Why fragmented data erodes trust in analytics

Your data problems aren't actually about data—they're X-rays revealing deeper organizational issues. Data struggles are not just broken dashboards or fragmented databases—they're revelations about how teams collaborate, how decisions flow, and how leadership shapes priorities.  👉 If Finance's spreadsheets can't talk to Marketing's dashboards, it's because Finance and Marketing aren't talking enough.  👉 Overengineered analytics pipelines emerge from fear of making bold decisions.  👉 Meaningless KPIs come from avoiding tough alignment conversations. Think of data health as an organizational early warning system—the cultural canary revealing hidden fault lines. When leadership ignores anomalies or fails to invest in proper governance, what looks like neglected data is actually a mirror of neglected organizational health. If you can't measure customer retention, that's not a data gap—it's a priorities crisis. Here's the kicker: This creates a vicious feedback loop. Poor data drives flawed decisions, which reinforces the problems that created the poor data. Take a marketing department working with unreliable lead attribution—they'll inevitably misallocate resources, deepening organizational inefficiencies and eroding trust in decision-making. When no one trusts the numbers, "the data is broken" becomes a convenient excuse for "We'd rather not face our internal misalignments." Teams retreat to gut instincts and outdated heuristics, further distancing themselves from reliable insights. Left unchecked, this pattern breeds a culture where finger-pointing trumps progress. The path forward requires treating data issues as leadership imperatives: 👉 First, create unified goals that demand cross-functional collaboration—shared KPIs that break down territorial walls. 👉 Second, elevate data literacy to the same level as financial fluency across your organization. 👉 Third, and most crucially, simplify. Complexity isn't sophistication—it's a tax on your organization's agility. The organizations that thrive won't be the ones with the most advanced tech stacks or the biggest data teams. They'll be the ones who recognize that data health and organizational health are two sides of the same coin. You can’t fix organizational issues by fixing the data.

Putting pressure on data science teams to deliver analytical value with LLMs is cruel and unusual punishment without a scalable data foundation. Over time, the best LLMs will be able to write queries as effectively or more effectively than an analyst - or at minimum make writing the query easier. However, the most cost-intensive aspect of answering business questions is not producing SQL, but deciding what the query inputs should be and determining whether or not the inputs are trustworthy. Thanks to the rapid evolution of microservices and data lakes, data teams find themselves living in a world of fragmented truth. The same data points might be collected by multiple services, defined in multiple different ways, and could actually be going in opposite and contradictory directions. Today, data developers must do the hard work of understanding and resolving those discrepancies, which comes in the form of 1-to-1 conversations with the engineers managing logs and databases. Very few if any service teams at a company have documented their data for the purpose of analytics. That results in a giant gap in documentation across 1000s of datasets across the business. Without this gap being filled, data scientists will ultimately have to manually hand-check any prediction that an LLM makes in order to ensure it is accurate and not hallucinating. The model is doing a job with the information it has, but the business is not providing enough information for the model to deliver trustworthy outcomes! By investing in a scalable data foundation, this paradigm flips on its head. Data is well documented, clearly owned, and structured as an API enforced by contracts that define the use case, constraints, SLAs, and semantic meaning. A quality-driven infrastructure is a subset of all data in the lake, which reduces the surface area LLMs need to make decisions only to the nodes in the lineage graph which have clear governance and change management. Here's what I suggest: 1. Start by identifying which pipelines are most essential to answering the business's most common questions (you can do this by accessing query history) 2. Identify the core use cases (datasets/views) that are leveraged in these pipelines, and which intermediary tables are of critical importance 3. Define semantically what the data means at each level in the transformation. A good question to ask is "What does a single row in this table represent?" 4. Validate the semantic meaning with the table owners 5. Get the table owners to take ownership of the dataset asn API, ideally supported programmatically through a data contract 6. Define the semantic meaning and constraints within the data contract spec, mapped to a source file 6. Limit any usage of an LLM to the source files under contract Good luck! #dataengineering

