Data Strategy

Self-Service Analytics Promised to Put Data in Everyone's Hands
It didn't. But I think we're finally getting it right.

I was a Tableau certified trainer in the thick of the self-service analytics movement. I believed in it completely. Here's what I actually learned from training people across manufacturing, healthcare, and aerospace, and why AI finally changes the picture.

6 min read Data strategy Point of view

The bet the industry made

Around 2010, the data world made a big bet. The idea was that you could put analytics tools directly in the hands of business users and get rid of the bottleneck. No more waiting three weeks for IT to build a report. No more translating your business question into a ticket that came back with something completely different. Tools like Tableau, Qlik, and later Power BI were going to change all of that. Drag a field, drop it on a canvas, see your data. Everyone becomes their own analyst.

It was a compelling vision. I believed it. I was inside it.

I was selling it, and I was all in

I spent years as a certified Tableau partner running training for organizations across the country. Big companies, small companies, manufacturing plants, hospital systems, aerospace. I was certified, I was teaching, and honestly I was drinking the Kool-Aid. The product was genuinely exciting. When you watched someone build their first dashboard and see their own data come to life in a way they couldn't before, it felt like a real breakthrough.

But I kept seeing the same thing happen. The training would go well. People would leave feeling capable. And then a few months later, the conversations would start. The dashboard isn't quite right. Can you help me add this calculation? The data looks off but I'm not sure why. And almost always: this is harder than I expected.

Tableau was really easy, until it wasn't. And for most business users, "until it wasn't" came pretty fast.

The wall that everyone hit

Here's what I came away thinking after years of training: we didn't have a tool problem and we didn't have a data problem. What we had was a time problem.

Tableau is genuinely one of the best products ever built in the analytics space. I've been using it for probably ten years now and there are still things that make me stop and think. And that's me, someone who has logged thousands of hours in the product. We can't expect a buyer, a planner, a sales VP, or an operations manager to get anywhere near that level of fluency. It just isn't realistic.

Tool fluency takes real time to develop

Knowing a tool exists is not the same as knowing how to use it. Getting confident in Tableau means months of consistent practice. A two-day training gets you started. It doesn't make you fluent.

Data literacy is a separate skill from tool skill

Before you can build anything meaningful, you need to understand your data structure. What's a dimension vs a measure. How joins work. Why two fields that look the same produce different results. That's a layer of knowledge most business users don't have and shouldn't need to have.

Skills go stale fast when you don't use them constantly

A buyer or ops manager might open Tableau once a quarter to answer a specific question. That's not enough to stay sharp. The tool gets rusty faster than they can relearn it. So they stop trying and go back to asking the analyst.

None of this is a criticism of Tableau. It really is the easiest sophisticated analytics tool that's ever been built. The issue is that "easiest sophisticated analytics tool" is still a pretty high bar for a purchasing manager who has twelve other things to do today.

What self-service 1.0 actually looked like vs what AI makes possible

The promise of self-service and the reality of self-service were two different things. The promise was: anyone can get answers from data. The reality was: anyone with significant training and consistent practice can get some answers from data, up to a certain level of complexity.

Self-service analytics: then vs now
Self-service 1.0 (2010–2020)
1
Learn a new tool: drag, drop, connect data sources
2
Understand data types, relationships, and how your ERP structures things
3
Learn calculated fields, LOD expressions, table calculations for anything complex
4
Build and maintain your own visualizations over time
5
Hit a wall, go back to the analyst, and repeat
Self-service 2.0 (AI-powered)
1
Know what question you want answered
2
Ask it in plain English
3
Follow up if the answer raises more questions
4
Done

AI finally changes the equation

I'm genuinely optimistic right now in a way I haven't been in a while. And I say that as someone who went through the hype cycle once and came out the other side with a clearer sense of where the real limits were.

What's different now isn't that the tools are smarter in a feature sense. It's that the interface is fundamentally different. Business users don't need to learn a new tool. They already know how to use the interface, because the interface is English. Ask a question, get an answer. Ask a follow-up. Drill in. That's it.

I'm excited to see a wave of tools built around this idea. At Marquis Data, we've been building Marquis IQ with exactly this shift in mind. Our goal has never been to teach business users a new tool. It's to make sure the data underneath whatever tool they already use is clean, connected, and ready to answer the questions that matter to them.

Self-service analytics finally makes sense when you don't have to learn anything to use it. That's what AI unlocks.

This feels like 2007 to me

There's an analogy that keeps coming to mind. When Steve Jobs introduced the iPhone in 2007, he said it was going to use one of the oldest pointing devices in history. One that everyone is born with ten of. Your finger.

