What Should Your Data Platform Actually Answer?
When teams evaluate a data platform, the conversation usually starts with features like connectors, dashboards, storage tiers, query speed. Those matter. But they’re the wrong place to begin. The better question is the one we ask every team that talks to us: what are the one or two questions you most need your platform to answer, and can’t answer well today?
It’s a clarifying question, because most of the hard ones share a shape. They reach across systems that were never designed to talk to each other, and answering them falls to a person, usually the most senior one in the room, putting things together by hand.
So, we have put together some of the questions worth holding on any platform.
“Where did this number come from?”
A figure appears in a report, a monthly emissions total, a production number, a service level measurement. An auditor, or an internal stakeholder asks where that number came from, end to end, along with proof of the data quality attached. How long does it take to have a solid answer?
For most teams, the honest answer is hours or days. A senior engineer has to piece everything together from the data historian, asset registers, maintenance logs, and spreadsheets, and even then, nobody is fully confident the result is right. A platform worth using lets you map out any number's entire journey in a single step. You can follow it back to the original raw signals it was computed from, seeing exactly which data points were trustworthy, which were filled in, and which were missing. When the regulators come knocking, that map is your instant audit trail.
“What actually happened and why?”
Something looks wrong. Maybe it's an anomalous reading, a sudden alarm, or a number that’s drifting. Figuring out "why" means linking that reading to the asset that produced it, the process that asset serves, recent events, the maintenance history, and even the operator on shift.
Today, that means a meeting, five different browser tabs, and hunting for the one engineer who knows how the systems connect. When looking at a data platform, the real question is: can you trace that whole story in a single click or prompt across linked relationships, following it from anomaly to asset to event to context, in one view? Or are you stuck manually piecing it together from scratch every single time? Because that’s the difference between solving the mystery in minutes or wasting days.
“What does this new source change?”
Operations never stand still. Whether it's a new sensor, a new upstream system, or a new regulatory requirement, every update expects to be integrated. The real question is how long it takes to go from “we need this” to “it’s flowing into reports and ready to use.”
If the answer is months, with rounds of rework every time, you don’t have a shared model. You just have a setup where every new source requires starting a whole new project. The platform you actually want lets you add a new source to your asset model just once, so every downstream report inherits the context automatically.This cuts integration time from weeks to hours, meaning your company's growth is no longer limited by how fast you can manually wire things together.
“Are we all working from the same picture?”
Here is the real question underneath it all: does a “customer,” an “asset,” or a “KPI” mean the exact same thing across your entire company?
When definitions live in isolated files and individual memories, you end up with two people bringing completely different numbers to the table. Instead of making decisions, the meeting turns into a debate over whose data is right. Having a shared model means using one single definition everywhere, and that is what makes the first three questions answerable at all and with it,real collaboration possible.
The category, and a different way to buy it
The questions above describe a very real solution with a real name: an operational data platform. It provides one shared model of your assets, the processes they run, the business context around them, and every single signal and event they produce, all while tracking how every value is calculated (Data lineage). Big enterprise vendors have sold this kind of setup for a decade. The catch is that it usually comes with a seven-eight-figure price tag, an 18-month rollout headache, and storage that locks you in for good. Because of that, many teams took one look, assumed it was out of reach, and went back to answering these tough questions by hand.
IntelliStream DataHub belongs in that same category, but we do things entirely differently. We are fully open source under AGPL-3.0, so you can read the code, run it completely air-gapped, or even fork it if we ever disappear. Because it is built on standard tech, you can deploy it in your own environment or use our own multi-petabyte Nordic infrastructure. Best of all, we routinely cost about 10 times less than the legacy giants.