Analytics Anonymous

That "quick question" over Slack has been the bane of data analysts forever. Imagine those are now handled by ChatGPT, giving quick and reliable answers to business users. Stakeholders are happy, and data analysts can focus on deeper, more impactful work. Are we about to finally see this happening? In this episode I talk with David Jayatillake (Co-Founder & CEO at Delphi Labs) about how large language models like GPT-4 are changing the way we work with data. What does this mean for data analysts or analytics engineers, and where do these new tools fit into the modern data stack? Key takeaways: A lot of tools already offer a text-to-SQL approach. While this can be very useful to increase productivity for data analysts or analytics engineers, it's problematic as an interface for business users. When the semantic layer is effectively generated on the fly with every new query, results are unpredictable and can lead to a loss of trust. With a semantic layer, analytics engineers and data analysts can implement business logic and and expose data and metrics to business users in a safe and reliable way. (For example, dbt offers a semantic layer, but a lot of BI tools like Looker or Metabase have their own as well.) Delphi builds on top of these existing semantic layers, offering a natural language interface for business users. Instead of digging through a BI tool, stakeholders can simply ask their question in Slack. The answers will be limited to what is defined in the semantic layer, therefore avoiding the risk of wrong results. When data analysts are freed from answering "simple" requests, they can focus on deeper, more complex work to generate insights and recommendations to the business. While AI might eventually be able to take over most operational tasks, David believes that strategic decision making will still require human oversight in the future. Besides building data tools, David is also very active in the data community. He hosts a Mastodon server for data folks, and you can find him on dbt Slack and Locally Optimistic. You should also check out his Substack where he's written a lot about semantic layers recently. Find David and Valentin on LinkedIn.

Have you ever dreamt of being your own boss? Work on projects that you choose, on your own schedule, at your own rates? If you already have experience working with data and analytics, becoming a freelance consultant is a great way to break out of the corporate grind. In this episode I talk with Jekaterina Kokatjuhha about how to become a freelance data & analytics consultant. She shares her personal experience and practical tips for how to get started, build a brand and audience, and overcoming uncertainty. Key takeaways: Before going solo, Jekaterina worked in various data roles in several different industries. The insights into different business models she gained there are helpful for understanding her clients problems now. Her first freelancing client resulted from a match on the Bumble dating app. She started consulting with one day per week, while still being employed full-time. When this was going well she decided to leave her job completely. When she started looking for new clients, she focused on D2C brands. This allowed her to capitalize on her deep experience in this area and target her communication to this audience. On LinkedIn, Jekaterina writes about common problems these companies face (e.g. which metrics to care about). When potential clients reach out to her, she asks them what content resonated most with them, so she knows where to put her focus. To increase her reach on LinkedIn, she posts around 8am and aims for 100 reactions within 1 hour, to get boosted by the feed algorithm. She also posts selfies occasionally. While this used to make her feel uncomfortable, it's important for people to connect her face to her content. Going beyond solopreneur freelancing, her next step is building a data agency. Her goal is always to help businesses extract more value from their data. Find Jekaterina and and Valentin on LinkedIn.

You've seen the pictures of people working on their laptops in beautiful, exotic locations. Exploring the world while you work – the digital nomad lifestyle is nothing new, but it's getting a lot more common, in particular since we all learned to work without an office over the last years. And the data analyst job is well suited for this lifestyle. In this episode, I talk with Melanie Dietrich about the benefits and challenges of working as a digital nomad data analyst. She also shares her story of breaking into data, coming from a business background. And how we can work on closing the gender gap in tech (and data). Key takeaways: Finding the right accommodation is a challenge when working on the go. You want a good desk and good wifi, but that's often not obvious from the descriptions on Airbnb. Going to a coworking space means additional expenses, but can help to connect with the local community of digital nomads. For internet access, it's good to have backup options. SpaceX Starlink works great for van life, but is too heavy for backpacking. Coming from a business background (audit consulting), Melanie wanted to move beyond Excel and taught herself data analysis with SQL and Python, using online courses. When looking for your first job in data, it's important to put yourself out there and demonstrate your knowledge. Networking is key. Helping business users solve their problems establishes your role in the team. Become the go-to person for their data questions and teach them how to use the available data tools themselves. Data science skills (e.g. machine learning) are often not so relevant in daily work. Data engineering skills are often more in demand. Melanie is a co-founder of the Women in Data x Business career network. They organize events and share experiences to encourage more women to chose a career in data. Find Melanie and and Valentin on LinkedIn.

