<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Ramya M on Medium]]></title>
        <description><![CDATA[Stories by Ramya M on Medium]]></description>
        <link>https://medium.com/@.ramya?source=rss-43e703f09ac6------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>Stories by Ramya M on Medium</title>
            <link>https://medium.com/@.ramya?source=rss-43e703f09ac6------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 18:56:55 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@.ramya/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Data 360 Series | Eight concepts, One company. The complete Data 360 journey.]]></title>
            <link>https://medium.com/@.ramya/data-360-series-one-company-the-complete-data-360-journey-a1732151590f?source=rss-43e703f09ac6------2</link>
            <guid isPermaLink="false">https://medium.com/p/a1732151590f</guid>
            <category><![CDATA[data-cloud-in-salesforce]]></category>
            <category><![CDATA[salesforce-data-360]]></category>
            <category><![CDATA[data-360]]></category>
            <category><![CDATA[salesforce]]></category>
            <category><![CDATA[salesforce-data-cloud]]></category>
            <dc:creator><![CDATA[Ramya M]]></dc:creator>
            <pubDate>Tue, 28 Jul 2026 12:38:52 GMT</pubDate>
            <atom:updated>2026-08-24T15:44:57.212Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Follow Mark, Arcy, and Devn — a marketer, an architect and a developer — as Brew360 Coffee goes from four scattered customer identities to a unified, AI-powered customer experience.</em></p><p>Everything from the previous eight chapters becomes real in this one. What follows is a story that walks through the full Data 360 process — every idea is grounded in one company’s real implementation. Meet <strong>Mark Brown</strong>, Head of Marketing at Brew360 — a specialty coffee brand with 200 stores, an online shop, a loyalty app and a growing wholesale business.</p><h3>Part 1: Mark’s problem — and why it isn’t just Brew360’s (↗<a href="https://medium.com/@.ramya/data-360-series-chapter-1-why-salesforce-data-360-934c3c4d3374">CH1 — Origin</a>)</h3><p>“How many of our Gold loyalty members also shop online?” Mark asked at a Monday data meeting. Three laptops opened. Fifteen minutes later: three different answers. His problem wasn’t a lack of data — Brew360 had plenty. The same customer, <strong>Ben</strong>, existed as four completely separate strangers across four different systems, with no shared memory of who he was.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*R7o-KtuU74RAvt-qvNdJZw.png" /></figure><h3>Part 2: Arcy’s whiteboard — the three layers (↗<a href="https://medium.com/@.ramya/data-360-series-chapter-2-under-the-hood-the-data-360-lakehouse-58e8d62c7153">CH2 — Architecture</a>)</h3><p><strong>Arcy Tara</strong>, Brew360’s Salesforce architect, grabbed a marker. <em>“Think of Data 360 like our roasting process. Green beans come in raw — that’s the DSO, the Bronze layer. They go through the roaster — that’s the DLO, the Silver layer, where we clean and map the data. What comes out is the finishedproduct: the DMO and Unified Profile, the Gold layer.”</em> <a href="https://medium.com/@.ramya/data-360-series-chapter-2-under-the-hood-the-data-360-lakehouse-58e8d62c7153">↗ Medallion architecture deep dive</a></p><p>The key idea Arcy shared: the DLO is the <strong>first place the data is stored permanently</strong>. It lands here in Apache Parquet format, checked against its expected structure. The DMO above it, is not a separate copy — it’s a live view that’s calculated on demand and always reflects the current DLO data. This difference matters a lot for both speed and cost.</p><blockquote><strong>Key insight:</strong> The most common mistake in a first implementation is treating Data 360 like a standard Salesforce object model. The DSO → DLO → DMO → Unified Profile pipeline works in a fundamentally different way. Plan for it from the start, or you’ll be redesigning things six months later.</blockquote><h3>Part 3: Getting data in — and what it costs (↗<a href="https://medium.com/@.ramya/data-360-series-chapter-3-getting-data-in-270-doors-into-data-360-19455a4089dd">CH — 3 Connect &amp; Ingest</a>)</h3><p>“Do we just bring everything in at once?” Mark asked. Arcy shook her head. “Every row costs credits. Streaming costs 2.5 times more than batch, per million rows. So step one: we figure out what the use case actually needs.”</p><p>For Brew360, most sources were batch. Only one use case genuinely needed streaming: detecting abandoned carts in real time.</p><p>Arcy identified six way to bring data in: <br>- Starter Data Bundles for CRM (up and running in two hours).<br>- Ingestion API for the non-Salesforce loyalty app.<br>- MuleSoft for the wholesale ERP system.<br>- BYOL/Zero Copy for three years of Snowflake order history. <br>- Inline transforms to fix email capitalization as it arrives. <br>- Interactions SDK as the only streaming source. <a href="https://medium.com/@.ramya/data-360-series-chapter-3-getting-data-in-270-doors-into-data-360-19455a4089dd">↗ All ingestion methods</a></p><blockquote><strong>The golden rule of ingestion:</strong> Only bring in what your use case actually needs. Every column you don’t use wastes storage. Every inactive customer record that goes through identity resolution uses up credits for no benefit.</blockquote><h3>Part 4: Teaching the systems to speak one language (↗<a href="https://medium.com/@.ramya/data-360-series-chapter-4-from-scattered-fragments-to-one-unified-view-52cfa7155be0">CH 4 — Harmonize &amp; Unify</a>)</h3><p>Arcy printed out a 200-row table — every source field from every system. <br>Column one: what the source system calls it. <br>Column two: what the C360 Data Model calls it. <br>Column three: which DMO it belongs to.</p><p>CONT_ID → ssot__PartyId__c → Party Identification. &quot;So we&#39;re just translating everything?&quot; Mark asked. &quot;Exactly. And once everything speaks the same language, that’s when things start to work.&quot;</p><p>Then came identity resolution — Data 360’s most important feature. Match rules (fuzzy matching on name, exact matching on email, standardized matching on phone) grouped matching records across systems.</p><p>Reconciliation rules decided which values to show on the representative profile, but all four original source records stayed exact as they were, linked to one Unified Individual UUID through identity link objects. Ben’s Key Ring: four source keys, one UUID. No record was overwritten or deleted. <a href="https://medium.com/@.ramya/data-360-series-chapter-4-from-scattered-fragments-to-one-unified-view-52cfa7155be0">↗ Identity resolution explained</a></p><blockquote><strong>Rule on profile imports:</strong> Only bring in active customers. Identity resolution costs more credits than any other Data 360 operation. Running millions of inactive records through it “just in case” is an expensive lesson to learn.</blockquote><h3>Part 5: The moment data becomes action (↗<a href="https://medium.com/@.ramya/data-360-series-chapter-5-data-without-action-is-just-storage-1926c3c8bf53">CH 5 — Segment &amp; Activate</a>)</h3><p>Six weeks in, Ben’s profile was unified. Mark typed out a description of a segment: <em>“Gold loyalty members who bought online in the last 90 days but haven’t visited a store in 60 days.”</em> Three hours later: 8,400 members — people Mark’s team had never been able to identify before. Activated to Agentforce Marketing (formerly Marketing Cloud), the personalized campaign produced a 34% increase in open rates.</p><p>The lesson from Brew360’s first segment publish: segmentation uses more credits than any other operation in Data 360. The default engagement lookback is 2 years — always set the lookback to what your use case actually needs, and spread out the publish schedules to avoid failures caused by running too many at once. <a href="https://medium.com/@.ramya/data-360-series-chapter-5-data-without-action-is-just-storage-1926c3c8bf53">↗ Segmentation &amp; activation guide</a></p><h3>Part 6: Giving the data a brain <strong>(</strong>↗<a href="https://medium.com/@.ramya/data-360-series-chapter-6-ai-models-agentforce-and-the-intelligence-layer-78100e1da246"><strong>CH 6 — The AI Layer</strong></a><strong>)</strong></h3><p>Arcy brought Mark a question: <em>“What if, instead of you deciding which segment Ben belongs to, Data 360 predicted what he was about to do, and acted on it automatically?”.</em> They trained a churn prediction model in Einstein Model Builder, based on Brew360’s unified profiles. Clean unified data went in and intelligent predictions came out.</p><p>The Agentforce integration changed everything. When Ben messaged “I haven’t received my order”, the agent didn’t run a live SQL query across multiple DMOs. It read from a pre-built <strong>Data Graph</strong> — a materialized JSON profile combining Ben’s loyalty tier, order history, past support conversations and churn risk score — stored in the Low Latency Store for millisecond retrieval. <a href="https://medium.com/@.ramya/data-360-series-chapter-6-ai-models-agentforce-and-the-intelligence-layer-78100e1da246">↗ Data Graph &amp; Agentforce</a></p><p>The agent responded: <em>“Hi Ben, I can see you’re a Gold member and this is your first issue with us in 14 months. Let me escalate to priority resolution and add 200 bonus points while we sort it out.”</em> His response: a five-star review.</p><h4>The question Arcy asked next: how “live” does live need to be?</h4><p>A few weeks after launch, Devn noticed a problem. Brew360’s mobile app had a “recently viewed” feature. The plan was for Agentforce agents to see what a customer had just looked at and use that during a conversation. But there was catch: the Data Graph only refreshed every few hours, not every second.</p><p>Here’s what it meant in practice. If Ben looked at espresso machines at lunchtime and messaged support that evening, the agent could see it — enough time had passed for a refresh. But if he looked at espresso machines and messaged supported five minutes later, the agent had no idea. The data hadn’t caught up yet. As Devn put it: “<em>The connections are all there. </em>T<em>he agent just isn’t seeing the newest ones yet.</em>”</p><p>Arcy’s first idea was one a lot of teams try — build a separate, super-fast database just for this one need, and have the agent check the database directly. She started sketching it out, then stopped herself. This approach would mean keeping the same customer information in two different places. Every time something changed in one place, someone have to remember to changes it in the other place too. Over time, that’s an easy way to end up with mismatched, unreliable data.</p><p>So instead, the team used a feature Data 360 already had for exactly this kind of problem: a <strong>Real-Time Data Graph</strong>. Instead of rebuilding everything, they simply told Data 360 which specific pieces of information needed to update instantly. In this case, recent browsing activity and cart activity. Those pieces now update the moment they happen, sent in through something called the <strong>Interactions SDK</strong>, which is just a tool that feeds live app activity into Data 360 as it occurs. Everything else about Ben — his loyalty tier, his order history, his past support cases — didn’t need second-by-second updates, so it stayed on the normal refresh schedule.</p><p>The lesson here isn’t “make everything update instantly”. It’s this: usually, only a smart of a customer’s profile actually needs to be that fresh. Data 360 already has a built-in wat to make just that small part update instantly, without needing a second, separate database to maintain.</p><blockquote><strong>Takeaway</strong>: When someone says “the agent needs live data, “it usually means only part of the answer needs to be live — not the whole profile. Before building a separate fast database for this, check whether turning on a Real-Time Data Graph for just the fast-changing fields solves the problem. See <a href="https://medium.com/@.ramya/data-360-series-chapter-7-apis-sdks-and-building-on-data-360-aa5d12dd40ec">↗ Data Graph vs Real-time Data Graph</a> for how this choice affects both speed and cost.