer. So you can really streamline the end to end workflow for uh the end user and
[16:57] you'll see that in a moment when Lisa demos it. And last but not least, as I mentioned, we've been bringing working very closely with a number of ID resolution partners to be able to build uh their ID
[17:10] resolution solutions directly on data bricks. And what we've done is we've actually prepackaged many of this uh clean room capabilities directly within data bricks. And when you go into clean
[17:22] rooms now, you'll see that there's a number of ID resolution partners that we've listed here. and you can instantly create a clean room with these partners and get going with the end-to-end
[17:34] solution. So instead of you having to go manually create a clean room, find the sharing identifier, think about the egress rules and uh policies that you need to align with. Now all you need to do is go find your favorite ID partner,
[17:48] click uh on the actual listing, hit install and the clean room is instantly created and you can sort of go go on about uh performing the ID resolution job.
[18:00] All right. So now we've talked about sort of the new capabilities that we launched for ID resolution and I sort of walked you through the foundation of what is required to perform ID resolution within data bricks. I'm going
[18:13] to now invite up Lisa to essentially give you an end toend workflow of what ID resolution looks like within the context of live ramp. So Lisa, come on stage. >> Thank you.
[18:26] Awesome. Okay. Well, excited to be here and I think Nquille really took us through a lot of the details on how we um work
[18:38] together to deploy an identity resolution capability. Um but I'm going to I'm going to take us back a step uh and talk a bit about um the why. Um so
[19:04] really trying to understand how they're thinking about their data foundation and the good news is as I talk to many in data bricks the data foundation is already set right you have the compute
[19:17] you have the data the governance is in place, but predictive analytics remains out of reach. And oftent times there are kind of like three primary reasons that
[19:30] continues to be the case um even today with all of the advancements we've made. So the first um is what we've termed the identity tax. Um, oftentimes depending
[19:44] on who the partner or or client may be, um, there could be a scenario where folks try and go it alone. They try to use string based matching. Maybe they try and do some probabilistic modeling
[19:58] and you spend cycles and cycles trying to build that out. Um, oftentimes without a reference graph, without a source of truth. um and that becomes uh that that really becomes kind of um
[20:11] effort that doesn't yield significant results. The second challenge that we see in many cases is around the privacy barrier. And this one Nikil already framed up, right?
[20:26] You have all your data already set within data bricks. And historically, you've maybe potentially sent that data out or used an API,
[20:38] but that adds risks, that adds latency, and it doesn't make sense. it breaks the data governance that you've already placed on your customer data. And then the third challenge we see often um is
[20:51] the one that marketers tend to feel the most um and this is what we call kind of the modeling gap. So the organization has made an investment in creating this
[21:03] data foundation but they're really not reaping the benefits or seeing the outcomes that everyone expected. And oftent time oftentimes that's because of uh these fragmented identifiers and not
[21:16] being able uh to connect that data effectively. And this has real results. So we've seen scenarios where 20 to 40% of
[21:28] programmatic spend continues to be wasted, 48% of audiences are mismatched and 15% of brand revenue is lost. So this has significant impacts as you
[21:41] think about uh fragmented identifiers in your customer data. Well, I'm excited that we have partnered with data bricks to bring embedded
[21:53] identity into the packaged clean room. Um and we're going to walk through, as Nquille mentioned, we're going to walk through a demo. I'm going to show you what it looks like um and how it would operate. Uh, and what this means is you
[22:06] now have access to our identity graph on demand. You can run it um with kind of a couple a couple configurations um a couple options of of how to run identity
[22:19] resolution, but you can run it on demand as needed as incremental data comes into your environment. Um, as you want a full refresh of data as as things change over time. um it's available for for access
[22:33] within your own environment. And then the second thing to know about this is as you think about your data foundation uh and use cases um that you want to deploy downstream, this ensures that all
[22:46] of those use cases can be deployed effectively with AI ready features and and that holistic customer profile um that Nquille mentioned earlier.
[22:59] So, we're going to take a look at each of those uh three challenges that I mentioned that continue to persist even with our technology advances. Um, and talk through that a bit.