It's easy to get started with SQL, but mastering it requires a deep understanding of database architecture and query optimization, which takes time and practice to develop. In this episode I talk with Michaël Scherding (Analytics Engineering Manager at SFEIR) about his journey to becoming a SQL expert. We also dive into the unique challenges of scaling an analytics platform to more than 100,000 employees. Key takeaways: Michael started his career working with Oracle systems. The technical limitations back then forced you to think about optimizing your code. He had a senior colleague with 25 years of SQL experience reviewing his code and learned a lot from that. While there are lots of resources to get started with SQL quickly, it takes years of practice to master it. When he experienced the ease and power of BigQuery, he immediately "fell in love" and is now a big fan of the Google Cloud Platform. Even when compute power is seemingly unlimited today, writing efficient SQL is still important when you care about performance and costs. He's currently deploying Looker at a company with more than 100,000 employees. The first phase is onboarding dozens of developers as pilot users. Implementing ownership and cost monitoring is critical to have in place from the beginning. To stay sane while working from home with two kids he's lucky to have his dedicated workspace. He also builds exercise and meditation into his daily routine, and plays drums in his free time. Some of his recommendations to learn and practice SQL are "Learning SQL" from O'Reilly and the Codewars community. Find Michaël and Valentin on LinkedIn.

You need to understand the business to be successful as a data analyst. But how do you learn this? And how can you best support a business at different stages? In this episode I talk with Olivia Höwing about her experience working as a data analyst at Project A. She gets to work with many different portfolio companies and learn about different business models and tool stacks. This position offers a unique vantage point and a great learning curve for a data analyst. Key takeaways: Olivia studied business, but wanted to work more with technology. With data, she found a field that bridges this gap. Working closely with internal and external business stakeholders is key to build your understanding of how the business works, and how you can use data to support it. The right tool stack depends on the stage and the needs of a company. When you're just starting up, you might just want to plug your data sources directly into a visualization tool. (Looker Studio is a super simple and free tool to get started, especially when you're already on Google Cloud Platform.) Building a full-fledged data warehouse, joining different data sources and building more complex data models comes later, when you have more data, more complex business needs, and performance considerations. (On GCP, BigQuery is the obvious choice. It has a free tier as well. And for data modeling, dbt is ubiquitous.) While it's interesting to explore new tools, you should only bring them in if it's really needed. Every additional tool doesn't just cost money, but adds complexity. And often the job can be done with tools that are already there. (For example, if a company is happy and productive working with Tableau, there's no reason to bring in another BI tool, even if better options might be available.) Data activation is about culture and building data into your daily routine and decision making. Close relationships with your business stakeholders are key here. If you want to learn and build your skills, it can be very helpful to go to local meetups or conferences. This year, the dbt conference Coalesce is coming to Berlin. Olivia supports Datacraft, a community to help people find their way into data. Find Olivia and Valentin on LinkedIn.

How do you break into a data career? What are the skills you need to learn? Where do you find your first job? There are many different paths that people take on this journey. And in the end, we often face the same challenges. And we can learn by listening to each others stories. In this episode I talk with Cynthia Ovadje (Junior Data Analyst at 1&1 Mail & Media) about how she realized her plan to become a data analyst, and the lessons she learned along the way. I had met Cynthia some years ago, when she was just starting out in this field. And I was really happy to hear how she's found her way into her current role since then. Key takeaways: The "analyst" title is used together with many different roles. Cynthia started out as an inventory analyst, making sure the right items were in the right warehouse at the right time. Looking at job offers and applying even if you don't meet the requirements can give you valuable experience and opportunities to learn about the job. Cynthia didn't let herself get discouraged by unsuccessful applications, but instead kept learning from the feedback she got. You can learn a lot of the technical skills with online courses. After several months of learning SQL and Tableau she landed her first job as a junior data analyst. Besides technical skills, strong communication is crucial for the data analyst role. The ability to translate business needs into technical solutions, and then explain complex analysis results in clear and simple language allows you to deliver real impact. Understanding how the business works gives you superpowers as a data analyst. You can learn this by joining meetings with many different stakeholders and building relationships with people from other parts of the business. A more established company (and data team) can be a better place to start at as a junior data analyst. There are usually more resources available to support you in your professional development compared to a startup. Working at personal projects can be super helpful to showcase your skills when you don't have the experience on your CV. Bonus topic: Cynthia shares her experiences in moving to Germany from Nigeria, and the unique challenges of living and working here as a woman of color. Find Cynthia and Valentin on LinkedIn.

