So you want to start a BI startup - read these first.
The Complete Guide to What's Wrong with BI Tools (For Builders).
Introduction: How it actually looks like
I've been on every side of the BI tools equation: I co-founded a BI startup that crashed and burned in 2011, spent time as a data engineer fighting with every tool you can name, and was Head of Marketing at Arch.dev bringing BI tools to market.
Here's what I've learned: If you're building BI tools today, you're fighting three different battles simultaneously. Most tool builders don't realize this, which is why most BI tools fail to deliver real value despite being technically impressive.
The Ghosts of the Past are still dragging your product down. The modular "Modern Data Stack" mentality created scattered tooling that brings joy to data teams but frustration to companies. This mindset still infects how you think about product architecture today.
The Ghosts of Today demand your immediate attention. The fundamental misalignment between what BI tools do and what decision-makers need has never been more critical. Meanwhile, everyone's rushing to add AI without understanding why current approaches fail.
The Ghosts of Tomorrow will determine whether your company survives the next five years. Platform expansion strategies, open source business models, and positioning decisions made today will either create defensible moats or leave you vulnerable to disruption.
I'm going to walk you through each of these battles, using lessons I've learned the hard way across a decade in this space. Consider this your field guide to the minefields ahead.
Let's exorcise these ghosts, one by one.
NOTE: This is a collection of my writing and you SHOULD read the original posts, but I’ll do my best and pull out the relevant parts so you can choose yourself which ones you’ll want to read in detail.
Ghosts of the past - the MDS modularity trap
The Problem: "Modularity can be dangerous. Try to consolidate your MDS; modularity leads to scattered stacks that bring joy to the data team but not to the company that employs the data team."
From "Breaking Down the Modern Data Stack":
Key Insight: The MDS created a mindset where every problem needs a separate tool. This modularity obsession still influences how builders think about product design.
The Damage:
"You can use Snowflake + dbt (self-hosted) + Meltano and top it off with Tableau. Ended up with a mix of DIY, SaaS, and lots of (usually well-working) integrations. And you can do so much more… Question, should you?"
Companies end up with 10+ tools that technically work together but create operational nightmares
Each integration point becomes a failure point for actual users
What This Means for Builders:
Stop thinking in terms of "best of breed" components
Start thinking about complete user workflows
Consolidation beats modularity for end-user experience
"Try to consolidate your MDS; modularity leads to scattered stacks that bring joy to the data team but not to the company that employs the data team"
Supporting Evidence:
Nordic data teams converge on simplified stacks: "dbt, the data build tool, used by 50% of the companies; BigQuery, the data storage system, used by 40% of the companies; Looker, the BI tool, is used by 30% of the companies"
Successful companies choose community size and ecosystem over modularity. You want to be the one that is chosen.
Ghosts of today - AI and the misalignment of the BI industry
The Fundamental Misalignment Crisis
The Core Problem: "The discipline of BI isn't about tabular data; it's about making better business decisions through data. There's a clear distinction between the discipline of BI and the industry of BI tools."
From "What's Wrong with BI":
Wrong 1: Building for the Wrong Users
"BI tools are made for analysts, the people inside the company who work with data, and the bridge between decision-makers and the technology... A BI tool should be one of the many tools that help to achieve the goals of the discipline of BI. A BI tool, thus, should help to drive better decisions through the use of data."
The Evidence:
You have a "negative feedback loop of analysts" where tools → analysts → more tools for analysts
Decision-makers are the ones who actually need the insights, but they can't use your tools
"Every self-serve analytics solution I've seen so far could be rebranded as 'self-serve analytics for analysts'"
Wrong 2: Working with Wrong Data
"Almost the entire BI industry focuses on providing and working with structured data. Yet it is also clear that 90% of the data that is generated this year, or the next, or the one after that, that could be available to your company is unstructured, event-based (time series), and real-time."
Wrong 3: Wrong Interface and Medium
"BI tools have come to mean two things: reports - a fancy word for tables with a few filters really, and dashboards, a fancy word for pie charts and graphs. I've heard millions of people say that Apple designs amazing products, and yet I've never ever heard anyone commenting on the great design of BI tools."
What to Do Instead:
"Get your data as close to decision-makers as possible; think about embedded BI, reverse ETL, and data activation"
"Don't try to build out central infrastructure at all; try to have the data in the hands of decision-makers from day one"
The AI Integration Trap
The Problem: "BI and generative AI still don't mix well. Generative AI is starting to creep into BI but not in the way we need it to."
