Your BI tool shows state beautifully. It still can't decompose a change — and that gap is where your afternoon goes.
Dashboards are great at one thing: showing you the current state of the world, sliced however you like. Revenue by region, conversion by channel, churn by product. So when a number moves and someone asks why, it feels like the dashboard should answer. It can't — and it's worth being precise about why.
State vs. the decomposition of a change
A dashboard renders levels: where each metric stands, period over period. "Why did it move?" asks for the decomposition of a delta: how much of the change each factor — and each combination of factors — is responsible for. Those are different computations.
A line that dips tells you that something moved. It can't tell you which factors drove the move, in what proportion, summing to the whole. Filtering the dashboard to one segment at a time gives you more levels, not a decomposition — and it quietly drops the interaction effects, where the real driver often hides.
The manual-SQL trap
So the dashboard punts to a human, and the human writes SQL. This is the throwaway-query trap every analytics engineer knows:
- A pull request that isn't really a pull request — a one-off query nobody will reuse.
- Half a day to a week, depending on how many dimensions are in play.
- Interaction terms either omitted (so the story's incomplete) or attempted and botched (so it doesn't reconcile).
- A final answer hedged with "directionally" because the numbers don't sum to the total.
Then the same metric moves next month and you do it all again from scratch.
What reconciled attribution adds
Reconciled attribution is the missing layer between "the dashboard shows it moved" and "here's exactly why":
- It decomposes the change, not the state — contributions, not levels.
- It includes intersections — region × product × channel — so the real driver doesn't get smeared away.
- It reconciles to the total — every contribution sums exactly to the observed move, with zero unexplained residual and no balancing plug.
- It's fast and repeatable — seconds, in your warehouse, the same way every time, not a fresh hand-built query each month.
Your dashboard answers "what is it now?" Reconciled attribution answers "what moved, and exactly why?" You need both — and most teams are missing the second one.
FactorPrism adds that layer as a Snowflake Native App: an exact, reconciled, intersection-level decomposition of any metric change, in seconds — your data never leaves your account.
Stop hand-writing the throwaway query. See the reconciled waterfall →