</blockquote><h3>Part 7: Devn discovers the two-step auth (↗<a href="https://medium.com/@.ramya/data-360-series-chapter-7-apis-sdks-and-building-on-data-360-aa5d12dd40ec">CH 7 — Developer Toolkit</a>)</h3><p><strong>Devn</strong>, Brew360’s developer, had his standard Salesforce access token ready. He made his first Data 360 API call. Error 401. After twenty minutes, he found the answer in the documentation. Data 360 has its own tenant and its own access token.</p><p>You login to Salesforce first, then exchange that token for a Data 360 token plus a DATA_CLOUD_INSTANCE_URL. Every API call after that goes to that specific URL — not the standard Salesforce endpoint. Two tokens needed, every time. Once you know it, you don’t forget it again.</p><p>The four OAuth scopes Dev needed: cdp_ingest_api, cdp_query_api, cdp_profile_api and refresh_token. He tested both auth steps using the official Salesforce Data 360 Postman collection before writing any production code.</p><h3>Part 8: What Brew360 learned — and what every architect should know (↗<a href="https://medium.com/@.ramya/data-360-series-chapter-8-governance-credits-and-implementing-like-a-pro-1dcd6bcd3cf7">CH 8 — The Architect’s Mindset</a>)</h3><p>Six months in, Mark presented the results: email open rates up 34%, in-store visits from online-only customers up 19%, Ben’s segment alone generated £280,000 in extra revenue, and the Agentforce support agent resolved 62% of tickets without needing a human to step in.</p><p>Arcy’s summary wasn’t about features. It was about three decisions made early on, that made everything else possible:</p><h4>1. Use case first, not data first</h4><p>Starting with “what business result do we want in 90 days?” determined exactly which sources, fields and identifiers mattered. Everything else waited.</p><h4><strong>2. Design for credit cost</strong></h4><p>Knowing that identity resolution costs the most and that streaming costs 2.5 times more than batch, shaped every architectural decision they made.</p><h4>3. Ship one use case fast</h4><p>A small, well-defined first use case delivered in 8 weeks built more momentum inside the company than any large, sprawling project could have.</p><h3>The epilogue</h3><p><em>Mark got a notification. Ben had opened Brew360’s app for the third time that week. An Agentforce agent had already sent him a personalized recommendation for the new single-origin Colombian blend — because his unified profile showed he’d rated a similar one, five stars six months ago. Mark hadn’t done anything. No campaign, no segment refresh, no meeting. The system had learned on its own. The data was doing the work. He smiled and made himself a coffee.</em></p><p>Six months later, Brew360 had three specialized agents running — one for loyalty queries, one for order management, one for product recommendations — all coordinated by a single orchestrator agent using Agentforce’s Multi-Agent Orchestration framework. Each agent read from the same unified Data 360 profiles, each followed the same consent rules and each was only as intelligent as the data behind it. The same data model Mark’s team had carefully designed in week one was now powering an entire team of AI agents. Arcy’s summary for the board: “<em>The data architecture is the agent architecture. Get one right and you get both.</em>”</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a1732151590f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data 360 Series | Chapter 8 — Governance, credits, and implementing like a pro]]></title>
            <link>https://medium.com/@.ramya/data-360-series-chapter-8-governance-credits-and-implementing-like-a-pro-1dcd6bcd3cf7?source=rss-43e703f09ac6------2</link>
            <guid isPermaLink="false">https://medium.com/p/1dcd6bcd3cf7</guid>
            <category><![CDATA[data-360]]></category>
            <category><![CDATA[data-governance]]></category>
            <category><![CDATA[salesforce-data-cloud]]></category>
            <dc:creator><![CDATA[Ramya M]]></dc:creator>
            <pubDate>Tue, 30 Jun 2026 15:27:48 GMT</pubDate>
            <atom:updated>2026-08-24T10:55:15.444Z</atom:updated>
            <content:encoded><![CDATA[<h3>Data 360 Series | Chapter 8 — Governance, credits and implementing like a pro</h3><p><em>“If you think good architecture is expensive, try bad architecture”. The decisions you make in first week shape your implementation for years, and governance is not something you add at the end.</em></p><p>We’ve covered the platform itself. Now let’s talk about how to implement it well. The architects who succeed with Data 360 share a common set of habits — none of them are specific to this platform. They’re about discipline, not features.</p><h3>Start with use cases, not data</h3><p>The most common mistake in an implementation is starting with “let’s get all our data into Data 360”. That’s the wrong order. The right first question is - <strong>What business outcome are we trying to deliver?</strong> Only after answering that, should you work backwards to figure out which data sources, fields, and identifiers you actually need.</p><p>A small, well-defined first use case delivered in 8 weeks builds more momentum inside a company than a huge data lake project that takes 18 months and still can’t answer a simple question like “who are our top customers?”</p><h3>Understanding the credit model</h3><p>Data 360 charges based on usage. Every operation uses up credits. Understanding what drives that cost helps you make architecture decisions that makes sense both technically and financially:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*iZokT2q5NknDasoC" /></figure><blockquote><strong>Digital Wallet — monitor consumption in near real time: </strong>The <strong>Consumption Cards</strong> tab in the Data 360 app gives admins near real-time visibility into credit usage. You can see usage trends, monitor shared capacity across contracts and detect unusual spikes in usage before they turn into billing surprises.</blockquote><blockquote>Build a regular review of the Digital Wallet into how the team operates — it’s the fastest way to catch a misconfigured segment publish schedule or a streaming activation that’s firing more often than it should.</blockquote><h3>The 8-step implementation framework</h3><p>This framework, derived from Salesforce’s own implementation guidance, gives you a proven sequence for any Data 360 project:</p><h4>1. Discovery &amp; prioritization</h4><p>Define use cases, identify stakeholders, prioritize by business value and how fast each can be delivered.</p><h4>2. High-level architecture</h4><p>Define source systems, data flows, org structure (Home vs Companion) and how Data Spaces should be organized.</p><h4>3. Field-level data mapping</h4><p>Map every source field to its target in the C360 DMO. Write down the primary keys, data categories and field types.</p><h4>4. Visualization planning</h4><p>Decide what dashboards, reports and BI connections (Tableau, Power BI) the team will need.</p><h4>5. Data ingestion &amp; config</h4><p>Setup connectors, create data streams, configure transformations and validate that data is landing correctly in the DLO.</p><h4>6. Activation &amp; orchestration</h4><p>Configure the first segments, activation targets, flows and data actions needed for the first use case.</p><h4>7. Test, remediate &amp; enable</h4><p>Check that unified profiles are correct, test segment member counts, review credit consumption and train the people who’ll use it.</p><h4>8. Launch &amp; iterate</h4><p>Go live with the first use case. Measure the results. Plan the next round of improvements based on what you learned.</p><blockquote><strong>The architect’s north star:</strong> Data 360 architects need to think across solution design, technical setup, data structure, enterprise and integration — all at the same time.</blockquote><blockquote>That breadth is intentional, not a flaw. It means a single well-architected Data 360 implementation can remove the need for four or five separate platforms. Design for the whole system, not just one piece of it.</blockquote><h3>Governance — not an afterthought</h3><p>Data 360’s governance framework is built directly into the platform, not added on top later. Understanding it, changes how you design Data Spaces, permission sets and data classifications from the very first day.</p><p>The authorization model is called <strong>ABAC — Attribute-Based Access Control</strong>. If you’re used to traditional role-based access (where “managers can see everything, analysts can see summaries”), ABAC works differently and in more detail: access decisions can be based on attributes of the user, the data and the situation, all at the same time.</p><p>For example: “field reps can see customer email addresses, but only for customers in their own territory and only during business hours”. This is built using the <strong>CEDAR policy language</strong> — an open-source language for writing access policies, originally created by AWS and adopted by Salesforce because it’s clear and easy to audit.</p><p>Governance is enforced at four levels:</p><ul><li><strong>Object level</strong> — controls access to entire DLOs or DMOs, based on classification tags</li><li><strong>Field level</strong> — restricts access to specific columns within an object (e.g. PII fields visible only to a compliance team)</li><li><strong>Row level</strong> — filters data so users only see the rows relevant to them (e.g. only show customers in their own region)</li><li><strong>Dynamic data masking</strong> — hides sensitive values at the moment they’re accessed, without changing the underlying data, applied consistently across queries, AI prompts, CRM enrichments, and activations</li></ul><p>Data 360 also supports <strong>automated classification </strong>— LLM and ML models automatically identify sensitive data categories (personal information, health information, financial data) across both structured DLOs and unstructured content. T<strong>ag propagation</strong> follows the data as it moves — a field tagged as PII. Email in a source DLO keeps that tag automatically as it flows into derived DLOs and DMOs.</p><p>For encryption, <strong>Bring Your Own Key (BYOK)</strong> lets you upload your own encryption key directly, giving you full control over your data security beyond what the platform provide by default. This maters especially for financial services, healthcare and public sector organizations that have strict rules about who controls encryption keys. You can set this up under: Setup → Platform Encryption.</p><h4><strong>Currency reporting for Data 360</strong></h4><p>Data 360 now supports accurate multi-currency reporting. Financial data (revenue, deal values, costs) can be shown in a user’s local currency using Salesforce’s standard currency conversion tables, so dashboards and calculated insights show consistent, currency-aware numbers across global organizations. Before this, multi-currency setups needed manual workarounds.</p><h4><strong>Playground Data Space — safe testing with real data:</strong></h4><p>A <strong>Playground Data Space</strong> is an exact copy of Default Data Space, created inside the same org. It lets you test schema changes, new identity resolution rules or mapping updates against real production data — but safely, inside a separate, isolated Data Space. Once you’ve checked the changes work, you can copy and merge them back into the Default Data Space.</p><p>This is the recommended way to test in a production org: never test a major configuration change directly in your Default Data Space. Playground Data Spaces close the gap between sandbox environments (which don’t have production data) and the risk of testing directly in production.</p><h4><strong>Governance design principle</strong></h4><p>Data Spaces are your broad governance boundary — use them to separate brands, regions or business units. ABAC policies then provide detailed control within a Data Space. Design your Data Space structure first, because every governance decision after that builds on top of it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1dcd6bcd3cf7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data 360 Series | Chapter 7 — APIs, SDKs and building on Data 360]]></title>
