When you sit down at Rick Bayless’ Topolobampo restaurant, you aren’t rushed into a menu. Instead, you start on a narrated journey. You hear about Chef’s own experience of Mexico. You are told and can read about that journey. As you go on this vicarious journey, it’s brought to life before you with each thoughtfully prepared plate and wine pairing. You understand the why before you taste the connection. You are invited to sit, enjoy, and savor the journey.
Most warehouse/analytics projects today start by picking a tool and slinging code. The story isn’t considered; future users are an afterthought, and there’s been no care in preparation.
NOTE: I’m going to use the word “warehouse” throughout this – I really just mean centralized data analytics project – warehouse/mart/lake/lakehouse/semantic model/etc. can all apply.
The other day I saw a LinkedIn post by Buck Woody, our new CDO whose first “in the office” day is today. It was a post about “small data” and some companies leaving the cloud. Since my wife is right, and my mind really does revolve around food, I thought about the slow food movement and left a comment that started as a joke but then turned into a serious thought.
I bet we all have a failed warehouse story… Here’s one of mine… For this one, eventually, the manager of the team doing the data engineering made the right call and quipped, “I’m not above canceling my own project” as he lobbied hard up the chain to have the initiative scrapped. It was a fast-food project – the tools were selected first (Oracle Warehouse Builder), the expensive consultants were brought in, some “token” planning and prework was done, but we were off to the races building ETLs without understanding the basics. Eventually, big discussions came up about assumptions and confusion. One debate was, “What is a member? What is a patient? What is a subscriber? Are these the same or are they different?” Arguments were still raging over what the data needed to be. Yet there we were in our cubes and war rooms, with dollars being poured out the window, furiously building something to someday slop down in front of the analysts, supposedly to be used in key decision-making.
What is the Slow Food Movement?
As I let this idea marinate a bit, I did some research, and it was pretty interesting. It’s not just an idea, but an actual movement. It apparently started in 1986, when a fast-food joint opened near the Spanish Steps in Rome. It started as a protest against the disappearance of local, culturally significant, authentic food traditions and seeing them replaced in place after place with industrialized, “factory-like” options. By 1989, they had a manifesto – and over time it spread to become a global movement. Check out the site and read more.
The founder of the movement Carlo Petrini has said, “everyone has the right to good, clean, and fair food.”
- Good – High in quality, flavorful, enjoyable to be savored by the guests.
- Clean – Natural, fresh, local ingredients – produced, harvested, transported and prepared with care rather than rushed and cheaped out on to squeeze margin.
- Fair – Priced right up and down the chain – for the guests at the restaurants, those involved in the trade, and the growers and suppliers. In a community where sharing and seeing all ships rise together mattered.
It’s Time for a Slow Data Manifesto
Every industry and offering is going through an enshitification process right now. That’s a whole different rant for another time, but you can find it everywhere. Maybe it’s time to push for slow data. I’m not saying we stop innovation and stop gleaning insights quickly. But we need to start slowly, build a foundation, and create good, clean, and fair data projects.
Good: It Satisfies the Users’ Needs
I went to Fallow restaurant in London recently and sat at the chef’s counter where the smells and sights were intoxicating. Will Murray, one of the chef-owners whose YouTube videos I love, came out to greet me and suggested some dishes. The waiter, Toby, was phenomenal – he listened, really listened to what I like. He brought me a splash of another wine to try, suggested one more small plate when I was ready. The food was exceptional. I can still taste the mushroom parfait 2 months later and I’d take a discounted client in London just to get that again, the confit celeriac… dishes I never thought I’d rave about by name alone. I could taste individual layers of flavor in each dish, while tasting the whole on a different level. The textures, the interactions with Toby, Myles, and Tia at their stations. I felt cared for. Unrushed. I left satisfied.
That’s what good data should feel like.
The users and those impacted by the decisions the data help them make should feel:
- Cared for – the data answers their questions, not just the ones you assumed they’d ask. Maybe it triggers more questions and ideas (like the Chapoutier Hermitage Toby knew I’d like without asking.)
- Unrushed – dashboards that don’t overwhelm and provoke anxiety and dread. Models that reveal complexity as its needed.
- Listened to – built with user needs discovered first, not dictated by what’s easy to build, fun to try, or sold at the golf course…
- Confident – they trust it because it makes sense, because someone thought about their journey, the data is accurate and checks out on cross reference.
- Satisfied – they got what they came for and need to come back.
Good data isn’t just accurate. It’s nourishing. It satisfies. Users should leave your dashboard the way I left Fallow, searching for an excuse to come back. (I’m only half-joking – if you need some on-site SQL advising in London drop me a line – and I’ll partner with you on a stupid low rate đ )
When was the last time anyone felt that way about a data warehouse?
