The Data-Driven Lie: Why Most Companies Fail at What They Claim to Do Best
Everyone claims to be data-driven. Almost no one is.
Swarnim Shrey
Founder, MindPalace
Open any company's investor deck, career page, or LinkedIn post. You will find it somewhere.
"We are a data-driven organization."
It has become table stakes. Nobody admits to being gut-driven or vibes-driven or loudest-voice-in-the-room-driven. Data-driven decision making is the only acceptable answer.
And yet.
Walk into the average leadership meeting. Watch what actually happens.
The meeting that gives the lie away
Sales says revenue is up 12 percent. Finance says it is up 8 percent. Marketing has a third number. Someone pulls up a dashboard. Someone else pulls up a different dashboard. The conversation drifts. A decision gets made based on who argued most convincingly.
That is not data-driven. That is data-adjacent.
We hear this same script from teams we talk to. Different industries, different stages, same conversation. Everyone has data. Nobody agrees on it. The decision still gets made by whoever talks loudest.
The infrastructure fallacy
Companies have spent the last decade building data infrastructure. The cloud warehouse. The ELT pipelines. The BI tools. The semantic layer. The data catalog. The reverse ETL. The metrics store. The AI copilot.
The Modern Data Stack became a badge of honor. If you had Snowflake and dbt and Looker and Fivetran, you were serious about data.
Infrastructure is not capability.
Having a warehouse does not mean you can answer questions. Having dashboards does not mean you can make decisions. Having data does not mean you are driven by it. Most companies confused the inputs for the outcome.
What "data-driven" was supposed to mean
The promise was simple. When a decision needs to be made, data informs it. Not politics. Not intuition. Not the CEO's gut. Data.
That requires three things.
1. The right data exists and is trustworthy. Everyone agrees on definitions. Revenue means the same thing to Sales and Finance. Churn is calculated consistently. The numbers are accurate. The KPI structure is MECE: no overlaps, no gaps.
2. The right people can access it without friction. A VP can answer their own question without filing a ticket. A regional manager can drill into their numbers without waiting three days. The data team is not stuck as the Human API.
3. The data connects to the decision. It is not just "revenue is down." It is "revenue is down because returning customers dropped because retention fell because onboarding completion cratered, and here is who owns each piece."
Most companies have achieved some version of #1, with caveats. Almost none have achieved #2 or #3.
Where it breaks down
The trust problem
Ask a simple question: "what was our revenue last quarter?"
In most organizations, you will get multiple answers. Not because anyone is lying. Because definitions differ and structures are not MECE.
- Gross vs net
- Booked vs recognized
- Including vs excluding refunds
- This product line vs all product lines
Finance has a number. Sales has a number. The dashboard has a number. The board deck has a number.
When leadership cannot agree on what happened, how can they possibly agree on what to do?
The access problem
Data exists. It lives in the warehouse. It is theoretically available. In practice, getting an answer requires:
- Knowing the right question to ask
- Knowing who to ask
- Waiting for the data team to prioritize your request
- Interpreting the output correctly
- Hoping the definitions match what you expected
This is why data teams spend 80 percent of their time on ad-hoc requests. The data team became the Human API. The bottleneck between data and decisions.
Some teams try to fix this with AI chatbots. AI without a grounding layer just hallucinates faster. Confidently wrong instead of helpfully uncertain. We covered that failure mode in why LLMs should never calculate your churn rate.
The context problem
This is the deepest failure.
Even when the data is trustworthy and accessible, it rarely connects to the decision. A dashboard shows that revenue is down. It does not show why. It does not trace the causal chain from North Star to driver to root cause. It does not tell you who owns the broken metric. It does not suggest what to do.
Dashboards show you what happened. They do not show you what it means.
A conversation we hear often, lightly anonymized
"Revenue is down 4 percent this quarter."
"Why?"
"We are looking into it."
"By when?"
"We have queued the analysis. Probably end of next week."
By the time the analysis arrived, the quarter was over. The cause turned out to be a single broken onboarding step that had been live for 17 days. Nobody noticed for 17 days because no dashboard was built to surface it. They were built to show what happened, not to flag what was breaking.
What it actually looks like: a Decision Context Graph
Closing this gap needs a different primitive. Not a better dashboard. A different structure entirely. That structure needs to capture three things:
- Relationships. How do metrics connect? What is the causal chain from North Star to drivers?
- Ownership. Who is accountable for each metric? Not the team. The person.
- Traversability. When something breaks, you can walk from symptom to cause to owner without guessing.
We call this structure a Decision Context Graph.
A Decision Context Graph is a living map of your business that captures what dashboards do not. Every metric connected to your North Star. Every number with a clear owner. Every relationship explicit and traversable. Every insight traceable to root cause.
When Revenue drops, you do not open five dashboards and start guessing. You traverse the graph. The causal chain is visible. The owner is clear. The decision is obvious.
That is data-driven. Not data-available. Not data-displayed. Data actually driving the decision.
This category is starting to get called the missing infrastructure layer for AI. Foundation Capital recently argued it could be one of the largest software opportunities of the next decade. We agree, but not because of hype. It is the missing layer that makes everything else work.
The path forward
The next generation of data-driven companies will not just have dashboards. They will have a Decision Context Graph. They will know how every metric connects to the North Star. They will know who owns every number. They will trace causality instead of guessing. They will make decisions based on structure, not synthesis.
That is what data-driven decision making was always supposed to mean. We are building the infrastructure to make it real.
If your leadership meetings keep arriving at the same disagreement, take a look at the product or read more about what we mean by Decision Intelligence.
Read this next
What is a Decision Context Graph? An Architectural Guide
A Decision Context Graph is the missing layer between your warehouse and your decisions. Here is what it is, how we build one in four hours, and why it matters now.
Why LLMs Should Never Calculate Your Churn Rate
Most AI-native BI tools let a language model calculate your business metrics. That is a category error. Here is the architecture we built instead, and why.