            <link>https://medium.com/@.ramya/data-360-series-chapter-7-apis-sdks-and-building-on-data-360-aa5d12dd40ec?source=rss-43e703f09ac6------2</link>
            <guid isPermaLink="false">https://medium.com/p/aa5d12dd40ec</guid>
            <category><![CDATA[development]]></category>
            <category><![CDATA[api]]></category>
            <category><![CDATA[salesforce-data-cloud]]></category>
            <category><![CDATA[salesforce-data-360]]></category>
            <category><![CDATA[data-360]]></category>
            <dc:creator><![CDATA[Ramya M]]></dc:creator>
            <pubDate>Sat, 06 Jun 2026 14:33:09 GMT</pubDate>
            <atom:updated>2026-08-24T10:17:36.620Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Building on Data 360 is different from standard Salesforce development in a few important ways: two-step authentication, a separate tenant address and a different packaging system. Here’s what you need to know.</em></p><p>If you’re used to standard Salesforce development, Data 360 will feel familiar in some ways and different in others. The biggest difference is that Data 360 has its own tenant, its own login step and its own query languages. Let’s go through it.</p><h3>The two-step authentication flow</h3><p>This catches almost every developer off guard the first time. To call a Data 360 API, you can’t just authenticate to Salesforce and start making calls. You need two separate tokens — one from Salesforce, one from Data 360:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*qC_qWcKYYW4fZzvX" /></figure><h3>The OAuth scopes you’ll need</h3><p>When creating an external client app, select the right scopes for your use case. Here are the four you’ll most commonly need:</p><ul><li>cdp_query_api — run ANSI SQL queries against Data 360 data</li><li>cdp_ingest_api — send data into Data 360 via the Ingestion API</li><li>cdp_profile_api — read and manage unified profile records</li><li>refresh_token — keep the app alive without re-authenticating</li></ul><h3>Four ways to query Data 360</h3><p>The official Data 360 Query Guide (<em>developer.salesforce.com/docs/data/data-cloud-query-guide</em>) describes four different ways to get data out, each built for a different person and a different purpose:</p><h4>1. Integrated Apps</h4><p>Visual, no-code exploration via <strong>Data Explorer</strong> and <strong>Query Editor</strong> inside the Salesforce interface. Best for admins, analysts and anyone validating data without writing code.</p><h4>2. Object-specific APIs</h4><p>CRM-style object queries using the <strong>Data 360 Connect REST API</strong> or <strong>Data Cloud REST API</strong>. Familiar to Salesforce developers. Good for structured DMO queries that follow relationships between objects.</p><h4>3. Language-specific libraries</h4><p>Connect Data 360 to your application using <strong>JDBC</strong> (Java), <strong>Python Connector</strong> (for data analysis and notebooks) or <strong>Connect API for Apex</strong> (for native Salesforce development).</p><h4>4. SQL Query APIs</h4><p>Full lakehouse-style analytics using the <strong>Data 360 SQL Query API</strong> (REST + Apex). Run ANSI SQL — complex aggregations, window functions and joins across billions of rows. <strong>Note</strong>: Calculated Insights and Streaming Transforms use a <em>different </em>version of SQL.</p><blockquote><strong>Start with Postman:</strong> Salesforce has published an official Data 360 API Postman collection at developer.salesforce.com. It’s the fastest way to check that your authentication is setup correctly and make your first API calls, before writing any production code.</blockquote><h3>Data Graph query vs. Query API — which one an agent actually calls</h3><p>Once the authentication works, the next question is usually: When Agentforce needs a customer’s profile is the middle of a conversation, what is it actually calling? Not a raw SQL join — that’s exactly what the Data Graph exists to designed to avoid (<a href="https://medium.com/@.ramya/data-360-series-chapter-6-ai-models-agentforce-and-the-intelligence-layer-78100e1da246">↗ Data Graphs</a>).</p><p>There are two different paths and knowing which one you’re using matters for both speed and cost:</p><ul><li><strong>Data Graph query</strong> — reads the precomputed JSON data directly from the Low Latency Store via cdp_profile_api. This is what Agentforce actions use behind the scenes for lookups made during a conversation. Every Data Graph query counts as an <em>accelerated data query</em> — a separate cost from normal batch usage.</li><li><strong>Query API</strong> (cdp_query_api) — runs a custom ANSI SQL query directly against the DMOs. Flexible, but this takes seconds rather than not milliseconds. Fine for debugging and analysis, but not for a live agent response.</li></ul><p>Working rule — If Agentforce action needs data during a live conversation, it should come from a Data Graph. If it’s for investigation — “why did this segment return that count” — it’s a Query API call from a notebook or admin tool, not something used in production agent code.</p><p>Cost pitfall: Data Graph queries and the batch refreshes that build them, are billed separately. A Data Graph that refreshes every hour against a large primary DMO can rack up batch transform costs even if nobody queries it that hour. Match refresh frequency to how often the underlying data actually changes, not to how often someone might theoretically ask for it .</p><h3>Developer capabilities changing how you build</h3><p>The recent releases have included several capabilities that significantly change the developer experience on Data 360:</p><h4>MCP Tools — 60+ Tools GA + Tableau MCP: Data 360 as an MCP server</h4><p>Data 360 now offers 60+ <strong>MCP (Model Context Protocol)</strong> tools. Think of MCP as a universal connector for AI— a standard way for AI coding agents to securely access files, databases and APIs without needing a custom integration built for each one.</p><p>With Data 360 as an MCP server, external coding agents (Claude Code, GitHub Copilot, etc.) can reach Salesforce customer data, unified profiles and insights directly from the command line — no screen to open, no manual export.</p><p><strong>Tableau MCP:</strong> Agentforce agents can now query Tableau analytics directly via MCP, with Data 360 as the underlying data foundation and Tableau’s row-level security controlling access.</p><h4>Flow Logs tab: Real-time visibility into flow performance</h4><p>The new Flow Logs tab in the <strong>Automation Lightning app</strong> uses Data 360 to permanently store flow run data.</p><p>From a single dashboard, you can see which flows are running slowly, hitting governor limits or throwing errors — no more digging through debug logs or building custom reports. Turn on persistent logging in Setup, then enable it on each flow you want to track. Especially useful for monitoring Data Cloud-triggered flows at large scale.</p><blockquote><strong>SOQL 12 MB query limit:</strong> In simpler terms, if your query asks for too many rows or columns at once, Data 360 will reject it. The limit is 12 MB of result data. The response will include a queryMore link. Use that to fetch the next page of results. If you&#39;re building scripts or applications that read large Data 360 objects, you need to handle the pagination pattern. Design your queries to select only the fields and row ranges you actually need, and test with realistic amount of data while you’re developing.</blockquote><h4>Data Kits</h4><p>When you’ve built something in Data 360 — DLOs, DMOs, mappings, relationships, transforms, connectors — and you need to move it between environments (from sandbox to production, or to share it with a partner), you use a Data Kit.</p><p>A <strong>Data Kit</strong> is a portable bundle of Data 360 metadata packaged as a reusable template. It’s the main way the Data 360 handles the ALM (Application Lifecycle Management). Without Data Kits, you’d have to recreate data models by hand across every environment — with a real risk of drifting out of sync and human error creeping in.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aa5d12dd40ec" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data 360 Series | Chapter 6 — AI Models, Agentforce and the intelligence layer]]></title>
            <link>https://medium.com/@.ramya/data-360-series-chapter-6-ai-models-agentforce-and-the-intelligence-layer-78100e1da246?source=rss-43e703f09ac6------2</link>
            <guid isPermaLink="false">https://medium.com/p/78100e1da246</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[salesforce-data-cloud]]></category>
            <category><![CDATA[agentforce]]></category>
            <category><![CDATA[ai-model]]></category>
            <category><![CDATA[data-360]]></category>
            <dc:creator><![CDATA[Ramya M]]></dc:creator>
            <pubDate>Wed, 03 Jun 2026 16:53:32 GMT</pubDate>
            <atom:updated>2026-08-24T09:31:29.773Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Data 360 is not just a data platform anymore — it’s the foundation on which autonomous AI agents think, act and interact with your customers.</em></p><p>This chapter is what sets Data 360 apart from any CDP that came before it. The AI layer turns Data 360 from a data management tool into an intelligent, agent-powered customer experience engine. How intelligent the engine is, depends entirely on the quantity of everything built in Chapters 2 through 5.</p><h3>The AI Models tab: the AI command center</h3><p>The AI Models tab (formerly called Einstein Studio) is Data 360’s built-in AI workspace. It’s the one place where you build, connect and manage every AI model in your org — predictive models, generative models and outside models.</p><p>Five kinds of tools live here:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*OopkZ37cM0_CQQ-8" /></figure><p><strong>BYOM vs BYO-LLM</strong> — these are two separate capabilities. BYOM (Bring Your Own Model) connects outside predictive ML models — SageMaker, Vertex AI, Databricks, custom Python endpoints, so they can make predictions using Data 360 data without copying that data anywhere.</p><p>BYO-LLM extends this idea to generative models, letting companies use their own fine-tuned or private LLMs inside Agentforce instead of the models Salesforce-provides.</p><p>Both work on a zero-copy principle: the data stays in Data 360 and the model connects to it.</p><p><strong>Foundation Models tab </strong>— new in the AI Models tab, this is where one connect external LLMs as a source for generative AI. After connecting one (OpenAI, Gemini, Anthropic, Amazon Bedrock), you can set it up and test the model in <strong>Model Playground</strong> before using it in to Prompt Builder or in production.</p><p>This is different from BYOM: Foundation Models are for generative features and grounding prompts, while BYOM is for bringing in your own trained predictive or generative model endpoints.</p><h3>Databricks BYOM — the deep integration</h3><p>We now have dedicated Databricks BYOM support that goes beyond a simple API connection. Data scientists and admins can access harmonized, near real-time Data 360 customer data <em>directly inside Databricks</em> — using Databricks’ full maching learning tools (Feature Store, MLflow, notebooks) to build predictive models.</p><p>Once a model is approved in the Databricks model registry, its predictions flow back into Data 360 to enrich unified profiles, power segmentation and personalize real-time interactions. This connects Databricks’ ML tools and Salesforce’s ability to act on the results, closing a gap that used to require building custom data pipelines.</p><h3>AI in Batch Data Transforms</h3><p>AI models from the AI Models tab can be used <em>directly within Batch Data Transforms</em>. On the transform canvas, add a prediction step, map your input DMO fields (matching the features your model was trained on), and the transform will score each record and write the results to a target Data 360 object either on a schedule or manually.</p><p>This lets you build AI directly into the data preparation stage, before segmentation even runs, instead of treating AI as something added on afterward. Use this to score customers in advance for churn, lifetime value or likelihood to buy, so those scores are ready as profile attributes whenever a segment or agent needs them.