Nora Wadman-Küthe

After all the setup and design work in Parts 1 & 2 of my Spotify Data Pipeline project, we’ve finally reached the payoff: getting our raw Spotify data into a structured database. This part is all about turning messy source files into organized tables so that everything is in place for the exciting steps still to come. In Part 3: Preparing Data the Right Way, we load tracks, artists, albums, and more into a fully functioning database that you can query and build on. Along the way, I cover: • How to decide when to lean on SQL versus Python so each tool does what it’s best at • Why an intentional, well-organized file structure in your project root will save you from headaches later (and how you can get clever with using these files in ways you wouldn’t expect) • How to design steps that accomplish multiple goals at once without making things harder to maintain • The importance of designing each step to build predictably on the last, so expanding later is painless By the end of this part, you won’t just have organized data. You’ll have a working database foundation that unlocks everything to come next. And here’s the payoff: Part 4 is where all of this groundwork comes alive. We’ll take that neatly organized database and use it to automatically find and log a music video on YouTube for every track. It’s the moment the whole pipeline has been leading toward and the reason we’ve been laying bricks one by one. https://lnkd.in/ghDjgE2F

1

"Saying 'you are wrong' is not curious. Saying 'why are your priors different than what the data is showing' is curious." Bryan Bischof on Data Renegades, on why the best data leaders share one trait: they exhaust your knowledge with follow-up questions. Not accepting the first dashboard. Not asking "what's our number?" and moving on. Following the thread until it reaches a question that changes behavior. He learned from a feedback class that feedback is a question, not a statement, and started applying that to everything. The best exec he worked with, Yonda, would always find the follow-up question Bryan hadn't anticipated. That's the question that changes operations. The contrast: leaders who hear data they don't like and spend all their energy finding errors in the analysis instead of asking what it means. Listen to full episode wherever you get your favorite podcasts. #DataScience #Analytics #DataCulture #Leadership

1

1 commentaar

In the world of Piano Analytics, every week brings a new improvement. Within Boards, it's now possible to customize the width of each chart or page element, and you can also adjust the height of each row of the board. This is a welcome evolution that gives maximum flexibility when defining dashboard zoning. It may seem minor, but these controls on rows really change the game (practiced and adopted this week building dashboards for a client): charts are no longer necessarily sized according to how many of them you place on each slice of the report. In the example below, I was able to create two rows of charts with a 1/3 - 2/3 proportion. And if I want to give more space to what is visible by default, I can increase the height of the product table just below, making it possible to see more than 5 table rows. I can also use a text zone filled with a blank space to preserve a two-column layout and make the page more airy. The design engine of Boards is becoming richer and richer while still keeping a certain structure that prevents bad practices. A good balance for a dashboarding module integrated into an Analytics suite. Looking ahead, and since Christmas is just around the corner, there are two features I would love to see added to Boards in the future🎄: 1️⃣ The ability to set a different time granularity at the chart level, so that upon loading the board, one chart could show a weekly trend while another shows a monthly trend, for example. 2️⃣ An option to select the metric(s) used to sort each chart, independently of the added report. That would greatly expand the possibilities for multi-view dashboarding. Just saying Mélanie Claisse.😁 #dataviz #analytics

37

2 commentaren

Tired of spending forever digging up data? Rik Pauwels explains how Conversational Analytics can help you ditch the dashboard-first approach and get answers in seconds with secure AI. Stop wasting time and unlock the hidden value in your data! Read the full breakdown to learn more.