[23:12] So, we've we've mentioned fragmented identifiers a number of times. Um, and this has been a problem for a long time, right? It it's not a new problem, but it continues to be an ever growing problem,
[23:26] right? Um, new data comes in from customer interactions with brands every single every single day, every single minute. Um particularly as you think about different channels and different
[23:39] ways that data data sources come into your own environment, you could see information from um PII based touch points with rich data across name and postal maybe from an uh an online uh um
[23:55] transaction. You could see something from inapp usage where they're unauthenticated. Um, and all of these have different touch points and different identifiers that at times can be really hard to
[24:09] figure out how to connect. Well, ramp ID um, and live ramp identity is built to help with that particular painoint. We're built with an offline graph at our
[24:23] foundation and that means that we can take any uh fragmented PII touch point and resolve it to the best representation of a person or a household. Uh we can also provide
[24:36] metadata around uh what we tend to call best X for Y, best email for an individual or best address for a household. All of that is kind of in the offline
[24:49] realm in order to give you a holistic view of that customer profile and kill opened with a question around how many of you can connect data to provide kind
[25:01] of insight on all the attributes associated with your consumer. Well, this gets you started with all those PII touch points um being able to be resolved to a single identifier.
[25:14] From there, data is more powerful when it can be connected, and that's what ramp ID is built for. We allow our clients to connect their data into the ecosystem. So, you take
[25:27] your first party view uh and you key it to a ramp ID which is encoded for your brand and then you can translate with any other party uh that is part of the live ramp ecosystem.
[25:40] In fact, we work with over 900 global partners and we've helped them build their data foundation. So, think about all of the walled gardens, all of the
[25:54] major social platforms, um many of the top brands across all of these different industries. As as you think about uh being able to connect with other trusted parties, as you think about activation, as you think about
[26:08] potentially even building out your own RMN or CMN or syndicating your own audiences. We connect across the ecosystem. Okay, so that was the first challenge
[26:22] that was I have problems connecting various fragmented consumer touch points um and you know what I termed earlier as the identity tax. The second challenge
[26:36] we often find around predictive analytics and being able to effectively deploy your data foundation is around the privacy barrier. Well, that is no longer the case. Now that we have
[26:48] embedded identity live within data bricks, you can resolve your data. You can translate your data to connect with other trusted parties. You can even use our data connectors um to bring that
[27:01] data into the live ramp platform for activation downstream. A really seamless process to ensure you can activate data to any destination.
[27:14] Okay, so those were the first two challenges. The third challenge is around the modeling gap or driving outcomes from your data. So I'm going to walk through a couple case studies um of
[27:26] what we've seen from folks in various industries. So the first one is a retailer and this retailer was struggling with inaccurate targeting and really not being able to optimize uh
[27:40] their activation strategy. And so what we did was we worked with them to enable our identity capabilities. And this ensured they
[27:52] could create high precision targeting. They could uh take a look at um at their customer data and start to build out holistic customer profiles that makes it
[28:05] easier to do lookalike and to reach those individuals more accurately. And the results and what they found was a 30% increase in marketing ROI and a
[28:18] 25% increase in customer engagement. So pretty significant results uh from deploying identity. The next one that I'll talk through um
[28:31] is really around um that privacy barrier. So, this insurance company um and this this would apply to many other regulated industries as well. Um this
[28:43] insurance company really wanted to be able to use their data across a variety of different use cases. Um and the way we worked with them was to resolve their
[28:55] data to our pseudonmous identifier, a ramp ID. As I mentioned earlier, ramp ID is encoded per customer. This can also work as you think about um uh use cases
[29:09] within an umbrella brand. So each division has its own ramp ID encoding, but you can translate between them. Well, what this insurance provider found
[29:21] is that by using RAMP ID, um, they were able to deploy a cross-ell and upsell capability. Um, and then they were able to do some advanced modeling against
[29:33] that data as well. And so that ended up in a 15% increase in their total number of quotes and a savings in their display spend as well. all from really
[29:45] understanding identity. Okay. And then the third one um also pretty common for this particular vertical um I think common for hospitality and travel is their customer
[30:00] data was siloed in different systems right so you had booking loyalty information um web visit and app information
[30:12] um and all of it was was uh fairly siloed. Um, additionally they had um, uh, different identifiers per system. So what we did is we worked with them to
[30:25] unify that view um, and we helped with some of their unauthenticated web traffic so that they could understand more about what was happening with those visitors and tie into the other
[30:37] information uh, that they had on their consumers as well. And this allowed them to create a real- time personalization strategy. Um, and it allowed them to start to think about their audiences
[30:50] differently, to start to think about audiences as an asset that could start to be syndicated with other partners um, and start to be used in collaboration use cases.
[31:03] Uh and the results they found specifically from applying identity was an 18% increase in incremental uh reach as well as a conversion lift.