</p><h3>Consuming predictions — two paths</h3><p>Once you’ve activated an AI model in the AI Models tab, there are two ways to use its predictions in the Salesforce environment:</p><ul><li><strong>Prediction Jobs</strong>: configure in Setup under the AI Models tab. The job scores records against the DMO on a batch schedule or in streaming mode, where a change to any input field on the DMO automatically triggers a new prediction. Results are saved to a Data 360 object and can be used in segmentation, enrichment and activation.</li><li><strong>Flow Actions</strong>: create an Autolaunched Flow in Flow Builder and call the model as a step in the flow. This gives you the full flexibility of Salesforce automation — trigger a prediction based on any event, schedule or record change, and route the output into any CRM field or downstream process.</li></ul><h3>Data Graphs — the Agentforce memory layer</h3><p>This is the most important AI concept in Data 360 and the one most often overlooked. A <strong>Data Graph</strong> is a pre-built, <em>materialized</em> (physically saved) JSON view of a customer’s profile — combining purchase history, browsing behavior, loyalty data, case history, calculated insights and any other related data into one pre-joined structure.</p><p>Why does this matter for AI? When an Agentforce agent needs to respond to a customer, it can’t afford to run a complex SQL join across multiple DMOs in real time, that would take several seconds. Instead, it reads from the Data Graph, which has already done all the joining and is stored in the Low Latency Store for millisecond retrieval.</p><p>How good your Data Graph is, directly determines how intelligently your agents can respond. A thin, poorly setup Data Graph means a generic, unhelpful agent.</p><h4>Note: “millisecond retrieval” isn’t the same as “real-time.”</h4><p>The Data Graph that serves an Agentforce conversation is a <em>materialized</em> view. It’s calculated ahead of time, on schedule, through batch data transforms, then saved in the Low Latency Store. Its fast to read, but only as fresh as its last refresh. If a customer abandoned a cart ninety seconds ago, that wouldn’t show up in the graph yet.</p><p>For truly instant freshness, where a page view or cart addition shown up on the very next query, Data 360 has a separate mechanism: <strong>Real-Time Data Graphs</strong>, fed by streaming data through the Interactions SDK.</p><p>Three things to know before using it:</p><ul><li>Every streamed event needs an Individual Id attached to it, or it will quietly fall back to the standard (non-real-time) process, a common cause of &quot;why isn&#39;t this showing up live&quot; problems.</li><li>Deletes aren’t supported in real time. New data and updates work, removing data still goes through the standard batch process.</li><li>Record caching for the real-time layer has limits up to 100 million records, kept for up to 180 days. Lower these numbers to reduce cost if you don’t need, every user, cached that long.</li></ul><p>Rule of thumb: use a standard Data Graph when “fast” is good enough. Use a Real-Time Data Graph only when the use case truly fails if the data is even slightly out of date — think live personalization or fraud detection, not a support agent checking a loyalty tier.</p><h3>Unstructured data pipeline — five stages</h3><p>For documents, PDFs, support transcripts, and other unstructured content, Data 360 offers a five-step process that you can configure yourself. Each step can be adjusted or extended using Python Code Extensions:</p><h4>1. Parse</h4><p>Extract raw text and structure from the source file — PDFs, Word documents, audio transcripts, images (via OCR), and video. This is the base that every later step builds on.</p><h4>2. Pre-process</h4><p>Remove noise, standardize language and understand images (OCR and captioning). You can chain several steps together. LLM-based processing is available for custom pre-processing logic.</p><h4>3. Chunk</h4><p>Split the content into meaningful segments for embedding. How you chunk it strongly affects search quality — pieces too large loses precision, pieces too small loses context.</p><h4>4. Post-process</h4><p>Add metadata, group similar content and use advanced techniques like Raptor chunking for building summaries of large sections. LLM-based processing is available here too.</p><h4>5. Embed</h4><p>Turn chunks into vector embeddings. Data 360 provides out-of-the-box embedding models and also supports custom or external models. These power keyword search, vector (meaning-based) search, and hybrid search —acts as a foundation of <strong>RAG (Retrieval-Augmented Generation).</strong> Instead of an AI agent relying solely on what it learning during training, it first searches the embeddings for relevant information, then uses that as grounding in its response. The result is accurate, business-specific answers drawn from actual documents.</p><h3>Agentforce and the real-time layer</h3><p>Agentforce is Salesforce’s autonomous AI agent platform. Data 360 is its <strong>memory and intelligence layer</strong>. When an agent needs context — a customer’s purchase history, churn risk score, loyalty tier, or recent conversation — it reads from Data 360’s real-time layer, which processes web and mobile events in millisecond and keeps the Data Graph up to date.</p><h4><strong>The real-time layer has its own pipeline</strong></h4><p>Ingest events → exact-match identity resolution → calculate streaming insights → check real-time segment membership → trigger real-time actions.</p><p>All of this happens in under a second. How good the result is, depends entirely on the quality of unified profiles, data mappings, and calculated insights built in Chapters 2–5.</p><h3>Einstein Trust Layer — how the governance works</h3><p>Every AI action in Data 360 goes through Einstein Trust Layer, which provides <strong>zero data retention</strong> with external LLMs (prompts are not stored or used to train models), <strong>dynamic data masking</strong> of sensitive fields before prompts are sent out (PII, financial data and health data are hidden before they leave, not afterward), and a full record of every AI interaction.</p><p>For team working in regulated industries, this means you can use any LLM — including external ones — without violating data residency or compliance rules.</p><h3>AI capabilities you need to know</h3><h4>Intelligent Context: Automatic grounding from unstructured business documents</h4><p>Intelligent Context automatically extracts, organizes and surfaces information from the unstructured business data — PDFs, support transcripts, product docs, contracts — so AI agents can deliver accurate, business-specific answers.</p><p>It replaces what used to be complex multi-step process with a low-code setup that can be done in hours. You can also setup context so the same document is interpreted differently depending on who’s asking — a product PDF can show different information to a sales agent than to a support agent.</p><h4>Tableau Semantics: Translating data into business language</h4><p>Tableau Semantics (powered by the Customer 360 Semantic Data model, SDM) lets Data 360 to translate raw data fields into terms and relationships that make sense to a business user. So, when an AI agent or analyst asks “show me top customers by revenue”, the platform understands what “top”, “customer” and “revenue” mean in your specific business.</p><p>It ensures consistent meaning across every query, report, and agent interaction — eliminating the “same question, different answer” problem common in data environments that aren’t well managed.</p><h4>Notebook AI: Your personal workspace for documents + Salesforce data</h4><p>Notebook AI is a workspace inside Data 360 where you can upload business documents, combine them with the specific Salesforce data and use AI to summarize, brainstorm or dig into complex problems.</p><p>Think of it as having a research assistant who has read all the internal documents <em>and</em> has full access to CRM, pipeline and customer data — and can reason across all of it in one conversation. Useful for analysts, strategists and anyone who needs insight from scattered sources without building a formal pipeline.</p><h4>Multi-Agent Orchestration: Agents working together, grounded in shared Data 360 context</h4><p>Summer ’26 Beta feature: connect specialized subagents under one orchestrator agent in Agentforce Builder. Each subagent handles a defined scope (e.g. one for loyalty queries, one for order management, one for product recommendations), while the orchestrator assigns work and coordinates between them.</p><p>This means the Data Graphs and unified profiles now need to serve <em>several </em>agent types at once, the data model quality is now the shared intelligence layer for a whole team of AI agents, not just one.</p><h4>Tableau MCP: Tableau analytics callable by Agentforce agents</h4><p>Tableau’s analytics engine is now accessible to Agentforce agents via MCP (Model Context Protocol), protected by the Agentforce Trust Layer. Agents can query Tableau views and certified datasets as a native tool and the answer comes back governed and with sources cited.</p><p>Data 360 is the data foundation and Tableau is now an intelligence layer on top, the agents can call directly. Row-level security and certified datasets in Tableau act as guardrails for both human analysts and AI agents accessing the same data.</p><p><strong>Agentic Setup and Data Management:</strong> Setup and manage the entire Data 360 data pipeline using <strong>natural language</strong>. Describe what you want to connect, transform or activate, and the agent provides smart suggestions for connector settings, transform logic and mapping configurations based on your intent. It guides you all the way from connecting a source to activating it, in one smooth process, while showing exactly what it’s doing at each step.</p><p>This makes it easier for admins and business users who understand the data problem but aren’t experts in Data 360’s technical setup and it speeds up experienced architects by handling the routine configuration work automatically. Turn it on through <strong>Setup → Setup with Agentforce</strong>.</p><blockquote>Data 360 is no longer just a foundation for human-facing analytics. It’s the foundation for autonomous AI agents that interact with customers 24/7. Every architectural decision made in Chapter 2 through 5 — the quality of data mapping, identity resolution match rules, calculated insight definitions and Data Graph configuration — directly determines how intelligent and useful those agents actually are.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=78100e1da246" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data 360 Series | Chapter 5 — Data without action is just storage]]></title>
            <link>https://medium.com/@.ramya/data-360-series-chapter-5-data-without-action-is-just-storage-1926c3c8bf53?source=rss-43e703f09ac6------2</link>
            <guid isPermaLink="false">https://medium.com/p/1926c3c8bf53</guid>
            <category><![CDATA[salesforce-data-360]]></category>
            <category><![CDATA[segmentation]]></category>
            <category><![CDATA[activation]]></category>
            <category><![CDATA[salesforce-data-cloud]]></category>
            <dc:creator><![CDATA[Ramya M]]></dc:creator>
            <pubDate>Tue, 02 Jun 2026 16:49:32 GMT</pubDate>