From "BI and generative AI still don't mix well":
The Cursor vs Loveable Problem:
"The big challenge of BI is to build the loveable.dev, not the cursor... While Cursor enables the existing coders to become 10x or 100x coders, loveable takes all that freedom away to enable EVERYONE (but coders!) to deliver great software products."
Current AI Approaches Fail:
"GoodData's implementation resembles a cursor for data engineers and analysts. It helps the 1% of people who already create dashboards work marginally faster. But what about the other 99% who should be working with data but can't due to technical barriers?"
The Real Opportunity:
"I want an AI version of a product designer married to a sales person and data analyst. I need a system that helps me think through the right decisions, identify meaningful metrics, and design experiments worth running."
Supporting Evidence from "What people are not seeing about TimeGPT":
"What people are not seeing about TimeGPT: TimeGPT is amazing news! Most people dismiss two important things: It's not just about time GPS; time series models are becoming available at large. Timed data is becoming more important; just because it wasn't in the past doesn't mean it won't in the future."
Ghosts of tomorrow - prepare for these
The Platform vs Point Solution Dilemma
From "Why dbt Labs acquired Transform":
The Expansion Pattern:
"Tech companies start by targeting a tiny specific customer segment. They try to find a pond they can be the biggest fish in. For the company dbt Labs that was the 'analytics engineering pond' they created from scratch."
Stage 2 Dangers:
"They build up enough momentum to target a second larger segment (that includes the first one!) and leverage something from their first segment to get into it. For Dbt labs, it is the fact that many companies have mixed data teams consisting of both analytics engineers and data scientists or machine learners."
The Positioning Confusion:
"Dbt Labs changed its product messaging about a year ago, away from analytics teams and engineers towards data teams and workflows. In addition, they specifically added ML models and operational analytics into their messaging."
From "Why Preset.io Is Playing Catch Up And How They Could Reverse The Game":
The Enterprise Trap:
"For almost two years, Preset has pushed out features that serve the enterprise segment almost exclusively. That's generally fine but isn't differentiated. Preset is playing catchup with Tableau and other tools on the market."
The Wrong Pond Problem:
"Preset is fishing in the big pond of 'enterprise customers' without any boundary applied. But the problem is, in this market, preset will never be a big player, and as such, preset simply cannot win the segment."
The Cheap Segment Distraction:
"Preset is also fishing in a second segment, that of startups with advanced data engineers. Why is that a bad segment? Because it's cheap! There's no money to be made with startups... But you can't even focus on the big fish when you're busy tendering small fish."
From "Airbyte vs Meltano Analysis":
"Both tools still are not quite there yet if the target customer is a mature data engineering team. Meltano, with the vision of being a DataOps OS should be just that: Easier to use than hand-coded anything. Apparently, it is not (yet)."
Key Lessons:
Expansion often dilutes core value proposition
Platform visions often exceed execution capabilities
"Analytics vs data" positioning matters more than you think
Competing in enterprise without differentiation = playing catch-up forever
The Open Source Business Model Minefield
From "Airbyte's wrong turn":
The Core Mistake:
"Airbyte rightly realizes that nailing the 'connector' problem (or data snowflake problem, as I like to call it) is key to the data integration space. But the fear of big companies providing a hosted Airbyte solution drives them in the wrong direction."
The Platform Incentive Problem:
"If you as a company want to have as much 'connector contribution' as possible, you'll have to create strong incentives for the developers of the connectors. That means as Tobi Lütke of Spotify puts it, to leave all the money on the table and give it to the developers."
The Real Strategy:
"That in turn means, getting as many consumers onto your 'platform' as possible, and that means, getting widespread adoption! Widespread adoption is if big companies come and host your product. Because you then get a lot of spreading for free."
From "The Power of Perseverance: Starburst's Journey" (Success Story Contrast):
What Success Looks Like:
"Technology and science are key differentiators for data companies. The success of both companies Borgman founded is deeply routed in his continuous search to turn cutting-edge technology and science into businesses."
The Perseverance Factor:
"Perseverance is key to building companies based on deep new technology. Turning cutting-edge technology into businesses often relies on markets adapting and technology advancing further."
From "TimescaleDB COSS Analysis":
"Before commercializing, you'll need broad adoption. Before commercializing, you'll need prime credibility."
Key Principles:
Community growth beats short-term revenue protection
Platform effects require giving up control
"Commoditization" fears often prevent platform success
Developer incentives must align with platform goals
Deep technology + perseverance beats incremental improvements
Synthesis: what this means for you
If you're building BI tools today, here are the hard-learned lessons from analyzing dozens of companies:
1. Fix the Fundamental Alignment First
Build for decision-makers, not analysts. The entire BI industry has this backwards. Every "self-serve analytics" solution is really "self-serve analytics for analysts." Until you solve the decision-maker problem, you're just adding to the noise.