[31:15] Awesome. Okay. So those were the three challenges we were trying to solve for the identity tax, the privacy barrier, and then driving outcomes. And I think um by applying identity, you're able to
[31:28] address all three of those challenges. So, how does this work in practice? Okay, so you've already seen this diagram, but we added uh what some
[31:40] customer data tables might look like um and how this would work with identity resolution itself. So, let's say you have some audience information um here labeled customer data. You have
[31:53] some PII touch points um and some attributes um included as well. you could bring that um into the clean room to resolve it uh and and get that data
[32:07] tied to ramp ID. Additionally, you could bring in add exposure logs. You could have um maybe add exposure logs tied to device identifiers um or add exposure
[32:19] logs already on someone else's ramp ID encoding. You could bring that in for processing within our identity resolution capability and get a full view of what's
[32:31] happening with campaign spend against a particular audience.
[32:43] Okay, so let's take a look at this in action. Okay, great. So in this example, um we're going to be using this uh data table. Um let's call it a loyalty data
[32:56] set. It just has PII and a number of audience attributes. Um all of this is synthetic data. Uh and you can see we have like name, postal, email, um and
[33:08] attribute information. Now, um I'm looking through this and I know it's going to be really hard to figure out how to do string matching uh against this data set. Um but I do think
[33:22] there's an opportunity to optimize what I'm deploying in my activation strategy. So I'm going to take a look at uh the number of records this table has. Okay,
[33:35] so we have a thousand records in in this table. We'll keep that in mind. uh for later. Okay, here is the metadata table. This is how you tell live ramp what to process. Basically, it has some
[33:48] information around the input table, um around the column headers to look at, etc. Here, this is what we talked through earlier. There's a tile available within the clean room page, and you can click
[34:02] on the live ramp uh tile. Um, and then from there, all you have to do is include um a proper clean room name. Um, and then it will get started
[34:15] setting up your clean room. Um, so right now I'm acting as the consumer. I'm going to switch over um to the live ramp side and we're going to see this same named clean room um being
[34:30] set up as well. Awesome. Okay, so back to the consumer side. Um, for demo purposes, I went into one that
[34:42] was already created. Um, and you can see there are two tables in here. The one that we've been talking about. Um, so the PII based loyalty data. Um, and then
[34:54] the metadata table. This is the configuration that tells um the process what to run. Within this, our notebook runs off of some parameters. Um, and the only thing
[35:08] you have to add is that metadata table. Uh, it includes information around which input table to use and it includes a designation for uh where to put the
[35:21] output once it's generated. Um, and the good news is you can reuse that metadata table over and over and over. Um, uh, as long as the schema is the same. If it's not, it's a simple change uh within the
[35:35] configuration there. Okay. Um so while it's running, you can see the progress in the runs tab. Um and then you can click directly on the output schema um to see uh the data set.
[35:52] So in here you can see we have the attributes are still available. Um but then it also includes ramp ID in place of the PII. Um and uh so here we go. Um
[36:08] great. So originally we came into this exercise thinking about how do I optimize my data table? How do I ensure uh my spend in my activation strategy is
[36:22] going to be used effectively? And so now what we're going to look at is across this data set, do we have any duplicates? And 250 rows were duplicates. So that
[36:35] means 25% of my data was duplicated within my data set. As I think about what this means, I want to consolidate uh th those rows, understand more about
[36:48] the attributes for that consumer, and then also tailor my spend so I'm not spending against those duplicated records multiple times.
[37:05] Okay. Um so, as we've talked about, um you know, data bricks creates the data foundation. Uh you're able to organize data there. you're able to query it. Lots of really great analytical tools. You're able to govern data there. It
[37:18] tells you what data you have. But as you start thinking about uh use cases, the key differentiator is being able to know how to connect that data. And that's
[37:32] what identity brings. The ability to connect across the data sets that you have. And so being able to pair that with a data a really strong data foundation within data bricks ensures
[37:44] you can drive outcomes. Um okay and with that I'm going to bring back up. >> All right.
[38:10] for ID resolution within the data bricks platform and ecosystem. And there's sort of two mean two main takeaways that I want people to think about. If you're a customer that has, you know, needs to be
[38:24] able to have your PII data resolved into an ID for you to have effective marketing campaigns, then all you have to do is go find that carousel in the clean rooms page, find your favorite ID
[38:36] partner, and then get started. And this is also going to be included in customer lake, which you probably also heard that we announced. So, a lot of the capabilities that you saw today will also appear on customer lake. So not
[38:49] only just can you resolve IDs, you can also do downstream activation on top of that. uh for partners that already have ID resolution solutions. Uh again, we've
[39:01] worked with a number of them, but if you are a partner that has built a solution where customers can go from PII to an identifier, now you can leverage many of the capabilities that we built, which is the jar task, package clean rooms, and
[39:15] uh even some of the carousel work to be able to build your solution and make it available on data bricks. So you have the end-to-end uh platform ready to go. With that, we'll conclude and thanks so
[39:28] much for attending.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.