            <atom:updated>2026-08-23T13:07:39.402Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Segmentation is where business rules meets data. Activation is where Data 360 reaches out and connects with every channel the customer uses.</em></p><p>We’ve ingested data, harmonized it, and built unified profiles. Now comes the payoff: using that data to create real business results. Segmentation and activation are where Data 360 proves its value.</p><h3>Four segment types — know when to use each</h3><p>The architect guide describes four different segment types. Each one is built for different speed and different use-case:</p><p><strong>STANDARD SEGMENT: Scheduled batch, 12–24 hrs</strong><br>The main segment type. Uses a drag-and-drop filter canvas with AND/OR logic. The standard schedule is every 12–24 hours. Rapid Publish runs every 1–4 hours for more recent engagement data.</p><p><strong>REAL-TIME SEGMENT: On-demand, milliseconds</strong><br>Finishes instantly, based on current profile data and recent activity. Built specifically for immediate personalization. It cannot use exclusion filters or nested segments. It’s designed for speed, not complexity.</p><p><strong>EINSTEIN SEGMENTS: Natural language → filters</strong><br>Describe the audience in plain English. Einstein Segment Creation turns it into to filter logic. This isn’t just converting words to syntax, it’s AI reasoning based on the actual unified Data 360 profile data. The result reflects what’s really in the data, not a generic guess. No SQL needed — built for marketing analysts.</p><p><strong>NESTED &amp; WATERFALL: Reuse &amp; prioritize</strong><br><strong>Nested:</strong> use an existing segment as a filter inside a new one — saves credits and keeps the criteria consistent. <br><strong>Waterfall:</strong> a priority order that makes sure each customer ends up in exactly one audience with no overlap.</p><h3>Calculated Insights: enriching profiles before segmentation</h3><p>Before you build a segment, you often need summary numbers — lifetime value, how often someone buys, days since their last purchase. Calculated Insights (CIs) are SQL-based calculations that run on the DMOs and attach the results back to the Unified Profile. Available in both batch (scheduled) and streaming (near real-time) modes.</p><h3>Activation: where your data reaches the world</h3><p>Activation is the process of publishing segment members, along with their supporting profile attributes, to a configured destination. Data 360 supports three different delivery methods, each suited to a different situation:</p><p><strong>BATCH ACTIVATION: Scheduled, high-volume</strong><br>Data is staged in Secure Internal Buckets (cloud object storage) or sent through Secure File Transfer, then picked up by the target system. Supports incremental refresh to reduce volumes. Best for large-scale email campaigns and refreshing advertising audiences.</p><p><strong>STREAMING ACTIVATION: Near real-time, event-driven</strong><br>Direct API calls sent to the destination as events happen. Best for automation triggers, real-time personalization, and any situation where seconds matter, not hours.</p><p><strong>ACTIVATION-TRIGGERED FLOWS: No-code, platform-native</strong><br>When complete, Data 360 fills an Audience DMO, which triggers a High Scale Salesforce Flow. The Flow uses External Services and MuleSoft connection to reach systems through their APIs. This is the fastest way to add new activation destinations without code.</p><h3>Communication governance — consent at the core</h3><p>Activation is powerful — but it only sends message to customers who have given their consent. Before any data leaves Data 360, it passes through a built-in consent-checking layer that’s always on and needs no custom code. A single, central consent system stores all customer preferences — global opt-outs, channel-specific consent and purpose-specific consent, on the Unified Individual Profile.</p><p>When activation runs, the platform automatically removes anyone who hasn’t given consent, before the data is sent out. It also picks the single best, most compliant contact method for the channel being used. Sensitive fields are automatically masked throughout the process. For team working in regulated industries, this built-in system is how Data 360 supports GDPR, CCPA, and similar rules without needing to write custom code.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*tcBy12EAzIczS7Rx" /></figure><p><strong>Related Lists in CRM:</strong> Data 360 profile information can now be pulled directly into standard Salesforce object pages (Accounts, Contacts, Leads, and more) as Related Lists — without a separate activation step. A sales rep viewing a Contact record can now see that person’s loyalty tier, last purchase date, and churn risk score right on the same page. Set this up through <strong>Setup → Enrichments</strong>. This is activation without a separate activation step — unified data shows up directly inside the CRM screens people already use.</p><p><strong>Test Data Cloud-triggered flows:</strong> One can now create and manage test scenarios for Data Cloud-triggered flows directly in <strong>Flow Builder → View Tests</strong>. Set up specific input conditions and the results you expect, then run the tests whenever you want — no need to manually set up debug settings each time. This matters for teams building reliable automations, where a failed flow means a missed customer interaction.</p><p><strong>Batch DMO activations and more in flows:</strong> Activation-triggered flows can now include <strong>batch DMO activations</strong> directly as steps in the flow, giving a no-code way to trigger batch data movement alongside the other flow logic. Recently support for <strong>more contact points and related DMOs</strong> is added in event-triggered flows, so one can act on more types of profile data in real time. Use the updated <strong>Wait Elements</strong> and improved <strong>Broadcast Flow</strong> features to build more advanced, timing-aware activation flows.</p><blockquote><strong>Credit alert:</strong> Segmentation uses more credits than any other operation in Data 360. The default lookback period for engagement data is 2 years, meaning every publish processes 2 years of events unless you set a filter yourself. Always set the lookback period to what you actually need. And spread out your publish schedules — running too many segments at the same time causes failures.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1926c3c8bf53" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data 360 Series | Chapter 4 — From scattered fragments to one unified view]]></title>
            <link>https://medium.com/@.ramya/data-360-series-chapter-4-from-scattered-fragments-to-one-unified-view-52cfa7155be0?source=rss-43e703f09ac6------2</link>
            <guid isPermaLink="false">https://medium.com/p/52cfa7155be0</guid>
            <category><![CDATA[salesforce-data-360]]></category>
            <category><![CDATA[harmonize]]></category>
            <category><![CDATA[salesforce-data-cloud]]></category>
            <category><![CDATA[data-model]]></category>
            <category><![CDATA[unify]]></category>
            <dc:creator><![CDATA[Ramya M]]></dc:creator>
            <pubDate>Fri, 29 May 2026 16:47:15 GMT</pubDate>
            <atom:updated>2026-08-23T12:40:33.339Z</atom:updated>
            <content:encoded><![CDATA[<p><em>This is where the real work happens. Bringing data together and matching it to the right person is the core skill of Data 360. It’s important to understand that a Unified Profile is not a single “master record” — it works differently, and that difference matters.</em></p><p>Ingestion is about getting the data <em>in. </em>Harmonization is about making that data <em>use the same language.</em> Unification is about recognizing that four different records, sitting in four different systems, might actually be the same person.</p><h3>Harmonization: mapping to the C360 Data Model</h3><p>The Customer 360 Data Model (C360) is Data 360’s shared, built-in data structure — a ready-made model with over over 300 standard objects that works for any industry. When you harmonize data, you’re mapping your source fields to this shared structure.</p><p>This step — mapping DLO fields to DMO fields, is where the team spend most of their time. Get it right early, and everything after it goes smoothly. Get it wrong, and you’ll spend months redoing identity resolution rules and calculated insights.</p><h3>Identity resolution: <strong>recognizing the same person across systems</strong></h3><p>This is the most important feature in Data 360. But first — <strong>a Unified Profile does not replace or merge your original records into one “true” version</strong>. Identity resolution doesn’t pick a winning value or overwrite the source data. Instead, it creates a <em>set of keys</em> that connect the source data — identifying all the records that belong to the same person.</p><p>The original records stays exactly as they are, linked together via Unified Profile’s unique ID (UUID). You then decide which source system’s data to use for each specific business use case.</p><p>The pipeline runs in four stages:</p><ul><li><strong>Candidate selection </strong>— Before comparing any records, the system uses <strong>Blocking Keys</strong> and <strong>Locality Sensitive Hashing (LSH)</strong> to shrink the number of comparisons needed. LSH is a mathematical method that groups records likely to be similar into “buckets”, so instead of comparing every record against every other record (which would take far too long with billions of records), it only compares records within the same bucket. Think of it like sorting mail by postcode before sorting by name.</li><li><strong>Deep matching</strong> — AI models calculate how likely two records are to be the same person, correctly handling misspellings, different formats and small variations.</li><li><strong>Clustering </strong>— Matched records are grouped together, including indirect matches (if record A matches record B, and B matches record C, then all three are linked to the same profile — even if A and C were never directly compared).</li><li><strong>Reconciliation </strong>— Field values from the grouped records are checked against rules you define (Most Recent, Most Frequent, Source Priority) to fill in a representative summary on the Unified Profile. <em>No source data is ever deleted or overwritten.</em></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*gqlIpXlcV9ShkD6U" /></figure><h3>Key Rings: the identity mechanism</h3><p>Behind every unified profile is what’s called the Key Ring mechanism. Every Unified Individual has a UUID (a unique ID that no other record shares). Attached to that UUID is a set of “keys” — one for each source system’s identifier. Picture an actual keyring: keys can be added, removed, or transferred as your data changes over time. All original records stay exactly as they are, linked to the UUID through identity link objects which map every source record’s ID to its matching Unified Profile ID.</p><h3>Near real-time processing and entity types</h3><p>Identity resolution doesn’t just run once — it keeps running continuously. Small batches of changes can be processed <strong>as often as every 15 minutes</strong>, so the Unified Profile always reflects the most current data available for personalization and Agentforce conversations.</p><p>The system supports three types of matching, each for a different use case —</p><p><strong>Individual matching: </strong>linking personal information across systems into one Unified Individual.</p><p><strong>Account matching: </strong>linking company records for B2B use cases, including matching similar company names.</p><p><strong>Household matching: </strong>grouping related Unified Individuals into household.</p><blockquote><strong>Best practice:</strong> Only bring <strong>active </strong>profiles into Data 360. Every record goes through identity resolution, which costs more credits than any other operation. Bringing in 5 million inactive customers just to exclude them later is a costly mistake.</blockquote><blockquote><strong>Expanded copy field enrichments:</strong> Data 360 can now show unified profile information as <strong>copy field enrichments</strong> on any Salesforce object — not just Contact and Lead. Set up enrichments for Accounts, Opportunities, Cases, or custom objects through <strong>Setup → Enrichments</strong>. Sales reps see churn risk scores directly on Opportunity records, and Service agents see loyalty tier directly on Cases — no extra setup step needed. <strong>Two improvements added:</strong> permissions now turn on automatically when you activate an enrichment (no need to assign them manually), and you can now use <strong>multi-dimensional Calculated Insights</strong> as a source for enrichments — before, only single-dimension CIs worked.</blockquote><blockquote><strong>Expanded copy field enrichments:</strong> Data 360 can now show unified profile information as <strong>copy field enrichments</strong> on any Salesforce object — not just Contact and Lead. Set up enrichments for Accounts, Opportunities, Cases, or custom objects through <strong>Setup → Enrichments</strong>. Sales reps see churn risk scores directly on Opportunity records, and Service agents see loyalty tier directly on Cases — no extra setup step needed. <strong>Two improvements added:</strong> permissions now turn on automatically when you activate an enrichment (no need to assign them manually), and you can now use <strong>multi-dimensional Calculated Insights</strong> as a source for enrichments — before, only single-dimension CIs worked.</blockquote><blockquote><strong>A common misunderstanding:</strong> Data 360’s Unified Profile is <em>not</em> a single master record. It does not send reconciled data back to the source systems, merge records together, or overwrite existing values. It’s a set of identity links — a key ring — that lets you <em>choose</em> which source system’s data to use for each situation. If you’re considering Data 360 as a replacement for Master Data Management (MDM), this difference matters a lot.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=52cfa7155be0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data 360 Series | Chapter 3 — Getting data in and shaping it on the way]]></title>