I keep having the same conversation with technology leaders across industries. "We're deploying agents." "We're scaling AI." "We're building copilots." Great. But when I ask about their master data, product hierarchies, supplier records, customer segmentation, the room goes quiet. Nobody wants to admit but agentic AI doesn't fix bad data, it weaponizes it. An agent that auto-triggers replenishment based on flawed demand signals doesn't save you money. It builds excess inventory with machine-like precision. An agent that reclassifies customers using inconsistent data doesn't improve personalization. It erodes trust at scale. And the numbers back this up. 81% of AI professionals say their company still has significant data quality issues, yet 85% say leadership isn't addressing them (Qlik, 2025). Gartner, Deloitte, and McKinsey all converge on the same finding: 70–85% of AI project failures trace back to data, not algorithms. IBM's 2025 CDO Study found that over a quarter of organizations lose more than $5M a year to poor data quality alone. And Salesforce report 84% of technical leaders say their data strategy needs a complete overhaul before AI can succeed. In retail, poor product master data skews demand forecasts and disrupts fulfillment. In finance, mismatched definitions of "active account" distort entire risk models. These aren't edge cases, they're the norm. So before you invest another dollar in your agent roadmap, ask yourself one question: Would I trust a new hire to make decisions using the data we have today? If the answer is no, that's your real priority. Your data strategy isn't supporting your AI strategy. Your data strategy IS your AI strategy. #AgenticAI #DataQuality #MasterData #AIStrategy #DigitalTransformation #EnterpriseAI #DataGovernance #AIROI #SupplyChain #RetailTech https://lnkd.in/evCerNx2

In my previous post, I discussed the inevitability of data silos. Today I want to focus on quantifying their true impact. Most conversations about data silos focus on the obvious costs: - Duplicate systems - Manual data entry - Reconciliation efforts While significant, these are merely the visible tip of a much larger iceberg. The more insidious costs remain hidden yet profoundly impact performance: 1) Decision latency: When information is fragmented across systems, decisions stretch into weeks as teams await complete data. Meanwhile, competitors who've solved this problem execute strategic pivots while others are still gathering facts. 2) Contradiction: When departments present conflicting "facts" about the same business reality, valuable executive time is wasted in reconciliation, eroding trust in data-driven decision making altogether. 3) Opportunity blindness: When customer data, product usage data, and financial information remain disconnected, the cross-functional insights that often represent your most profitable opportunities remain invisible. 4) Innovation tax: When each initiative requires custom integration work, innovation becomes prohibitively expensive. Teams either create quick, disconnected solutions (tomorrow's silos) or delay projects awaiting proper integration, neither supporting the rapid experimentation needed for growth. 5) Analytics confidence gap: When analysts spend 80% of their time acquiring and cleaning data rather than interpreting it, their analyses become superficial. The resulting insights rarely challenge established thinking or reveal counterintuitive opportunities. 6) Regulatory exposure: When crucial information is confined to isolated systems, compliance efforts are hindered by fragmented data views. This leads to missed deadlines, inaccurate reporting, and potential penalties. How can we quantify these costs? While challenging, it's not impossible: - Measure decision cycle times, tracking time spent on data collection versus analysis - Calculate hours consumed reconciling conflicting data sources - Audit innovation projects for delays directly attributable to data access issues - Track the percentage of analytics capacity dedicated to data preparation versus insight generation - Document financial penalties from regulatory reporting delays or inaccuracies In my next post, I'll outline practical steps to address these costs without requiring a complete organisational restructure or technology overhaul. What hidden costs have data silos created in your organisation? Have you found effective ways to measure their impact? #DataStrategy #DataGovernance #DigitalTransformation #Management #Innovation

Your biggest reporting problem is political. When multiple systems produce multiple “truths,” teams stop trusting the numbers, and meetings turn into debates. Over time, frustration builds. Not because people are misaligned, but because the system forces them to be. That’s the hidden cost of fragmented data. The solution isn’t another dashboard. It’s a single, governed source of truth. When everyone works from the same numbers, conflict drops and clarity returns. If decisions feel harder than they should, fix the foundation first.

Finance says revenue is $10M. Product says $9.8M. The CEO asks, “Which one is right?” and everyone stares at their laptops. You spent $500K to unify your data platform. Now you have two conflicting versions of the same KPI. What went wrong? I've seen this exact scene play out in boardrooms more times than can be counted. Different teams. Same Microsoft Fabric platform. Completely different numbers for the same metric on the same day. Finance pulls from the Warehouse. Product queries the Lakehouse. Both think they're showing "Revenue," but the definitions drifted weeks ago, and nobody noticed until the executive meeting. Now trust is broken. Not in the people. In the data itself. This isn't a technical failure. It's a governance failure. The problem is simple: organizations deploy both Lakehouses and Warehouses without deciding which one owns what. Different teams build parallel pipelines, apply their own transformations, and create separate versions of customers, orders, and revenue. Microsoft Fabric gives you the flexibility to do this. But without clear contracts, flexibility becomes fragmentation. Here’s what happens: • Power BI reports connected to the Warehouse show one number • Python notebooks reading from the Lakehouse show another • Executives stop trusting both and go back to Excel You can't fix this with better Slack communication. You need three things: •Define source of truth per domain. For Customers, Products, Orders, Finance- pick one authoritative source. Everything else derives from it, not from the operational system. •Introduce data contracts. Document schemas, refresh SLAs, and change rules. Breaking changes require notice to downstream consumers. •Centralize metrics, not infrastructure. Define "Revenue" once in a shared semantic layer. Let teams consume it however they want- Power BI, Python, Excel- but the definition stays consistent. The hardest part isn't technical. It's getting Finance and Product to agree on what "Revenue" actually means. But once that happens, the architecture is simple. If a Fabric deployment feels like it's technically working but trust is still low, this is why. The platform was built- the contracts were not. 2026 is the year where "We're on Fabric now" stops being an excuse and starts being a commitment: one source of truth, one definition of Revenue, disagreed once and agreed in writing. Save this 🛟 if you're dealing with conflicting KPIs across Fabric services. The fix isn't choosing Lakehouse vs. Warehouse. It's defining who owns what- and enforcing it. #MicrosoftFabric #DataStrategy #Lakehouse