5

Okay - picking up from last week. The new company is DataFabrIQ. Site is live today: https://datafabriq.co/ The short version: most BI tools ask you to trust the data. We think you shouldn't have to. The longer version goes back to what I wrote about a few days ago. Execs, finance, and RevOps teams at growing mid-market companies are running their most important workflows — MRR close, pipeline review, commission runs, variance reporting — in a mess of Google Sheets, scripts, and Slack threads sitting on top of Salesforce, Hubspot, Shopify, NetSuite, and whatever else. Every month they rebuild the same arithmetic, argue about whose number is right, and lose days to reconciliation. I lived this at Meal Ticket. Every operator I've talked to in the last five months is living it now. The usual answer is "buy another BI tool" or "hire more people to fix it." Neither works. A dashboard chart is the last mile. The problem is everything underneath. So we built the underneath. One auditable data model across billing, CRM, accounting, and payments. Every number traces back to a source record. Every transformation is versioned SQL anyone on the team can read. And on top of that, purpose-built workflows for the jobs that actually eat your team's week — so the MRR number on Monday's pipeline review is the same number that lands in the board deck. Live today: Pipeline Review, MRR Rollforward, Commission Runs, Variance Reporting, plus A/R and broken-payments reporting. The list keeps growing as our design partners tell us what's next. One thing I want to say out loud: none of this was buildable a year ago, at least not at the price and speed we can deliver it now. AI changes what one small team can reason about across a messy data stack.   It's a big part of why we think this is the moment, and it is the most fun I’ve ever had building a company. If your week involves a giant spreadsheet, a wall of vLookups, or a pivot table you're afraid to touch - you should see what we built. Comment, DM, or hit "Request access" on the site. Twenty minutes, no deck. Shoutout to my amazing cofounder Bill Hoffman, and our early investors for giving us a shot to build this.  And, to our design partner teams allowing us to build on top of their real data, and to all of you who have quietly supported and helped over the last five months — thank you. You know who you are. So excited about the future!

82

18 commentaren

I recently managed to turn a 5-hour GA4 data crawl into a 3-minute sprint, but the path there wasn't exactly straightforward. This is a short engineering post-mortem on a bottleneck that had me stuck for a while. I’m sharing the specific, somewhat undocumented Enzyme flags I eventually used to manage 5.8 TB of history in Databricks. Hopefully, this helps anyone else hitting similar walls with DLT. You can find the full technical retrospective on Medium: https://lnkd.in/d5cNMnxg

9

Most pricing tools stop at exact matching. That's not a strategy — that's a spreadsheet. Exact matching tells you what your identical product costs elsewhere. Useful? Sure. Enough? Not even close. It misses the alternatives shoppers actually compare. It ignores refurbished channels eating into your margins. And it's blind to competitor promotions until the damage is done. Real pricing intelligence needs four layers — not one. #PricingStrategy #CompetitiveIntelligence #Ecommerce #RetailTech #PriceMonitoring

5

Wow, what a day at Snowflake Summit! 🤯 The keynote yesterday was absolutely packed with game-changing announcements, especially around AI. My mind is still buzzing! Here are the top 5 updates that really blew me away: 𝗦𝗻𝗼𝘄𝗳𝗹𝗮𝗸𝗲 𝗔𝗜 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲: Imagine being able to talk to all your data – whether it's neat tables or messy documents – super easily. This is a huge step towards making data analytics a lot simpler and easier! 𝗖𝗼𝗿𝘁𝗲𝘅 𝗔𝗜𝗦𝗤𝗟: This one is wild! Now, we can use simple SQL (the language many of us already know) to understand and analyze things like images, documents, and even audio files directly in Snowflake. For anyone dealing with lots of unstructured data, this is a total lifesaver 𝗢𝗽𝗲𝗻𝗙𝗹𝗼𝘄: Data just got way easier to bring into Snowflake. They've made it simple to connect to all sorts of data sources seamlessly, cutting down tons of manual work. 𝗪𝗼𝗿𝗸𝘀𝗽𝗮𝗰𝗲𝘀: This is fantastic for data teams! Now, we'll have dedicated workspaces within Snowflake that can directly access dbt projects and integrate with Git. Everything in one place, making our workflows so much simpler! 𝗔𝗱𝗮𝗽𝘁𝗶𝘃𝗲 𝗖𝗼𝗺𝗽𝘂𝘁𝗲: Snowflake can now automatically adjust its power (like cluster size and number) based on demand. This means less infrastructure headaches and potentially big cost savings. So many incredible innovations! I'm really looking forward to seeing how these new features empower us to do even more with data and AI. Check out this link for more information https://lnkd.in/gPs3HRMx What are you most excited about from these announcements? Let me know in the comments! 👇

233

5 commentaren

Your model is a decaying asset the moment you deploy it. We learned this the hard way. A customer churn model, validated with 94% AUC, went into production. Three weeks later, its precision had dropped over 20%. The culprit: data drift. A new marketing campaign completely changed the distribution of our user acquisition features. Our post-mortem fix wasn't a better model. It was a better system. We built a monitoring service using a daily Spark job. It computes the Population Stability Index (PSI) on key features, comparing the production inference data against the training set baseline. If PSI on a critical feature exceeds a 0.25 threshold, it doesn't just send a PagerDuty alert. It triggers a Jenkins pipeline that automatically retrains the model on the last 30 days of data and stages the new artifact for deployment review. A model isn't a project; it's a system. Don't ship a model. Ship a system that monitors, retrains, and redeploys that model. #MLOps #DataEngineering #MachineLearning