            <link>https://medium.com/@.ramya/data-360-series-chapter-3-getting-data-in-270-doors-into-data-360-19455a4089dd?source=rss-43e703f09ac6------2</link>
            <guid isPermaLink="false">https://medium.com/p/19455a4089dd</guid>
            <category><![CDATA[salesforce-data-cloud]]></category>
            <category><![CDATA[data-connectors]]></category>
            <category><![CDATA[ingestion]]></category>
            <category><![CDATA[salesforce-data-360]]></category>
            <category><![CDATA[data-ingestion]]></category>
            <dc:creator><![CDATA[Ramya M]]></dc:creator>
            <pubDate>Tue, 26 May 2026 18:46:15 GMT</pubDate>
            <atom:updated>2026-08-24T10:19:28.959Z</atom:updated>
            <content:encoded><![CDATA[<p><em>270+ connectors, two way to share data across systems, two types of transformation, and one framework that controls it all. Bringing data in Data 360 is more than just plugging in a source. It’s the first big decision that shapes everything that comes after.</em></p><p>One of the most common questions one ask at the start of a Data 360 project is — “How do we get the data in?” The answer depends on where the data lives, how fast it needs to arrive, and what shape it needs to be in before it’s useful.</p><p>This chapter covers four things: the framework that all connections are built on, the different types of connectors available, the cost decisions around how data arrives, sharing data both in and out without copying it (zero-copy), and the process that shapes data after it arrives.</p><h3>The foundation: Data 360 Connector Framework (DCF)</h3><p>Before looking at individual connectors, it’s worth understanding what they’re all built on. The <strong>Data 360 Connector Framework (DCF)</strong> is the shared system that every connector is built on top of — from CRM connector to custom integrations using the Ingestion API. It gives us a consistent experience for setting up connectors, storing metadata, extracting the data, and delivering it into the Lakehouse.</p><p>It also applies the same enterprise-level security options everywhere, including private links, VPNs and secure tunnels, so the same security controls apply no matter which connector we use. All 270+ connectors are built on DCF, which means one can expect the same behavior across every connection they setup.</p><h3>ELT, not ETL — a deliberate choice</h3><h4><strong>ETL </strong>(Extract, Transform, Load)</h4><p>means clean and reshape data first, then store it — like ironing a shirt before putting it in the wardrobe.</p><h4><strong>ELT </strong>(Extract, Load, Transform)</h4><p>means store it first, shape it later — hang it up now, iron it later when we actually need it.</p><p>Data 360 uses an <strong>ELT-first approach</strong>: get data into the DLO quickly, in its raw form, then transform it afterward. This means data becomes available faster, a failed transformation doesn’t block the data from arriving, and we always keep the original raw data as backup reference. It also makes it easier to change how we transform the data later, without having to pull it again from the source.</p><h3>Batch vs. streaming — the cost decision one must make first</h3><p>Before setting up any connector, answer one question first: <strong>does this use case really need data within seconds, or is minutes-to-hours good enough?</strong> The answer is almost always “minutes-to-hours” — and that matters a lot for cost.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*p-8xMqo7vwQXyKWG" /></figure><h3>The connector landscape — categories</h3><p>The 270+ connectors in Data 360 fall into below meaningful categories. Knowing which category the source falls into shapes the architecture decisions around scheduling, transformation, and cost.</p><p><strong>SALESFORCE NATIVE — CRM, MC, Commerce, Service</strong><br>Deep integration with Salesforce products that needs no setup. <strong>Starter Data Bundles</strong> come with CRM connector already set up meaning — DLO creation, DMO mapping, and key field relationships all happen automatically. This is the fastest way to get the first unified profile.</p><p><strong>DATABASES &amp; DATA WAREHOUSES — Redshift, BigQuery, Snowflake, RDS</strong><br>Connect to relational databases and cloud data warehouses. For platforms like Snowflake and BigQuery, one can choose to either ingest (copy the data into the DLO) or federate (query it where it already sits, without copying). The trade-off is speed vs cost, explained in next session.</p><p><strong>STREAMING &amp; MESSAGING — Kinesis, Ingestion API, SDKs</strong><br>For true real-time data streams. Amazon Kinesis for existing AWS pipelines, Ingestion API for custom REST-based systems (using the cdp_ingest_apiscope), and the Web/Mobile Interactions SDK for tracking website and app behavior.</p><p><strong>EXTERNAL SAAS APPLICATIONS — Marketo, Dynamics 365, Mailchimp</strong><br>Simple, click-based connectors for business apps outside the Salesforce ecosystem. Managed directly in the platform, usually running in batches. Useful for pulling together — marketing, CRM, and engagement data from tools outside Salesforce.</p><p><strong>CLOUD OBJECT STORAGE — Amazon S3, GCS, Azure Blob</strong><br>Bring in structured files (CSV, Parquet, JSON) or unstructured content (documents, PDFs, audio, video) stored in cloud storage. The unstructured path feeds Data 360’s UDLO process for chunking, embedding, and search indexing.</p><p><strong>INTEGRATION PLATFORMS — MuleSoft Anypoint</strong><br>For complex integrations that need routing, orchestration or transformation <em>before</em> the data enters Data 360. MuleSoft connects to hundreds of additional systems and is the right choice when source data needs a lot of reshaping before it lands in the DLO.</p><h3>Zero Copy — works both ways, not just one</h3><p>Zero Copy is one of Data 360’s most important capabilities, and it works in <strong>both directions</strong> — something the standard documentation often doesn’t explain clearly enough. Data 360 can pull <em>in </em>data from external platforms and share data <em>out</em> to them, all without physically copying or moving the data.</p><h3>Bringing data in — two ways</h3><p>When we connect to an external data warehouse or lakehouse, we can choose between two approaches:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YgHZYwk37tQlk0EmKpeHIw.png" /></figure><p><strong>Query federation</strong> uses JDBC and smart SQL translation — Data 360 rewrites the query into the source system’s SQL language, lets the external system do the actual processing, and only returns the results. An optional caching layer (query acceleration) keeps a near-real-time local copy so repeat queries are faster.</p><p><strong>File federation</strong> works differently. The source needs to expose an Iceberg REST Catalog, and Data 360’s own engine reads the underlying Parquet files directly. There’s no translation and no sending work to another system — just direct file access. This is simpler and more efficient for Iceberg-native platforms like BigQuery and Databricks.</p><h3>Sharing out — Data 360 as a data source</h3><p>Zero Copy also works the other way around. External platforms can query Data 360’s data via an exposed <strong>JDBC driver</strong> for live, on-demand access, or use the <strong>DaaS (Data as a Service) API</strong> to access file containing DLOs, DMOs and Calculated Insight Objects. The DaaS API provides access to both metadata and underlying data tables, while keeping Data 360’s governance rules in place — meaning the system reading the data sees trusted, controlled data, not raw files.</p><blockquote><strong>Decision point:</strong> Ingest or federate? Ingesting copies data into Data 360’s Lakehouse — queries run faster, full identity resolution is supported, and it works even if the source system is temporarily offline. Federating leaves the data, where it is cheaper, always up to date, but depends on the source availability and can be slower to query. For identity resolution and segmentation, you need to ingest. For extra analytics and enrichment, federation is often the better choice.</blockquote><h3>After landing: the data transformation pipeline</h3><p>Ingestion brings data into the DLO. Transformation shapes it so it’s ready to map to a DMO and becomes part of the unified profile. Data 360 offers three types of transformation, applied at different points along the way:</p><ol><li><strong>Inline transforms — at ingestion time</strong><br>Applied before the data even reaches the DLO. Simple, formula-based fixes: PROPER() — fixes capitalization, TRIM() — removes extra spaces, LOWER() and string concatenation. This catches the most common data problems — inconsistent capitalization, extra spaces, name fields split incorrectly — as early as possible. If you can fix it here, fix it here.</li><li><strong>Batch Data Transforms — scheduled, large-scale reshaping</strong><br>Set up using a visual, low-code pipeline builder. Designed for more complex structural changes: <strong>pivoting</strong> (turning one wide row with many columns into several narrow rows — useful for cleaning up produce line data), <strong>flattening</strong> (restructuring nested JSON into tabular form), joins and aggregations. Runs on a schedule against source DLOs and writes the results to new DLOs. Supports full sync and incremental modes — always choose incremental whenever possible. Batch Data Transforms also support <strong>custom code through Code Extensions</strong> — so complex logic that can’t be done visually, can be written in Python and run as part of the same managed pipeline.</li><li><strong>Streaming Data Transforms — continuous, near real-time</strong><br>Written as a SQL SELECT query that runs continuously against the incoming record changes. Supports cleaning and standardizing data (checking personal information, normalizing formats), enriching it through joins, and importantly <strong>Streaming Lookup JOINs</strong> against reference data that rarely changes for instant profile updates. Behind the scenes, this uses a High-Density job design that packs several streaming transforms for one customer into a single compute job, keeping the cost efficient even at large scale. This is the right choice for monitoring events, updating profile immediately, and real-time scoring.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*F0ypd-CrN3fQ40HhUfk91Q.png" /></figure><h3>Ingesting unstructured data — a dedicated path</h3><p>Beyond structured records, Data 360 has a separate path for bringing in unstructured content — PDFs, Word documents, sales call transcripts, knowledge articles, support chat logs, and any text-heavy files that don’t fit into a normal relational DLO. This content feeds directly into the vector search process that powers Agentforce’s grounded, fact-based answers (covered in depth in Chapter 6).