--- title: "Fragmented data became a decision system | Dagny" description: "Dagny unified fragmented data in BigQuery and built the Looker analytics, LTV, and experimentation systems leadership used to make decisions." canonical: "https://dagnylabs.ai/case-studies/health-analytics" last-updated: "2026-10-25T13:46:10Z" --- # Fragmented data became a decision system | Dagny [HTML page](/case-studies/health-analytics) [All customer stories](/case-studies) [Mogl](/case-studies/mogl-billing) # Fragmented data became a decision system Dagny unified fragmented data in BigQuery and built the Looker analytics, LTV, and experimentation systems leadership used to make decisions. Health data & analytics · 6 min read BigQueryUnified data foundation LookerAnalytics layer LTV + A/BDecision models ## Key outcomes - Data foundation: Unified in BigQuery - Analytics: Modeled and explored in Looker - Decision systems: LTV model + A/B testing platform ## Context A health technology company had valuable data across the company’s application, Salesforce, A/B testing platforms, survey platforms, Airtable, and Google Sheets. Each system held part of the picture, but the company did not have one analytics foundation that connected product behavior, customer context, experiments, feedback, and operating data. Teams stitched together exports, reconciled definitions, and rebuilt context for each analysis. The CMO and CEO had access to data, but not a shared decision system built on consistent models and governed business logic. Dagny unified the source data in BigQuery, built a complete analytics layer, made the analytics layer usable through Looker, and then extended the foundation with an LTV model and an A/B testing platform. ## Fragmented decision data - Scattered sources - Manual joins - Mixed definitions - Slow analysis - Harder decisions The fragmented data environment created recurring problems: - Product and application data was separated from Salesforce account and commercial context. - Experiment exposure and survey feedback were disconnected from the company's core product and customer models. - Airtable and Google Sheets contained useful operating logic that was not governed in a shared analytics layer. - Metric definitions varied across tools, teams, and individual analyses. - The CMO and CEO lacked a common foundation for evaluating performance, customer economics, and experiments. ## The analytics layer Dagny began with the data foundation. The source systems were unified in BigQuery, where shared models could connect product behavior, customer context, experiments, feedback, and business operations. On top of that foundation, Dagny built the complete analytics layer and exposed the analytics layer through Looker. LTV and experimentation systems then extended the same governed models into recurring leadership decisions. ### 1. Unified in BigQuery Dagny brought together data from the application, Salesforce, A/B testing platforms, survey platforms, Airtable, and Google Sheets. The resulting BigQuery foundation gave the company one place to connect the full customer and business picture. ### 2. Built the analytics layer Dagny modeled the company's core entities, relationships, and metrics so analysis could reuse shared logic instead of reconstructing definitions for every question. This became the governed layer between raw source data and business decisions. ### 3. Served through Looker Looker made the modeled data available for exploration, reporting, and recurring analysis. Leaders and operators could work from the same definitions while moving from a high-level result into the underlying evidence. - Source systems - BigQuery - Analytics models - Looker ### 4. Built an LTV model Dagny built a reusable lifetime-value model on top of the unified foundation. Customer economics could now be evaluated through a shared methodology tied to the same product, commercial, and behavioral data used elsewhere in the analytics layer. ### 5. Built an A/B testing platform Dagny also built an experimentation platform that connected experiment setup, exposure, outcomes, and analysis to the core data model. Tests could be evaluated in the same context as customer behavior, business performance, and long-term value. ## Connected systems SystemRole in the system ApplicationProduct events, usage, and behavioral data. SalesforceAccount, relationship, pipeline, and commercial context. A/B testing platformsExperiment exposure, variants, and outcomes. Survey platformsCustomer feedback, sentiment, and research signals. AirtableOperational records and team-maintained context. Google SheetsSupplemental operating data and analysis inputs. BigQueryUnified warehouse and modeled analytics foundation. LookerAnalytics exploration, reporting, and decision support. Source systems , then BigQuery , then Analytics layer , then Looker - LTV model - A/B testing platform - CEO + CMO decisions One governed analytics layer connected source data, business definitions, analysis, and decisions. ## Results ### One decision-ready foundation Data from six source categories was brought together in BigQuery and organized through a shared analytics layer. Teams could work from connected data and consistent definitions instead of rebuilding the context for each analysis. ### A reusable LTV model The company gained a common model for evaluating customer economics, grounded in the same data foundation used for product and business analytics. ### An integrated experimentation platform Experiment setup and analysis became part of the core analytics system, connecting test outcomes with customer behavior and longer-term value. ### Shared decision support for leadership The CMO and CEO could evaluate performance, customer value, and experiments through a connected set of models and analytics rather than separate views assembled for each decision. ## Before and after - Before: Data was fragmented across the application, Salesforce, experiments, surveys, Airtable, and Sheets - After: Source data unified in BigQuery - Before: Teams joined exports and reconciled definitions for each analysis - After: Shared analytics models reused across the business - Before: Reporting logic lived across tools and individual workflows - After: A complete analytics layer served through Looker - Before: Customer economics lacked a shared, reusable model - After: An LTV model built on the unified foundation - Before: Experiment data was disconnected from the core analytics model - After: An A/B testing platform tied experiments to trusted outcomes - Before: The CMO and CEO made decisions across separate views - After: A shared decision foundation for leadership ## Takeaway Dagny turned fragmented data into decision infrastructure: unified in BigQuery, modeled as a complete analytics layer, accessible in Looker, and extended through LTV and experimentation systems built for leadership decisions. ## Health technology company An anonymized Dagny customer story Industry / teamHealth data & analytics Reading time6 min read In this story [Context](#context) [Fragmented decision data](#fragmented-decision-data) [The analytics layer](#the-analytics-layer) [Connected systems](#connected-systems) [Results](#results) [Before and after](#before-and-after) [Takeaway](#takeaway) [How Dagny works](/our-process) ## More customer stories [All stories](/case-studies) [Finance operations Hours to days to 10–15 min Invoice generation, reported by QBR's founder - Workflows - Applications - Rails ### QBR Media brought billing and collections into a shared operating system QBR Media](/case-studies/billing-operations) [Insurance brokerage 60 min to <5 min Time per proposal - Workflows - Applications ### From Epic records to a client-ready proposal Riemer Insurance Group](/case-studies/insurance-proposals) [Customer success operations 2 weeks Scope to solution - Data & Integrations - GTM Systems ### Customer health needed a trusted foundation High-growth SaaS company](/case-studies/customer-health) ## What’s the first workflow you’d fix? Tell us where your team gets stuck. We’ll discuss what a first build could look like. [Book a call](/book-a-call)