7

AI on relational data? It’s not just possible—it’s powerful. 🤯 Kumo's co-founder Hema Raghavan is living proof. She's spent her career bringing predictive modeling to the enterprise, and now, with Kumo inside Snowflake, she's doing it with zero data movement or governance headaches. See how you can: ⚡️ Build predictive models directly on relational tables. 🧠 Incorporate unstructured data with LLMs. 🔒 Keep governance and security locked down. 👩‍💻 Seamlessly deploy through Snowflake Marketplace. Ready to do AI better? Watch the full video to learn more. http://bit.ly/3VHoT65

134

7 commentaren

Is your data inspiring real company growth? 📉 Here's the secret: Even if you have great data flows and slick dashboards customized for your team, the REALITY is that people are not robots. They get busy, distracted and often only check their dashboards 5 minutes before the review. So how do you get your data to create ACTION? How do you get people the care? The answer is that numbers alone don’t inspire change. You need to make it PERSONAL so they really care and focus on what's most important for the company. Making it personal means: 1. Make your data come alive: Tell stories and make it real. 2. Directly linking your bottom line to their personal interests 🧠 When data is framed as a narrative, what’s happening, why it matters, and what to do about it, people listen. Storytelling boosts data engagement by 22% and turns raw numbers into real decisions. Want to bring your data to life? Try these storytelling moves: Use trend lines to spotlight opportunities 📈 Focus visuals on one key message at a time Frame charts as “chapters” in a bigger story 📘 Match your message to what your audience cares about 🎯 Want to make it personal? You need to gamify it: 🎖 Reward team members for creating high-impact stories 🏆 Run a “Best Insight of the Month” challenge 📊 Score dashboards for clarity, relevance, and actionability 📣 Let teams vote on weekly standout data stories When people feel connected, they start to care, and that’s when change happens. How do you make your data more human? Drop your best tips below 👇

1

This week taught me something uncomfortable. I realize that products lose ARR long before dashboards show them. And many times, most users don’t churn because a feature is broken. They churn because something felt slow, unresponsive, or “off” in the moment they needed it. And you only see the damage days or weeks later. When your product is slow, that's not a technical issue, it's a revenue leak. That’s why AIOps matters. It predicts issues, prevents frustration, and protects the trust your PLG depends on. The truth is, Your product experience is shaping your ARR… even when you’re not watching. Where does your product feel slow right now? #AIOps #PLG #SaaS #GrowthStrategy 

5

🚫 Metadata ≠ Insight Too many teams launch a “data catalog,” stop at column names + descriptions, and wonder why analysts still drown in 17 look-alike customer tables. The real unlock? Search that adds context A modern metadata platform should behave like a world-class search engine: • Prioritise fresh & popular assets people query today • Boost 🟢 certified datasets ahead of ⚠️ drafts • Understand natural intent (“churn dashboard”, “high-value users”)—not just exact table names • Honour access controls so results match your permissions How to get there → treat metadata as a knowledge graph Connect the dots between datasets, owners, metrics, dashboards, and queries. When every relationship becomes an edge in the graph, magic happens: • Usage + lineage → Smart ranking of the right table, first time • Stored joins + query examples → LLM-generated SQL that actually runs • Tag-to-owner links → One-click access requests and auto-approvals • Metric-to-dashboard paths → Root-cause data-quality issues in minutes Quick litmus test for any catalog 1. Search “revenue by cohort.” Does the certified dataset appear on top? 2. Hover the result. Do you see usage stats, owners, and sample joins? 3. Ask it to generate SQL. Does the query run without edits? If not, you have metadata—just not the context that turns it into insight. Building graph-based search right now? Let’s swap lessons learned. 💬 #AI #Innovation #Technology #DataCatalog #DataEngineering #KnowledgeGraph