</p><p>To bring in unstructured data: connect from <strong>Amazon S3, Azure Blob Storage, or Google Cloud Platform</strong> (or use any existing Data 360 connector) — this creates an <strong>Unstructured Data Lake Object (UDLO)</strong>. Or, set up a data stream directly from any connector to bring in unstructured files continuously. Once the data lands in the UDLO, we create a <strong>search index configuration</strong> — this decides how the content is split into chunks and turned into searchable vectors, producing the objects needed for semantic (meaning-based) search in Prompt Builder, Agentforce, and Tableau. A few real examples: improving service reply suggestions by grounding them in knowledge articles, writing personalized emails by referencing a customer’s past emails, and showing agents similar resolved cases so they can solve new ones faster.</p><blockquote><strong>Problem Records DLO:</strong> When a record fails to pass ingestion checks (wrong data type, missing required field, format mismatch), Data 360 now automatically saves it in a dedicated <strong>Problem Records Data Lake Object</strong> instead of quietly dropping it. This is a big improvement for finding data quality issues — we can query the Problem Records DLO directly to see exactly which rows failed and why, then fix the source data or adjust our transform logic. Before this, failed rows would simply disappear with no way to see them. Now they’re kept so we can investigate.</blockquote><h3>Private connectivity — for regulated environments</h3><p>Many companies in regulated industries (financial services, healthcare, public sector) block all public internet access to their data lakes. Without a private connection, Data 360 can’t connect to those environments at all. The platform solves this with support for <strong>AWS PrivateLink </strong>— a network-level connection that keeps all traffic entirely inside AWS’s own network, using private addressing that never touches the public internet.</p><p>For companies using multi-clouds, where the data lake sits on Azure or Google Cloud, Data 360 extends the private connection using <strong>cross-cloud interconnects</strong> — either set up by the customer (Azure ExpressRoute, Google Cloud Interconnect, Equinix Fabric) or fully setup by Salesforce. In both cases, Data 360 connects to external data sources as if they were local, while keeping the same guarantee that no data touches the public internet.</p><blockquote><strong>Ready-to-Use Data Lake Objects:</strong> Ready-to-Use DLOs are pre-built Data Lake Objects for the most common Salesforce source objects. Instead of manually defining the structure and mappings yourself, these ready-made templates get the first DLOs created and ready to map within minutes — cutting down the time between setting up a connector and getting the first unified profile. Especially useful for teams doing their Data 360 implementation.</blockquote><blockquote><strong>The golden rule of ingestion:</strong> Only bring in what the use case actually needs. Every row ingested consumes credits. Every column you don’t need is wasted storage and added cost in identity resolution — the most expensive operation in the platform. Define the use case first, then work backwards to the exact source fields required.</blockquote><blockquote><strong>Architect pattern — incremental everywhere:</strong> Whether you’re using batch ingestion, batch transforms, or full refreshes, always turn on incremental mode when its available. A full refresh reprocesses the entire dataset every time it runs. Incremental mode only processes what has changed since the last successful run— made possible by Data 360’s Iceberg-based architecture. At large scale, the difference in cost and speed between the two can be enormous.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=19455a4089dd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data 360 Series | Chapter 2 — How Data 360 is built: objects, storage, and the org model]]></title>
            <link>https://medium.com/@.ramya/data-360-series-chapter-2-under-the-hood-the-data-360-lakehouse-58e8d62c7153?source=rss-43e703f09ac6------2</link>
            <guid isPermaLink="false">https://medium.com/p/58e8d62c7153</guid>
            <category><![CDATA[architecture]]></category>
            <category><![CDATA[salesforce-data-360]]></category>
            <category><![CDATA[data-cloud-in-salesforce]]></category>
            <category><![CDATA[data-cloud-consultant]]></category>
            <category><![CDATA[salesforce-data-cloud]]></category>
            <dc:creator><![CDATA[Ramya M]]></dc:creator>
            <pubDate>Mon, 25 May 2026 17:47:30 GMT</pubDate>
            <atom:updated>2026-08-21T17:28:04.919Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Before you build anything on Data 360, you need a clear picture of where data lives, how it flows, and what governs it at each layer.</em></p><p>Data 360’s architecture follows one main idea: bring in data from anywhere, shape it into one standard form, unify it around the customer, and let every other system — AI, marketing, sales, service — benefit from it. Let’s walk through each layer.</p><h3>The Medallion Architecture: Bronze → Silver → Gold</h3><p>Data 360 uses a <strong>Medallion Architecture </strong>— a common pattern in modern data platforms, where data gets better quality as it moves through each layer. Think of it like making a coffee: raw green beans go in, they get roasted and ground, and what comes out is the finished espresso.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*iIkyYszjzALZcI3F" /></figure><h3>The key objects we’ll work with every day</h3><p>Data 360 has a specific vocabulary. Understanding these terms is non-negotiable before we touch any configuration:</p><h4><strong>Data Source Object (DSO)</strong></h4><p><strong>— the true raw landing zone. </strong>Every external data stream arrives here first in its native file format (e.g. CSV, JSON). It is <strong>temporary staging only</strong> — no transformations, no persistence. Salesforce-native connectors skip the DSO and land directly in the DLO.</p><h4><strong>Data Lake Object (DLO)</strong></h4><p>— the <strong>first storage layer that saves data permanently</strong>. It’s created automatically from a DSO (or you can create it manually).</p><p>Data is stored in <strong>Apache Parquet</strong> format — a way of organizing files by column instead of by row, which makes it much faster to query than a CSV (it only reads the columns you ask for, and skips the rest). These files are stored in Amazon S3 cloud storage, organized as tables by <strong>Apache Iceberg</strong> — a table-management layer that adds safe reliable writes (ACID transactions), the ability to look at yesterday’s data (time travel), and the ability to add a new column without rewriting the whole table (schema evolution).</p><p>This is also where the system checks that data matches the expected format (schema enforcement) and labels it by category — Profile, Engagement, or Other. Any data transformations happen here, before the data is mapped to a DMO. Supported field types: text, number, date, boolean, email, URL, phone, and percent.</p><h4>Unstructured Data Lake Object (UDLO)</h4><p>— a storage object built specifically for <strong>unstructured content</strong>, meaning content that isn’t organized into rows and columns such as PDFs, Word documents, support call transcripts, audio file details, and emails. It connects from storage (Amazon S3, Azure Blob Storage, Google Cloud Platform) or via any existing Data 360 connector.</p><p>UDLOs feed into a five-step process (Parse → Pre-process → Chunk → Post-process → Embed) that turns raw content into searchable AI data called vector embeddings. The matching model object is called a UDMO (Unstructured Data Model Object). <em>This becomes important in the AI chapter — come back after reading chapter </em><a href="https://medium.com/@.ramya/data-360-series-chapter-6-ai-models-agentforce-and-the-intelligence-layer-78100e1da246"><em>Chapter 6</em></a>.</p><h4>Data Model Object (DMO)</h4><p>— the organized, mapped layer. DLOs get mapped to DMOs and DMOs follow the C360 Data Model, which has 300+ standard objects that work across any industry.</p><h4>Data Space</h4><p>— A way to divide and <strong>control access to</strong> data, metadata, and processes — by brand, region, or department — all within one Salesforce org. Data Spaces control who can see what: each DLO and DMO belongs to exactly one Data Space, and you can’t query across Data Spaces unless you’ve explicitly given the permission.</p><p>Permission sets are directly tied to Data Spaces, and every action inside them is logged for compliance purpose. Think of it as several separate workspaces living inside one Salesforce org.</p><h3>Home Org and Companion Org — the two-org model</h3><p>New to Data 360? Here’s the simple version: all your data lives in one Salesforce Org (the Home Org). Other Salesforce orgs that connect to it can read the unified data without making their own copy. The details below is for designing deployments with multiple orgs.</p><p>Data 360 works using two-org model. This is important to understand for any enterprise (large-scale) implementation:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*1Vm56MBDIbBYmyZ8" /></figure><blockquote><strong>Key insight:</strong> The most common mistake in a first implementations is treating Data 360 like a regular Salesforce object model. It isn’t one. The DSO → DLO → DMO → Unified Profile pipeline works very differently. The platform also has <strong>two</strong> storage tiers — the Lakehouse (Parquet/Iceberg on S3) for historical and batch data, and the <strong>Low Latency Store</strong> (fast SSD storage) for real-time in-session data. Plan for both from the start, or you’ll need to redesign your system later.</blockquote><h3>The two storage tiers you need to know</h3><p>The Medallion diagram above shows the <em>logical</em> layers — how data is organized conceptually. The actual physical storage is split into two tiers that work together:</p><h3><strong>Lakehouse layer </strong>—</h3><p><strong>Apache Parquet + Iceberg on S3</strong><br>Large scale, cost-efficient storage for historical and batch data. Handles analytics, ML, and long-term persistence at petabyte scale. This is where DLOs, DMOs, and Calculated Insights live.</p><p>Data 360’s primary storage foundation is built on Apache Parquet files stored in Amazon S3, organized and governed by Apache Iceberg.</p><p><strong>S3</strong> — infinitely scalable, cheap cloud storage. Stores the files.</p><p><strong>Parquet</strong> — column-oriented file format. Makes storage efficient and queries fast.</p><p><strong>Iceberg</strong> — table format on top of both. Adds ACID transactions, time travel, schema evolution, and snapshot-based incremental processing.</p><p>Together they form what the industry calls a <strong>data lakehouse</strong> — the scale and flexibility of a data lake, with the reliability and query performance of a data warehouse. That’s exactly what Data 360’s DLO layer is built on.</p><h3>Low Latency Store (LLS) —</h3><p><strong>Fast SSD — millisecond access</strong><br>A reliable, large-scale SSD cache that sits above the Lakehouse. Live Agentforce conversation data, real-time profile information, and frequently used data are stored here so they can be retreived in under a millisecond. Data automatically moves to the Lakehouse for long-term storage.</p><p>Because S3 access takes tens to hundreds of milliseconds, the LLS exists for data that needs to be retrieved in single-digit milliseconds — active Agentforce conversation history, customer profile context pre-fetched for real-time personalization, and hot in-session data.</p><p>It acts as a durable, intelligent buffer: in-session data flushes down from memory into the LLS, and frequently-needed Lakehouse data gets pulled up into it. Data eventually migrates to the Lakehouse for long-term persistence.</p><p>Together, the three tiers — memory, LLS, and Lakehouse — form a seamless storage hierarchy that lets Data 360 serve both millisecond-latency agentic interactions and petabyte-scale batch analytics from the same unified platform.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ULvURR9sp2dyRyQU" /></figure><blockquote><strong>Spring ’26 — Salesforce Unified Lineage:</strong> Data 360 now includes a visual data lineage map — a full view showing exactly how data moves, from its original source, through ingestion, transformation, and unification, all the way to where it gets used (activation). Think of it as a GPS trail for your data. If something looks wrong in a segment or an activation, you can trace it back to where it started within seconds. This is a major tool for governance and troubleshooting that used to require third-party software.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=58e8d62c7153" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data 360 Series | Chapter 1: Why Salesforce Data 360?]]></title>