Clean: You Know Where it Came From
On that trip to Fallow, I sat at the chef’s counter, as I try to do when I head out to a restaurant that has one. I love the tastes, smells, and sounds of a restaurant, but I love even more to watch the orchestra of the service kitchen in operation. The evidence of the mise en place and prep done hours before opening, the choreography between the stations, and the use of time. It’s such an experience to watch experts who care craft something they love.
That combination of craftsmanship, planning, and careful coordination is the difference maker our data projects need.
- You can trace the lineage – Every number, every metric – you know where it came from, what transformations it went through, what systems touched it.
- The prep work is visible– Like the mise en place in that front of house kitchen, you can see the foundation. The data models, the orchestration, the organization… It’s not just thrown together at the last minute in a big heap.
- Quality first – The entire team should have an emphasis on data quality. The commis chefs at Fallow care about balancing sustainability with quality. They reject bad ingredients from the trucks before letting them eventually wind up on a plate. We need to care deeply about quality at the beginning. Or the users will taste the issues and the warehouse is shelved by users moving back to excel.
- Built to last, not just to launch – Clean code, clear documentation, maintainable pipelines. You’re not creating technical debt that poisons the next project.
- Picture the end, but start from scratch with careful planning – We need to understand why we are building our project and what the users need – but we need to know the full layout of our ingredients. We need to build on a model and shared understanding between the different folks involved in it – or we’re just building a dozen different outcomes and the users will not use it.
We need a culture of quality, coordination and planning. With everyone on board. The full team needs to be empowered to say “STOP!” if they notice we’re missing the mark.
Fair: Customers are Protected and Teams are Empowered
I don’t have a mouth-watering intro for “fair” and I don’t think I want one anyway. Fair means realizing our work carries real stakes. Our data needs to be accurate and useful – fair to the users. Our data needs to be secure and governed – fair to the customers whose PII we store. Our data needs to be built with a plan in mind – fair to the teams who build the warehouse.
- Governance, Security, and Compliance FIRST – Data breach letters are getting pretty old to receive. These roles need to be at the table when the projects start. Before the first query is even written.
- Usability Throughout – We can’t forget why we are building this. We need to understand our users will ask questions they can’t think of yet – we need an approach that lets them self-service and discover questions. But we need a solid base that can answer the basics and make them think of the next.
- Costs and Budgets Matter – When we bring the consultants and tools in before we know why or what we are building, we’ll be spending a lot more than we ever anticipated. Know what you are building. Don’t be afraid of proofs of concept, but don’t use that phrase as an excuse to skip being good, clean, and fair.
- Shared Language across Many Sources – If we aren’t speaking the same language about our own data (our various systems already don’t, guaranteed) before we’ve finished – we’re building a project destined to be shelved. Some integration team manager will get to brag about not being above canceling their own project one day…
If we can’t build it good, clean, and fair, we shouldn’t build it at all.
The Dessert Course
I think it’s time for a slow data manifesto. We’ve made way more money at Straight Path from folks who tried it the fast-food way. It’s great to have a bias to action, and sometimes we need it, but from my vantage point after 14 years of consulting to companies in trouble, if folks just thought for even a little bit first, they would have been better off. Don’t get me wrong, every so often I want a Quarter Pounder with cheese and some greasy friesâand sometimes we have to get a quick answer out of our data. That’s fine – but the commodity approach shouldn’t be the norm. Data that really makes the change and impact on our organizations and customers needs to be crafted with care.
It needs to be good, clean, and fair. Let’s make that change start with us. We data geeks can be the force that rises against enshitification spreading from industry to industry. And maybe our data, when crafted with care, will show the executives that there’s more money to be made by doing the same with the rest of their departments than rushing and squeezing to get just another quarter percent of profit.
I’m excited that Buck is starting today – I think the CDO is a role that we’ll see as one of the most important roles in the C-Suite when we look back twenty years from now – and I love that we will help bring that to clients who maybe can’t justify one full-time.
Maybe I’m off here, and he’ll tell me why the world doesn’t need a slow data manifesto, and perhaps you’ll agree with him. I want to hear it… If I’m catastrophizing, as I sometimes do, you can tell me. But maybe it’s time to build with quality in mind from the start. Maybe it’s time for someone to start a slow data movement.
Being first isn’t as exciting or important as you think it is. Being first may get you somewhere for a little while. But being accurate, safe, trusted, and useful? That’s where true growth lies, that’s where the real recurring revenue is. We can and should do so much better. And we all have the power to push up and make our companies care about good, clean, and fair data.
Please tell me where I’m wrong or right in the comments. I’d love your thoughts. What do we do about it?
Bon Appétit!