2. Answer the One Critical Question
From Starlake's experience: "Give me an example of a company that uses this and is super stoked about it. Why are they so stoked? How is this new approach 10x better?" If you can't answer this instantly, you don't have product-market fit. Everything else is distraction.
3. Pick the Right Pond
Don't fish where you can't be the biggest fish. Preset failed by targeting the "big pond of enterprise customers" where they'd never dominate. dbt Labs succeeded by creating their own pond ("analytics engineering") where they could be #1. Choose markets where you can win, not where you want to compete.
4. Your Biggest Competitor is DIY
More often than not, your competition isn't other vendors—it's customers building their own solutions. This is especially true in open source, where experienced engineers can just deploy your tool themselves. Plan for this reality from day one.
5. Consolidation Beats Modularity
The Modern Data Stack created "scattered stacks that bring joy to the data team but not to the company." Users don't want to manage 10+ tools. They want complete workflows. Stop thinking "best of breed" and start thinking "complete user journey."
6. Don't Add AI to Add AI
Current AI integration approaches serve the existing 1% better, not the 99% who should be using data tools. Focus on democratizing analysis, not optimizing expert workflows. Build the "Loveable" for everyone, not the "Cursor" for existing power users.
7. Platform Strategy Can Kill Focus
Expansion often destroys what made you successful. dbt Labs, Preset, and others diluted their core value proposition chasing larger markets. The "analytics vs data" positioning shift matters more than you think. Stay focused until you dominate your first market.
8. Get Open Source Incentives Right
Community growth beats short-term revenue protection. Airbyte's licensing mistakes show what happens when you optimize for revenue capture over ecosystem growth. "Leave all the money on the table and give it to the developers" if you want platform effects.
9. Deep Technology + Perseverance Wins
Starburst succeeded because they built on genuine technological breakthroughs, not incremental improvements. "Technology and science are key differentiators for data companies." Turning cutting-edge technology into businesses requires perseverance as markets adapt.
10. Solve Specific Pain, Not Generic Value
Avoid "generic data fluff" like "drive better decisions" or "central hub for analysis." Mode and GX failed because they couldn't identify the specific elephant in the room. Find the one painful problem you solve exceptionally well, then build everything around that.
11. Bridge the Data Engineering Gap Smartly
There's a real gap between data producers and data users. But the solution isn't always more tooling—sometimes it's organizational. Focus on whether your approach actually closes this gap or just adds another layer.
12. Time Your Market Entry
Some markets are entering "fast-moving water" where growth accelerates rapidly. Time series data, real-time applications, and democratized AI analysis are examples. Position yourself where the current will carry you, not where you have to swim upstream.
The Bottom Line: The companies that understand these lessons won't build faster horses for the existing 1%—they'll build cars for the other 99%. Most BI tools fail not because they're technically inferior, but because they solve the wrong problem for the wrong people at the wrong time.
Start by asking: Who desperately needs what you're building? Why is your approach 10x better? Can you dominate this specific market? If you can't answer these clearly, you're not ready to build.
This guide navigates through my key insights on BI tools. For full context and details, read the original articles linked throughout.
Complete Reference: All Articles Covered
Core BI Industry Critiques:
"What's Wrong with BI" - The fundamental misalignment between BI discipline and BI tools
"BI and generative AI still don't mix well" - Why current AI integration approaches fail
"Breaking Down the Modern Data Stack" - The modularity trap and its consequences
Company Strategy Failures:
"Why Preset.io Is Playing Catch Up" - Enterprise positioning mistakes and market selection
"Dissecting What Makes a Data Strategy Fail" - Mode and GX strategy analysis
"Airbyte's wrong turn" - Open source licensing and platform strategy mistakes
"Why dbt Labs acquired Transform" - Platform expansion strategy analysis
Tool Comparison & Market Analysis:
"Airbyte vs Meltano Analysis" - Platform vision vs execution reality
"Measuring data teams, dagster & airbyte & meltano" - Tool evaluation frameworks
"dbt snapshotting and technical analysis" - Technical implementation insights
"Nordic Data Architecture analysis" - Market consolidation trends
Business Model & Strategy:
"The Power of Perseverance: Starburst's Journey" - Success story of deep technology + perseverance
"TimescaleDB COSS business models" - Open source commercialization principles
AI & Technology Trends:
"What people are not seeing about TimeGPT" - Time series AI and its implications for BI