Matias Rodriguez Frank

How data people at different levels communicate: Junior Data Person: - Built 30 dashboards in one year, driving operational transparency. - Wrote 500 lines of SQL code for data cleaning, improving data accuracy by 15%. - Shared reports with 10 departments, impacting 100+ stakeholders. - Automated 4 data sources, saving 10+ hours of manual work weekly. Medior Data Person: - Developed integrated, user-centric dashboards, boosting decision-making efficiency by 20%. - Collaborated with key stakeholders to build cross-functional reports for better alignment. - Mentored 2 interns and 1 junior engineer, enhancing team capabilities. - Automated workflows, saving 30+ hours of manual reporting weekly. Senior Data Person: - Designed a scalable data architecture, reducing query execution time by 40%. - Mentored 3 medior and junior team members, improving team productivity by 25%. - Presented actionable insights to C-suite, directly contributing to a 15% increase in revenue. - Led end-to-end data strategy for a major project, integrating 10+ complex data sources seamlessly. Manager (Data Leader): - Defined and implemented the data strategy for the organisation, aligning it with business goals. - Spearheaded the transition to a modern data stack, reducing operational costs by 30%. - Fostered a data-driven culture, increasing data utilisation across teams by 50%. - Managed a diverse team of analysts and engineers, delivering projects 20% ahead of deadlines. #data #analytics #dataengineering #ai #ML

290

10 Comments

📉 Data Engineering is misunderstood even in 2025. Startups still think they don’t need Data Engineers. Until they’re drowning in messy dashboards, missing reports, broken pipelines, and paying 10x in Snowflake bills 🫠 Let’s be clear: ❌ Data Engineering is not just writing SQL queries. ❌ It’s not just about building dashboards. ❌ It’s not about “connecting a few APIs” and calling it a day. Data Engineering is the foundation of a modern data-driven company. It’s the system behind the insights, the engine behind ML, the reason your dashboards load fast and stay accurate. A Good Data Engineer can: ✅ Design scalable & resilient architecture ✅ Build robust and automated ETL/ELT pipelines ✅ Handle schema evolution without breaking downstream models ✅ Optimize storage using the right partitioning, formats, and governance ✅ Ensure data quality, versioning, access control, and monitoring Tired of looking for a perfect resource to learn data engineering? Well, I have built a one-stop solution for people like you 🫵🏻 6 Course, 16+ Projects and Complete Interview Preperation Platfrom ALL AT ONE PLACE, check below ⬇️ #dataengineer #dataengineering

869

14 Comments

𝗣𝗼𝘄𝗲𝗿 𝗕𝗜 𝗗𝗲𝘀𝗶𝗴𝗻 𝗜𝘀𝗻’𝘁 𝗳𝗼𝗿 𝗬𝗼𝘂 - 𝗜𝘁’𝘀 𝗳𝗼𝗿 𝗧𝗵𝗲𝗺 You understand the data. You know the model, the logic, and the layers of DAX behind every tile. But here’s the reality: 𝗠𝗼𝘀𝘁 𝗣𝗼𝘄𝗲𝗿 𝗕𝗜 𝘂𝘀𝗲𝗿𝘀 𝗱𝗼𝗻’𝘁 𝘁𝗵𝗶𝗻𝗸 𝗹𝗶𝗸𝗲 𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿𝘀. Studies by Nielsen Norman Group show that users often misread dashboards when the design doesn’t match how people naturally process information - usually because of poor layout, unclear labeling, or overwhelming information. The takeaway? 𝗝𝘂𝘀𝘁 𝗯𝗲𝗰𝗮𝘂𝘀𝗲 𝗶𝘁 𝗺𝗮𝗸𝗲𝘀 𝘀𝗲𝗻𝘀𝗲 𝘁𝗼 𝘆𝗼𝘂, 𝗱𝗼𝗲𝘀𝗻’𝘁 𝗺𝗲𝗮𝗻 𝗶𝘁 𝘄𝗼𝗿𝗸𝘀. When we build Power BI reports, we often do it from our perspective: • “This layout feels clean to me.” • “I know where to click.” • “The logic is obvious.” But for users, the experience is completely different: • They don’t know what to focus on first • They hesitate to click in case it “breaks” something • They don’t trust numbers they can’t trace or explain 𝗗𝗲𝘀𝗶𝗴𝗻𝗶𝗻𝗴 𝗳𝗼𝗿 𝘆𝗼𝘂𝗿𝘀𝗲𝗹𝗳 𝗶𝘀 𝗮 𝘀𝗵𝗼𝗿𝘁𝗰𝘂𝘁. 𝗗𝗲𝘀𝗶𝗴𝗻𝗶𝗻𝗴 𝗳𝗼𝗿 𝘂𝘀𝗲𝗿𝘀 𝗶𝘀 𝗮 𝘀𝗸𝗶𝗹𝗹. If you want adoption, trust, and action: • Watch users interact with your report (even just once) • Ask them what they found confusing • Use their behavior-not your instinct-to guide layout and interaction Remember - You’re building a 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻 𝘁𝗼𝗼𝗹 for someone who doesn't live in the data like you do. #PowerBI #UIUX #DataViz

