Joel Labes

Emma McGrattan breaks down the 1-10-100 rule in such a clear, practical way. Such a great explanation of why catching issues early truly pays off. Give it a watch and see how proactive data observability can make all the difference. #DataQuality #DataObservability #DataGovernance #AIreadyData

2

1 Comment

The Tech Lead role that requires stakeholder management skills as much a technical capability? Your Principal Data Engineer who needs to understand enough about various product features to critically evaluate your tech choices, and help train new team members? Any difficult role where you’re seeing candidates, but they’re not quite right because they’re just not quite ‘good enough’ and you can’t put your finger on why? There’s a high likelihood that the brief you’ve provided to your recruiters is missing some key CONTEXT. ✅ Context for your recruiter means being able to provide a deeper, and therefore more compelling, message to take to market – which increases the interest levels of in-demand talent. It allows them to evaluate your role on its merits, rather than a job spec ✅ Context means recruiters have clearer parameters. But clearer parameters usually mean wider parameters, where they can be more creative with the candidates they target, over and above a word matching exercise ✅ Context means your recruiters should be able to screen and then pitch their candidates to you in a deeper way – reducing the likelihood of you rejected a promising candidate based on something arbitrary that isn’t in their CV. So if you can’t fill your role – step number 1 – have a look at how the context for the role has been communicated, received and understood.

4

For data tools to really drive the business I’m finding that they need to embed more closely to where work is actually done. Instead of reports or data tools being in one platform and expecting a business user to view them and then take action in the other platform connections should be made directly where possible. The tool itself should be able to apply the insights. I’m finding between modern AI tools and the ease of deployment with tools like Palantirs foundry or Vercel you can build an app that not only provides insights but can make them immediately actionable in the same platform. #data #analytics #datadriven #businessinsights

1

🚀 2-Day Databricks Workshop We recently completed a 2-day Databricks workshop at the organisation, focusing on strengthening our data engineering foundations and getting everyone aligned on core concepts. The goal was about ensuring we’re all speaking the same language when it comes to: 🔹 Lakehouse architecture fundamentals 🔹 Spark-based distributed processing 🔹 Data transformation workflows 🔹 Notebook collaboration & best practices 🔹 Scalable pipeline thinking Workshops like this are so valuable because strong data teams aren’t just built on tools - they’re built on shared understanding. When everyone is aligned on architecture, workflows, and best practices, execution becomes smoother, faster, and more reliable. Grateful for the collaborative discussions and the opportunity to reinforce core principles together. Strong foundations → better systems → bigger impact. #Databricks #DataEngineering #DataTeams #BigData #Lakehouse #TechLearning #AI

4

You've used Inmon. You've used Kimball. Now let's talk about Data Vault. If you've spent time in either camp, Data Vault will feel strange at first. There are no facts. No dimensions. No normalised enterprise model sitting at the centre. Just Hubs, Links, and Satellites. But once it clicks, you'll understand why it exists. Inmon gives you a clean, normalised source of truth. Great for consistency, painful to change when the business does. Kimball gives you fast, intuitive models built around business processes. Great for analysts, but you're making design decisions upfront that are hard to undo. Data Vault says: what if we stopped making those decisions upfront entirely? Hubs capture business keys, nothing else. Links capture relationships between those keys, nothing else. Satellites carry everything that changes over time, fully historised by default. The result is a model that absorbs change without breaking. New source system? Add a Hub. Business process changed? Add a Link. New attribute arrives? Add a Satellite. Nothing you built before needs to touch. This matters a lot in modern pipelines where sources are unreliable, requirements shift constantly, and you need full auditability. The tradeoff is real though. Data Vault is not easy to query directly. You still need a consumption layer on top, whether that's dimensional models, wide tables, or whatever your BI tool prefers. Think of it as your integration layer, not your serving layer or if you're going for medallion architecture then think of it as your silver layer, not Gold layer. If Inmon is about correctness and Kimball is about usability, Data Vault is about survivability. For teams building long-lived enterprise data platforms, that's a very compelling reason to consider it. #DataEngineering #DataVault #Inmon #Kimball #DataModelling #DataWarehouse

4