A multibillion-dollar enterprise with a direct sales force spread across three continents, selling into a market that had just absorbed a demand shock. The CEO needed to know, with certainty, what was actually in the pipeline. Not what reps said. Not what the CRM claimed. What was actually there, backed by signals.
The problem no one wanted to name
The revenue operations team produced a forecast every week. It went to the board. It went to the CEO. It drove hiring decisions, inventory planning, and investor communications.
And it was wrong. Consistently. Not by a little.
The forecast carried a 40% variance in the final month of each quarter. In a $2B revenue business that is not a rounding error. That is a credibility crisis waiting for a trigger. The CFO knew it. The CRO knew it. But the system that produced the number was so entrenched, so politically protected, that nobody wanted to be the person who said the quiet part out loud.
I was brought in to build "better dashboards." That was the brief. What I found was a data architecture duct-taped together over six years by five different owners, each with their own conventions, their own definitions of "opportunity stage," and their own quietly hacked SQL that nobody else understood.
The architecture of distrust
The CRM had one set of opportunity data. The billing system had another. The product usage database, which was supposed to signal buying intent, lived in a separate cloud environment the BI team could not even access directly.
Reconciling the three sources took a full-time analyst two days every week, and even then the output was greeted with suspicion. Sales leadership trusted their spreadsheets over the official BI. Finance built shadow models. The CEO asked both teams for numbers and went with his gut when they conflicted.
This is what happens when data infrastructure grows organically without an architectural north star. Everyone builds their own reality, and eventually the organization loses the ability to have a single conversation about what is true.
Building the provenance layer
I started with the data model. Not the dashboard. The dashboard is the last 5% of the work; start there and you are painting over rot.
The goal was a single source of truth for pipeline health. That meant:
- Unifying CRM, billing, and product signals into one analytics warehouse.
- An immutable event log of every opportunity touch, stage change, and forecast commit.
- Derived tables that captured not just what was in the pipeline, but how it got there.
The key insight: most pipeline visibility problems are not math problems. They are provenance problems. People do not trust the forecast because they cannot see the chain of custody. When a number changes, they cannot trace why.
So we built lineage into every derived metric. If the Q3 forecast moved, you could drill to the exact opportunity, the exact rep, the exact email thread that triggered the change.
The resistance
The technical build took three months. The cultural build took six.
The first version of the dashboard was accurate. It was also rejected. Sales leadership looked at it and saw a threat: a system that exposed the blind spots in forecasting methods refined over years of tribal knowledge.
This is the hardest part of data leadership. You are not just building systems. You are redistributing authority, and people who hold authority under the old system will fight to keep it.
I spent more time in one-on-ones with department heads in those six months than in the previous two years combined. We did not launch company-wide. We launched with one regional sales manager who was fed up with the old process. He became the internal advocate. Then another joined. Then another.
By month six the CRO was using it in executive standups. By month nine the board was asking for it directly.
The outcome
Forecast variance dropped from 40% to under 12% in the first full quarter after adoption. That translated into eight figures of working capital efficiency, because the company could finally stop hedging against its own uncertainty.
But the number that mattered more to me: for the first time in years, Sales and Finance presented the same forecast in the same meeting. And both teams believed it.
That is what a single source of truth is supposed to do. Not just aggregate data. Restore the organization's ability to trust itself.