330

19 Comments

Looking to take your data visualization to the next level? The ggdist package extends ggplot2 with specialized tools for visualizing distributions, uncertainties, and comparing data groups. It’s a powerful add-on to your ggplot2 toolkit, especially useful for communicating complex insights with ease. Key benefits of using ggdist: ✔️ Enhanced Distribution Visualizations: Create rich visuals like eye plots, slab intervals, and gradient intervals, making it easier to interpret complex data distributions. ✔️ Detailed Uncertainty Displays: Add flexible uncertainty intervals to your charts, helping to convey the full range of your data without clutter. ✔️ Improved Comparison Options: Effectively compare groups or distributions in a visual style that remains clear and interpretable. ✔️ Seamless Integration with ggplot2: Works smoothly with ggplot2 syntax, so you can incorporate it into your existing workflow without extra setup. With ggdist, you can go beyond the basics of ggplot2 to build plots that convey more information without losing clarity. This package offers specific solutions for data sets with high levels of uncertainty or multiple comparison groups, ideal for fields like statistics, finance, or any data-driven research. The visualization shown here is an example from the ggdist package website, highlighting its unique capabilities: https://lnkd.in/eWfqDqBP If you want to explore this and many other related topics, check out my online course, Data Visualization in R Using ggplot2 & Friends. Learn more: https://lnkd.in/ehRrzdJ7 #datasciencecourse #ggplot2 #dataviz #package #datastructure #analytics #statisticsclass #rstudio #tidyverse #datavisualization

59

𝗧𝗵𝗲 𝗣𝗼𝘄𝗲𝗿 𝗕𝗜 𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 𝗹𝗮𝗻𝗱𝘀𝗰𝗮𝗽𝗲 𝗶𝘀 𝗮𝗯𝗼𝘂𝘁 𝘁𝗼 𝘀𝗵𝗶𝗳𝘁 𝗱𝗿𝗮𝗺𝗮𝘁𝗶𝗰𝗮𝗹𝗹𝘆, 𝗮𝗻𝗱 𝗺𝗼𝘀𝘁 𝗼𝗿𝗴𝗮𝗻𝗶𝘇𝗮𝘁𝗶𝗼𝗻𝘀 𝗮𝗿𝗲𝗻'𝘁 𝗽𝗿𝗲𝗽𝗮𝗿𝗲𝗱 𝗳𝗼𝗿 𝘄𝗵𝗮𝘁'𝘀 𝗰𝗼𝗺𝗶𝗻𝗴. While companies debate AI governance and use cases, the practical integration points are already here. I've built a working demonstration that connects Claude AI directly to Power BI Desktop through MCP protocol - enabling real-time model exploration, DAX development, and performance optimization. What this means for BI teams: → 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 𝘃𝗲𝗹𝗼𝗰𝗶𝘁𝘆: Routine measures, calculations, and optimizations handled automatically → 𝗞𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗱𝗲𝗺𝗼𝗰𝗿𝗮𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻: Junior developers get instant access to senior-level DAX patterns and best practices → 𝗤𝘂𝗮𝗹𝗶𝘁𝘆 𝗮𝘀𝘀𝘂𝗿𝗮𝗻𝗰𝗲: AI can run performance traces and identify bottlenecks faster than manual review → 𝗗𝗼𝗰𝘂𝗺𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻 𝗱𝗲𝗯𝘁: Models can be documented and maintained automatically I watched Claude analyze a multi-million row model, create optimized DAX measures, run performance diagnostics, and generate documentation that would take hours manually. The accuracy was impressive when properly constrained to work within the actual data model. Yet, AI hallucination is real - I caught it trying to shortcut with JavaScript instead of proper DAX. This technology requires validation frameworks and proper guardrails, not blind trust. Besides that, as multiple people have pointed out, feeding real data into LLMs, that will later be used for training - is a no-go. However, providers already offer solutions to that - business plans siloed from post-training. Organizations that establish AI-assisted development workflows now will have significant competitive advantages in analytics delivery speed and quality. Those that wait will be playing catch-up in 12-18 months. This isn't about replacing analysts or developers - but about removing friction from technical tasks so we can focus on business insight and strategic value creation. ____ #PowerBI #MicrosoftFabric #LLM #MCP #AI #FutureOfWork