20

What do you do when you're flooded daily with AI how-to posts, plenty of “you’re missing out if you’re not using this” and you have limited time, due to personal circumstances, to keep up with everything new and shiny? You choose a few short courses to broaden or update your perspective, sharpen some skills, and hopefully gain guidance on which of the many AI options are actually worth investing in. The Prompt Engineering Masterclass from Maven Analytics is one of those courses. @chrisbruehl explains a range of more advanced concepts and provides many demos, all framed around the BRIDGE model for accurate and complete prompting. Although the course uses ChatGPT Pro, which I don’t have, the core ideas translate easily to other models, whether as practical steps to take in prompting or good-to-have (or must have) features regarding topics like security. That made the course very worthwhile for me right now. #data #AI #prompting #bridge

6

1 commentaar

𝐒𝐨 𝐦𝐚𝐧𝐲 𝐭𝐞𝐚𝐦𝐬 𝐰𝐚𝐧𝐭 𝐆𝐞𝐧𝐀𝐈, 𝐝𝐚𝐭𝐚 𝐦𝐞𝐬𝐡, 𝐬𝐡𝐢𝐧𝐲 𝐀𝐈… 𝐛𝐮𝐭 Their data isn’t even clean yet. I sat down with Maaike Van Den Branden, a data & AI leader with nearly 20 years of experience building high-impact data teams, and she broke things down in such a practical, no-buzzword way. She helps organizations turn data strategy into real business outcomes — not just dashboards that nobody uses. 𝐇𝐞𝐫𝐞’𝐬 𝐰𝐡𝐚𝐭 𝐰𝐞 𝐠𝐨𝐭 𝐢𝐧𝐭𝐨 ⬇️ Why leaders chase the “next thing” before fixing basics The real blockers: data quality, governance, adoption — not tools Her “MumsGPT” example: 1 week of work → minutes ⚡ What a new CDO should do in the first 90 days 3 KPIs CEOs actually care about: revenue, adoption, innovation The biggest red flag: everyone knows data is bad… but no one says it 𝘔𝘺 𝘧𝘢𝘷𝘰𝘳𝘪𝘵𝘦 𝘱𝘢𝘳𝘵 ❤️ “People say they’re data-driven… but they don’t want to open the reports.” That’s where GenAI actually helps — make data easier to ask, easier to understand, easier to adopt. 🎥 Full episode dropping soon on The Executive Outlook. Your turn: What’s REALLY stopping your org from getting value from data — tech, people, or data quality? Tell me in the comments. #DataLeadership #AIinBusiness #DataGovernance #GenAI #TheExecutiveOutlook #ComplereInfosystem

15

When was the last time your dashboard actually changed a decision? Great BI isn’t about how the data looks—it’s about how well it guides people. - It simplifies complexity - Aligns teams with a shared version of truth - Turns noise into direction - Builds trust through clarity, not just visuals Because at the end of the day, BI isn’t there to impress. It’s there to inform, empower, and help you make the next right move. Drop your thoughts in the comments. Let’s talk about what really makes BI valuable. #DataAnalyst #DataScientist #BusinessIntelligence #BIwithImpact #DataDrivenDecisions

19

13 commentaren

Hi there! Meet my new article: 'Archiving Automatisation on Tableau Server: Tableau + Python' at Gett Tech blog. Here you can find all steps on how to implement this on your Tableau server 😎 The link is in the comments.

28

1 commentaar

Tableau Bites 271st Edition is out today! ✔️ Top 2000 in Perspective – Behind the Scenes - Johan de Groot ✔️ Team Datafam Data + Maps Tableau User Group 12 Dec 2025 - Dennis Kao, Rod Sardari, PhD, GISP and Agata Ketterick ✔️ D29 Tableau IronViz Secrets: Storytelling, Colour Theory and Cultural Connections (Part 2) - Sarah Burnett and Fi Crocker ft. Brian Moore, Varun J., Kinsey Miller and Louis Yu ✔️ Tableau Next Tips #4: Features - What You Can and Can’t Do - Tarun Gupta ✔️ Two Approaches to Software Development: Lessons from Tableau's Evolution - Craig Bloodworth ✔️ Tableau Liquid Glass Dashboard - Samantha Cohn ✔️ Our Favorite Vizzes of 2025 - Kevin Flerlage and Ken Flerlage ✔️ My Top-5 Posts of 2025 - What You Liked Most - Anastasiya Kuznetsova Find the link in my profile, or Google “Tableau Bites” to prepare for your lunch break read today! #Tableau #TableauTips #TableauTraining

5