Before the iPhone, smartphones existed. They had powerful features. But to use them well you needed a stylus, a keyboard, and a manual. The barrier to the power wasn't the power itself. It was the interface in front of it.

That's exactly where analytics has been for the past fifteen years. The power was there. Tableau, Power BI, Qlik, these are genuinely capable tools. But the interface required too much of people who just wanted an answer. AI is the finger. It removes the interface problem entirely.

The people who needed data to do their jobs better never stopped needing it. They just couldn't get to it reliably without help. Now they can.

I'm not just optimistic. I'm excited.

At Marquis Data, our mission has always been "helping people make better decisions, faster." That's not a tagline that came out of a branding session. It's what I've believed since the start of my career in this space.

What makes this moment different is that we're not working in a vacuum. Excel, Tableau, Power BI, Claude, OpenAI. The ecosystem of tools available to business users right now is genuinely remarkable. And the companies building those tools are serious about the same thing we are: getting data into the hands of the people who actually make decisions. We're not competing with any of them. We're complementing them.

Marquis IQ is a data platform built for decision makers, by decision makers, to support decision making. We handle the foundation: the integrations, the master data quality, the enrichment. Whatever tool your team reaches for, an Excel pivot, a Tableau dashboard, a Power BI report, or a plain English question to an AI, it gets better answers when the data underneath it is clean and connected. That's what we build.

Self-service analytics fell short the first time because nobody fixed the foundation first. That's the problem we've been solving. And now that the tools on top of that foundation are finally where they need to be, I'm more optimistic about this industry than I've been in a long time.

It's a genuinely good time to be in data. We're just getting started.

Marquis IQ · The foundation that makes AI analytics work
AI tools need clean data underneath them. That's what we build.

AI analytics tools are only as good as the data they sit on. Marquis IQ handles the integration, master data quality, and enrichment layer for manufacturing companies so when your team asks a question, they get an answer they can trust. Not a number that requires a follow-up call to the analyst to verify.

Certified ERP connectors for Epicor, Dynamics, Sage, Infor, SAP, and more
Master data enrichment that creates one clean record per customer, supplier, and item
IQ Insights: AI-generated intelligence built directly into your operational data

FAQ

Questions we hear from manufacturing teams

The honest answers when companies ask about self-service analytics and what's actually changed.

Why did self-service analytics tools like Tableau fail to democratize data?

They didn't fail because they were bad products. They fell short of true democratization because business users still needed significant tool fluency, data literacy, and sustained practice to use them well. A buyer or operations manager who opens a dashboard tool once a quarter can't stay proficient. Getting genuinely good at Tableau requires consistent use over months, not a two-day training. True self-service means no new tool to learn at all, which is what AI-powered analytics is now delivering.

What is self-service analytics and why does it matter for manufacturing?

Self-service analytics means business users can answer their own data questions without waiting on a BI analyst or IT team to build a report. For manufacturing companies this matters a lot, because the people who most need data insights, ops managers, buyers, pricing leads, and plant managers, are rarely the same people who know how to build dashboards. When those users can ask a question and get a reliable answer on the spot, decisions happen faster and with better information. AI analytics tools are making that possible in a way that drag-and-drop BI tools couldn't.

How is AI-powered analytics different from traditional self-service BI?

The fundamental difference is the interface. Traditional self-service BI replaced IT-built reports with drag-and-drop tools that users could theoretically operate themselves. The problem is that drag-and-drop is still a skill that takes real time to develop. AI-powered analytics uses natural language as the interface. Users ask questions in plain English and get answers. There is no new tool to learn, no training to stay current, and no wall to hit when the questions get more complex. The AI handles what used to require a power user.

Does AI analytics still require clean data to work properly?

Yes. AI analytics tools are not a shortcut around data quality problems. They make the interface easier but they don't fix bad underlying data. A natural language query against messy, inconsistent ERP data will still produce unreliable answers. Data integration, master data quality, and enrichment all need to be in place before any analytical tool, AI or otherwise, can produce results your team will trust. What AI changes is the interface layer, not the data layer.

What's the right way to think about AI analytics for a mid-market manufacturer today?

Think of it in two layers. The foundation layer includes data integration, master data quality, and enrichment. This needs to be built correctly and maintained over time, and it's best sourced from a specialized data platform built for your industry. The analytics layer sits on top of that foundation. This is where AI tools shine: giving buyers, planners, ops managers, and executives the ability to ask questions and get reliable answers without learning anything new. Get the foundation right first, then let AI do what it's genuinely good at.

See what AI analytics looks like when the data underneath it is actually ready

Bring your ERP stack and your biggest data frustration. We'll show you what an industry-aligned data platform makes possible.