Salesforce acquiring Informatica for $8B isn’t just another deal — it signals the real truth about AI: 👉 The AI race is actually a data race. Salesforce didn’t buy a new LLM or a ChatGPT competitor. They bought data infrastructure. Because the uncomfortable reality is this: AI fails when data fails. Gartner estimates that 85% of AI projects collapse not because models aren’t smart, but because data isn’t clean, connected, or governed. I’ve seen this across industries — companies spend $50M+ on PhDs, platforms, and agents, only to realise they built an AI skyscraper on data quicksand. When AI agents operate on fragmented, outdated, or inconsistent data, you don’t get intelligence… 👉 you get confident mistakes at scale. Marc Benioff said it best: “To get your AI right, you must get your data right. Without trusted data, AI is just hallucination.” This acquisition is a clear message to the entire market: Foundations matter more than the model. If an AI agent must answer a customer query by checking CRM, Finance, Warehouse, and Service systems — and each one has gaps or outdated records — then even the most expensive AI is only guessing. The winners of 2025 won’t be the ones with the flashiest algorithms. They’ll be the companies doing the “unsexy” work: • Cataloging data assets • Building high-quality pipelines • Governing access & lineage • Creating a unified data layer for agents • Embedding Responsible AI across workflows Before deploying one more agent, ask the $8B question: Is our data ready to be trusted — or are we just automating confusion? As we move toward autonomous agents, data is no longer just an asset; it’s the engine of the enterprise.

Most hospitals are losing millions every year — not from care gaps, but from data gaps. I once sat with a clinical leader trying to answer a basic question: “Did this patient ever get their follow-up?” She clicked through five different systems — payer portal, EHR, scheduling tool, lab reports, discharge notes. By the time she pieced it together, the patient was already in the ER. This isn’t just an IT headache. It’s a: • Finance problem → lost reimbursements, write-offs • Care problem → delayed interventions, preventable admissions • Trust problem → clinicians burned out, patients frustrated Studies show data fragmentation costs U.S. hospitals $8.3B annually in inefficiencies and duplicate testing. Connected records + real-time insights don’t just make ops cleaner — they save lives and protect margins. The real cost isn’t in building the pipes. It’s in not doing it fast enough. If your team had seamless data access tomorrow, what would change first? #DigitalHealth #HealthcareInnovation #DataInteroperability #HealthTech #AIinHealthcare #PatientCare

One of the biggest structural problems in logistics today is fragmented data, which quietly undermines efficiency, visibility, and innovation across the entire ecosystem. Every company experiences this issue, but few discuss it openly. Data is stored in too many disconnected places: TMS, WMS, ERPs, carrier portals, supplier spreadsheets, dispatch tools, and warehouse applications. None of these systems aligns, flows together, or speaks the same language, resulting in complications far beyond simple manual work. Many carriers and shippers use non-integrated systems, and smaller fleets often rely on multiple, uncommunicative tools. As a result, 82% of companies see fragmented data as the biggest barrier to AI and analytics readiness. Additionally, multi-tier suppliers often go unnoticed in networks, posing risks that standard dashboards can’t identify. What are the consequences of this fragmentation?  - Slower decision-making - Manual reconciliation - Inaccurate reporting - Broken workflows - Compounding errors - AI models that fail before they can even begin - A network that consistently reacts too late This is why integration and unified data layers are essential. Companies that are addressing fragmentation are transitioning to: - API-first, real-time data flow - Centralized data layers instead of scattered spreadsheets - Clean, consistent master data - Fewer tools with deeper connections - Cross-functional visibility instead of operating in silos Because once the data layer works, everything gets easier: execution, visibility, billing, claims and yes AI 🙌