When business users complain about the data, that's a good sign! It means they actually want to use it. In this episode I talk with Uzi Blum (VP Data at Taxfix) about the four layers of data quality. Key takeaways: Three steps to quickly gain trust with your stakeholders: (1) Show them you understand their problems, (2) deliver results quickly (within weeks, not months), (3) focus on getting the most important metric right first. Weekly active data users (WADU) in the organization is a good proxy metric for the trust people have in data. An aspirational metric might be share of decisions taken that are based on data. Data quality can be measured by the share of incident-free days (reactive), or the share of data assets that are compliant with your quality standards, have monitoring in place, and are covered in the glossary (proactive). To ensure quality on the row layer, we can use unit testing (to cover expected cases) and monitoring (to cover unexpected cases, e.g. with Great Expectations). To discover problems before your stakeholders do, it can be effective to have a data team member on call to check data quality issues in the morning and give a "green light" when it's good to use. Having a glossary with aligned definitions of all metrics can go a long way. Ideally, this is linked to your BI tool, to help users with the right context. Guidelines for creating effective dashboards can also help with providing context (e.g. having clear titles and labels, highly visible filters and consistent color codings). For more on this, check out these blogs by Uzi and the Taxfix team: Four Shades of Data Quality Why and how we created the Taxfix glossary of terms and metrics Find Uzi and Valentin on LinkedIn.

How much time will you need for this new analytics project? You might want to underpromise and overdeliver. In this episode I talk with David Jayatillake (Chief Product and Strategy Officer at Avora.com) about scoping analytics work. Key takeaways: Scoping analytics work is hard, because it involves a lot of exploration and back and forth. Questions often evolve with the knowledge we gain about the data. Don't try to estimate effort in days, but simply group tasks by t-shirt size: S (super easy, less than 1 hour), M (we fully understand what we need to do, less than 1 day), L (effort unclear, could be days or weeks). To understand the data, we need to understand the metadata. Technical aspects like tracking implementation, lineage, freshness. But also business context such as outages, one-off events, seasonality, usual drivers of change. To make this metadata available together with the "main" data can unlock a lot of value. Analyses are never really finished, there's always a follow-up question. Great analysts anticipate this and "over-engineer" their solutions to allow stakeholders to explore more. The goal of self service is not to eliminate work for the analysts. The more data you make available, the more questions you will get. And that's a good thing! Some tools we talked about in this context: Avo (tracking plan management) Monte Carlo (data observability) Atlan (data catalog) Lightdash (Looker alternative on top of dbt) We also shared our favorite sources of inspiration: Podcasts: Analytics Engineering Podcast Data Engineering Podcast Analytics Power Hour The Right Track Blogs, newsletters: Analytics Engineering Roundup Benn Stencil Sarah Krasnik Stephen Bailey Prukalpa Mikkel Dengsøe Emily Thompson Shitposts: Pedram Navid Taylor Murphy Josh Wills Seth Rosen And of course you should subscribe to David's newsletter! Find David and Valentin on LinkedIn.

Wouldn't it be great if you had only that One Metric That Matters? So clear. So simple. Turns out life – and product management – is not so simple, and we need to navigate this complexity. In this episode I talk with Jonas Adrian Eriksen (Group Product Manager at Gorillas) about North Star metrics and data-informed product development. Key takeaways: Northstar metrics provide focus, but create the risk of tunnel vision. They're best used on a team level to align on yearly or quarterly goals. They need to be leading indicators, not lagging. On a company level, we should consider a constellation of multiple "star" metrics and need to understand their interdependencies. The best setup for product analytics is to have data analysts embedded in the cross-functional product teams. The analyst can assume the role of "chief of staff" to the product owner who understands the business problem and is proactive in defining the goal and finding solutions. Bonus topic: We talked about the "Great Resignation", and both our personal stories of why we have quit our jobs (twice) in the past months. Caring about your mental health plays a big role here. Find Jonas and Valentin on LinkedIn.

That "quick question" over Slack? Yeah, that's probably not the highest value your analytics team can deliver. In this episode I talk with Elizabeth Press (Head of Business Intelligence at Gymondo) about using DataOps to make analytics work more efficient, reliable and valuable. Key takeaways: With DataOps, we want to move away from "quick questions" over Slack and piles of ad-hoc requests or self-inflicted "emergencies". It all starts with gaining an overview of our tasks and projects (for example with a Kanban board) and establishing healthy communication patterns. We can use product thinking or design thinking to understand customer needs and context. The task of the data or analytics team is then to design good solutions. It's important to measure the quality of our work as data or analytics team. How much value are we delivering, how much are we held back by fixing broken systems? Analysts are the face of the data organization. Their strength is in understanding the data, the customer needs and context, and then telling effective stories to drive better decisions. Find Elizabeth and Valentin on LinkedIn.