184

22 Comments

NEW UPDATE - Tired of losing your flow in BigQuery just to peek at a schema? Ever been halfway through writing SQL and thought “what was that column called again?” 🙋‍♀️ BigQuery have just rolled out a shiny new Reference panel that lives right beside your query editor, and it’s a game-changer for productivity. Here’s what it gives you: ✨ The last tables you viewed (no more hunting through datasets) ✨ Full schema for those tables (column names + data types right where you need them) ✨ A one-click preview of the data (without writing extra SQL) No more flipping between tabs or digging through the Explorer panel while you’re in flow. It’s all right there, side-by-side with your code. Honestly, it feels like BigQuery is slowly turning into the IDE we always wanted. Hopefully next we'll be able to rename our tabs!!! 📖 Official docs if you want to explore: https://lnkd.in/dKygsnbM 👉 Have you spotted this update yet? I’d love to know if you think it’ll change the way you write SQL. #BigQuery #SQL #DataAnalytics

76

10 Comments

The reason many data teams FAIL with Data Quality. This is a very common scenario: - One of your datasets break - Investigation reveals that the problem is upstream - The reason is a change introduced by the data producers At this stage the data team has two options: 1. Go to the data producers and make them fix the issue 2. Put a plaster on the problem by correcting the data in the pipeline In too many cases the team will go with Option 2. Why? Because it's EASIER. It's in their control. But this will almost always be the WRONG way to go about it. By doing this you: - Accept poor data at source - Add unnecessary complexity to your pipelines - Accumulate technical debt you'll need to pay for in the future - Lower the standards and show your producers it's ok to create poor data Don't be that team. Settling for poor data will have really bad long-term implications. Set a high bar.

145

57 Comments

At work, the product team had launched new features. We as BI analytics had to closely monitor numbers, adoption and metrics. Three weeks into it, we noticed mixed adoption and occasional dips in engagement. Our dashboard would tell us the numbers, but not why. So this was not enough. 𝐓𝐡𝐞 𝐜𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞: Well, the UAT (user acceptance testing) hadn’t caught many of the real-world problems customers would face, even if we are launching it to 10% of 400 million user base we have. I needed a way to get closer to the ground truth, get something good to take this forward, I wanted to above and beyond for these new launches. 𝐇𝐞𝐫𝐞'𝐬 𝐰𝐡𝐚𝐭 𝐈 𝐭𝐡𝐨𝐮𝐠𝐡𝐭: Instead of waiting for structured surveys, or ratings, I pulled a mix of sample Play Store reviews and what people complaint in the tickets they created. This was a huge pile of unstructed data, no joke. And you cant really look at them directly and get something. I set up a framework to: -Tag complaints by product/feature. -Spot recurring keywords (e.g., “crash,” “payment stuck,” “too many steps”). (used few python libraries I knew) -Filter specifically for new product launches where early feedback mattered most. Surprisingly, some reviews were not just complaints but customer-suggested improvements. For example, one review about the wallet feature exposed an onboarding glitch, other had trouble creating tickets(api failures and since this was small, nobody really spotted) while another pointed to confusing eligibility on how to register for Rupay-CC that matched a drop we were seeing in funnel conversion. Now, let's talk about the solution I worked around: - I compiled these insights into a tracker that tied review themes back to dips in numbers. Shared with product teams, it gave them a direct line from customer voice → product fixes. This was not done at once, took some time, mails, and the product loved these solutions and how I stepped out to help. 𝐓𝐡𝐢𝐬 𝐰𝐚𝐬 𝐭𝐡𝐞 𝐨𝐮𝐭𝐜𝐨𝐦𝐞: This surfaced over 10+ actionable product issues in a short span. Not all were taken up but it gathered a lot of attention, and fixing them not only stabilized metrics but also created goodwill. Because customers obviously feel heard when the issues they flagged got resolved/lead to innovation in the product. We're talking about an app that had 400 million + user base. 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠𝐬 𝐟𝐨𝐫 𝐚𝐥𝐥 𝐝𝐚𝐭𝐚 𝐚𝐬𝐩𝐢𝐫𝐚𝐧𝐭𝐬 & 𝐩𝐞𝐨𝐩𝐥𝐞 𝐢𝐧𝐭𝐞𝐫𝐯𝐢𝐞𝐰𝐢𝐧𝐠:- As a business analyst (or later as a senior data professional), your role can go far beyond just analyzing and creating dashboards. Look at how proactively I stepped in, saw a problem, created an untraditional approach (hunted for unstructured data), created a method to systematize it, and linked it to business metrics and voila, happy customers! *plays Wake me up by The Weeknd* in bg

130

7 Comments