From Dashboards To Decisions: Why Business Intelligence Needs A Reset

Abeer Shrivastava is a data analytics and AI leader at Tredence Inc.

getty​”Data is a precious thing and will last longer than the systems themselves,” Tim Berners-Lee said. Most companies I’ve worked with have kept the data. Far fewer can use it to decide anything faster.

After 15 years in business intelligence, I believe most companies don’t have a visibility problem. They have a decision problem, and it comes from three misalignments: metrics nobody agrees on, data with no context for action and no one who owns trust in the numbers. More dashboards fix none of them.

At two large tech companies I worked at, I watched finance, marketing and operations bring different revenue figures to the same executive meeting. Stakeholders were used to pulling numbers straight from source systems, and those figures didn’t match the centralized lakehouse behind the executive dashboards. Each leader defended their own number, and nobody could say which one was wrong. The meeting ended with an action item to “reconcile the data,” and the actual decision waited another week. Nobody lacked data. The meeting went to deciding whose number was right, and leadership left frustrated.

What I took from it: The discrepancy wasn’t a pipeline bug. Each number was correct for the system it came from. What was missing was a decision about which definition the company would stand behind, and no engineer could make that call alone.

Tools can’t settle this. Someone has to decide what “revenue” or “active customer” means, write it down and make every team use it. A semantic layer, such as Looker’s modeling layer, enforces that decision. It can’t make it.

Takeaway: Before building anything new, ask three teams to define your most-quoted metric. If you get three answers, fix that first.

A report can show that margins fell last quarter. It can’t tell the plant manager what to change on Monday. The reports that work put the metric beside the action: the threshold, the owner, the next step. At a BFSI client, flagging high-net-worth customers who showed signs of leaving gave leadership something to act on, and attrition in that segment fell 20% in relative terms, keeping billions of dollars in assets with the firm.

Takeaway: For every dashboard, name the decision it supports and who makes it. If you can’t, retire it.

At a retail client, the enterprise data warehouse was nearly abandoned. Stakeholders and power users didn’t trust the data, so they went back to building their own reports. The breaking point came when two regional teams presented conflicting sales numbers in a review, and both said the warehouse was the reason they’d stopped relying on it.

My instinct as a data leader was to fix quality from our side. We could have kept auditing pipelines and tightening checks, and honestly, that was my first plan. But the users weren’t saying the data was wrong. They were saying they couldn’t tell whether it was right, and no amount of testing on our end could answer that for them. Some on the team worried that handing certification to the business would slow everything down, and some stakeholders weren’t sure they wanted the accountability. We had the stakeholders certify the datasets themselves anyway. They became accountable for the data and not just consumers of it. Active users grew 36% within six to nine months and kept climbing.

What I learned: We’d been treating trust as a quality problem, and it was an ownership problem. People trust what they’ve signed off on. Once the business put its name on the datasets, the data was theirs, and they defended it instead of working around it.

Takeaway: Trust can’t be delivered to people. Give the business ownership of certifying the data they depend on.

I think the future of reporting is conversational, and I’m implementing it now at a large retail client. Leaders are already asking for AI agents they can chat with instead of static dashboards. I’d go further and say most teams should cut their dashboard count in half, and many should eventually retire the rest.

But conversational BI is where the three misalignments get expensive. A chatbot on top of conflicting definitions, context-free metrics and datasets no one owns gives you confident wrong answers instantly. The same foundations decide whether it works.

Takeaway: Before you invest in AI agents, check whether your metrics, context and ownership could survive a question asked in plain English.

The companies that pull ahead won’t be the ones with the most dashboards or the most AI. They’ll be the ones where everyone works from the same numbers and knows who answers for them.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?