            <link>https://medium.com/@.ramya/data-360-series-chapter-1-why-salesforce-data-360-934c3c4d3374?source=rss-43e703f09ac6------2</link>
            <guid isPermaLink="false">https://medium.com/p/934c3c4d3374</guid>
            <category><![CDATA[salesforce]]></category>
            <category><![CDATA[salesforce-data-360]]></category>
            <category><![CDATA[salesforce-data-cloud]]></category>
            <category><![CDATA[data-360]]></category>
            <category><![CDATA[cloud-data]]></category>
            <dc:creator><![CDATA[Ramya M]]></dc:creator>
            <pubDate>Sun, 24 May 2026 17:01:21 GMT</pubDate>
            <atom:updated>2026-08-21T13:24:59.151Z</atom:updated>
            <content:encoded><![CDATA[<h3>Data 360 Series | Chapter 1 — Why Salesforce Data 360 (formerly Data Cloud)?</h3><p><em>This chapter explains what the problem was with Salesforce platform to handle big data at scale, how Data 360 was built to solve it, and where it stands today.</em></p><p>Let’s start at the beginning — not with Data 360 itself, but with the problem it was built to solve. Understanding that problem makes the rest of the platform much easier to follow.</p><h3>The problem: the same customer, invisible to itself</h3><p>Salesforce built the world’s most successful CRM (Customer Relationship Management) software. A CRM records sales deals, tracks support cases, and managing contacts. But it had one big limitation: it was not built to handle very large amounts of data.</p><p>Meanwhile, over almost two decades, Salesforce acquired many other successful companies. Marketing Cloud (through ExactTarget, for email and campaigns), Commerce Cloud (via Demandware, for e-commerce), Pardot (for B2B marketing). Each of these ran on its own database, with its own way of structuring data, and its own customer ID system.</p><p>The result? The same customer — let’s call Ben Parker — had four different IDs across four different systems. No one could see the whole picture of who he was. A customer who had already bought twice online still received “first purchase” discount emails. A Gold loyalty member got generic, impersonal service replies. The data existed. It was just locked away in separate systems that couldn’t talk to each other.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*38Hg8BzQVGs6T2kV" /></figure><h3>The rebranding timeline one need to know</h3><p>Before we go further, let’s clear up with the name confusion. This platform has had five names in six years. Each new name marked a real jump in what the product could do — not just a marketing change:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*J9KyUJGzwbRwb0sO" /></figure><h3>What Data 360 actually is —</h3><p>In simple terms, Salesforce Data 360 is a <strong>petabyte-scale data lakehouse</strong> (a combination of data lake storage and data warehouse-style querying — more on this in <a href="https://medium.com/@.ramya/data-360-series-chapter-2-under-the-hood-the-data-360-lakehouse-58e8d62c7153">Chapter 2</a>). It has built-in identity graph, a shared data model, and a system for sending data to other tools — all running on Salesforce’s trusted infracture. It can ingest data from 270+ sources, process billions of records, and deliver that data to AI agents, marketing campaigns, and sales workflows almost instantly.</p><p>Think of it this way — its a customer data platform and AI foundation built on metadata, running on Hyperforce (Salesforce’s cloud infrastructure). Importantly, it <strong>does not replace</strong> the data tools you already use — Snowflake, Databricks, BigQuery, Redshift. Instead, it connects to them, making their data trustworthy and usable inside Salesforce, without copying it.</p><p>Data 360 is now part of <strong>Agentforce 360 platform </strong>— Salesforce’s end-to-end AI platform. This means Data 360 isn’t just a data management tool anymore. It’s the memory and intelligence layer behind every AI agent in Salesforce. Through <strong>Salesforce Headless 360</strong>, Data 360 is also available as an API and an MCP (Model Context Protocol) server. This means AI coding agents and external tools can access Salesforce customer data directly, without opening any screen.</p><h3>The five-phase lifecycle</h3><p>Everything in Data 360 follows the same basic sequence. Every use case — from personalizing an email to grounding an AI agent’s answers — flows through some or all of these phases:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Otfo761g_v6Byq0g" /></figure><blockquote><strong>Key takeaway:</strong> Data 360 is not just a product update — it reflects a whole new way of thinking about the platform. The name “360” means this is no longer just about customer data in the cloud, it’s about a complete 360-degree view of every customer, powered by AI and grounded in a unified data foundation.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=934c3c4d3374" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Connect DBeaver to Salesforce Data Cloud (Data 360)]]></title>
            <link>https://medium.com/@.ramya/connect-dbeaver-to-salesforce-data-cloud-bab6f3024649?source=rss-43e703f09ac6------2</link>
            <guid isPermaLink="false">https://medium.com/p/bab6f3024649</guid>
            <category><![CDATA[salesforce-data-cloud]]></category>
            <category><![CDATA[data-cloud-in-salesforce]]></category>
            <category><![CDATA[data-360]]></category>
            <category><![CDATA[database]]></category>
            <category><![CDATA[dbeaver]]></category>
            <dc:creator><![CDATA[Ramya M]]></dc:creator>
            <pubDate>Mon, 18 Aug 2025 17:01:55 GMT</pubDate>
            <atom:updated>2026-04-04T03:43:45.992Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/895/1*HIA_iC6CE8gdl5AtcQ18FA.png" /></figure><p>DBeaver is a SQL software application and a database administration tool, you can use to Salesforce Data Cloud. It is easier to navigate because of clean UI that helps to run and save SQL queries, browse tables and visualize results.</p><h3>Salesforce Data Cloud Instance:</h3><p>Follow these steps to connect the DBeaver tool to Salesforce Data Cloud -</p><h4>1. Enable OAuth Settings</h4><p>On Data Cloud —Setup, navigate to OAuth and OpenID Connect Settings page and turn on ‘Allow OAuth Username — Password Flows’.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*cxDAKaS22nimsp492Rpb6g.png" /><figcaption>Fig. OAuth and OpenID Connect Settings</figcaption></figure><p>This will helps users to login to Salesforce from an OpenID provider and authorize Salesforce to access data. Here, OpenID Connect acts as an authentication provider.</p><h4>2. Connected App</h4><p>Create a new connected app to enable DBeaver to integrate with Salesforce Data Cloud.</p><p>a) From Salesforce Setup, Go to App Manager and click on New Connected App.</p><p>b) Provide appropriate name (Ex: Data Cloud DBeaver Connection), API name and contact email.</p><p>c) Under the API section, mark ‘Enable OAuth Settings’ checkbox and enter following callback URLs -</p><p><a href="http://localhost:55555/Callback">http://localhost:55555/Callback</a><br><a href="http://localhost:55556/Callback">http://localhost:55556/Callback</a><br><a href="http://localhost:55557/Callback">http://localhost:55557/Callback</a><br><a href="http://localhost:55558/Callback">http://localhost:55558/Callback</a><br><a href="http://localhost:55559/Callback">http://localhost:55559/Callback</a></p><p>d) From the available OAuth Scopes, select the following options -</p><p>• Manage Data Cloud profile data (cdp_profile, api)<br>• Manage user data via APIs (api)<br>• Perform ANSI SQL queries on Data Cloud data (cdp_query_api)<br>• Perform requests at any time (refresh token, offline access)</p><p>e) Save and then click Continue</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*IOj0raK8ZyHobBFnzMjiZA.png" /><figcaption>Fig. Connected App</figcaption></figure><h4>3. Consumer Details</h4><p>In case you didn’t taken backup of the consumer details, from the connected app, click on the ‘Manage Consumer Details’ button and copy the credentials -</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*CvPKmoydRiktNocjjmIIqw.png" /><figcaption>Fig. Consumer Credential Details</figcaption></figure><h3>DBeaver:</h3><h4>1. Install DBeaver</h4><p>Download and install the DBeaver application -<a href="https://dbeaver.io/download/">https://dbeaver.io/download/</a></p><h4>2. New Database Connection</h4><p>On successful software installation, from the navigation menu, select <strong>Database </strong>Menu and click on ‘New Database Connection’</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*KvnHWL1kzVCNnTBf_ogazg.png" /><figcaption>Fig. New Database Connection</figcaption></figure><h4>3. Select Database</h4><p>On the Connect database panel, search for Data Cloud and select the right database (<em>Make sure to select latest version and not legacy</em>) -</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*ZsNuqmjnTfwu1ugKc9YHgA.png" /><figcaption>Fig. Salesforce Data Cloud</figcaption></figure><h4>4. Connection Settings: Authentication</h4><p>On the connection settings panel, under Authentication section, enter your salesforce Org login id as the Username and salesforce password with org. security token as the Password.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*ZlhbrNBZU0c7_Vd4ww636w.png" /><figcaption>Fig. Authentication</figcaption></figure><h4>5. Configuration Settings: Driver Properties</h4><p>a) Navigate ‘<strong>Driver properties</strong>’ tab and click on ‘<em>Add user property</em>’ option below to enter the consumer credentials</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*ntbkUoy72simiuMXL5QeHQ.png" /><figcaption>Fig. Driver Properties</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*3PDlidvSEus-ef8s0n9uPQ.png" /><figcaption>Fig. User Properties</figcaption></figure><p>b) Enter the copied Salesforce Consumer Key as the value for <strong>clientId </strong>and Consumer Secret as the value for <strong>clientSecret </strong>-</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*zT9Jrh_Mou9dM5UQ7ZHTGw.png" /><figcaption>Fig: Credentials</figcaption></figure><h4>6. Test connection</h4><p>Navigate to main tab and click on Test connection button to successfully connect DBeaver to Data Cloud database -</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*QgAtz3aJ_yWLxRc4GFYEKg.png" /><figcaption>Fig. Database Connection</figcaption></figure><p>On click of OK button, on the left pane you would see the successfully connected Database</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/290/1*YT5cDtZzVTAtJ4c02_ow3g.png" /><figcaption>Fig: Database</figcaption></figure><p>Expand the Database and click on table to see the list of Data Cloud objects present in Data Cloud.</p><p>On selection of specific object table, you would have the ability to view the object fields under Properties tab, records under Data tab and UML under Diagrams tab -</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*UgLZB4ocfiH4cqjHdYA5lw.png" /><figcaption>Fig. Sample DLL</figcaption></figure><p>Additionally, if you’d like to rename the Database, right click on the newly created database, select <strong>Edit Connection</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/798/1*_bpe1j_S1vSlxtHeAA1z-g.png" /><figcaption>Fig: Edit Connection</figcaption></figure><p>Under Connection Settings, Go to <em>General </em>Tab and update the Connection name.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*PyBnlLFl_NaVMVo1cc0nWg.png" /><figcaption>Fig: Rename Connection</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bab6f3024649" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>