The business model is versioned, without duplicating your data
"The business model is drawn visually, versioned and shared between data and business teams." The word worth pausing on is "versioned": here is what it means in practice, and why it was a harder problem to solve than it looks.
The problem with naive versioning
Volume grows with time, not with the business
The simplest way to keep a history is to save a full snapshot on every update. It works, until the day it doesn't: an object that hasn't changed in six months ends up duplicated 180 times.
Infrastructure costs balloon, queries slow down, and the history becomes unusable right at the moment it starts to have value.
One version per change, not per load
Stable data costs nothing more
Every concept and every relationship in the graph is stored as a sequence of versions, each valid over a specific period. A version stays open as long as nothing changes; as soon as a real change is detected, and only then, the old one closes and a new one opens.
Volume no longer grows with time passing, but with the actual rate of business change. You can query the state of the model at any past date — "what did this product range look like six months ago?" — without ever having had to duplicate anything. It's the same data, with a memory.
Structure and content, cleanly separated
Exploration stays fast, even after years
The graph separates what describes how things are connected from what describes what they contain and since when. The structure stays light and fast to traverse, while the full history (content, versions, validity dates) lives in a dedicated layer.
What history makes reliable
A dashboard, an ad hoc exploration or an AI assistant answer all query the same versioned memory — so the same truth, on the same date, everywhere.
Knowledge that holds
A concept, a definition, a rule captured once stays traceable through all its changes, instead of getting lost in successive copies.
One truth per date
Two different tools pointed at the same date return the same answer: no more drifting copies.
Privacy follows along
Written at the concept level, rules apply automatically to every historical version, with no configuration to redo.
What it changes in the numbers
But the real benefit isn't on the bill
It's that an enterprise AI or a dashboard built on this memory can finally answer a question that is as simple as it is rigorous:
"What was true, on this date, in this context?"
No approximation, no drifting copy, and never asking the same question twice for two different tools.
See history in action on your own data
We connect a real scope and replay its state at several past dates, to measure what versioned memory changes for you.
Subscribe to our newsletter
Insight on knowledge graphs, straight to your inbox.