Sales fall—or costs spike
The result is obvious. The reason is not, and someone needs an answer before the next meeting.
The manual investigation problem
A pivot only finds the answer if you guess the right cut.
Sales dropped. Was it region, division, state, product, SKU, channel, customer—or a combination? Even a simple factor can sit at the one level you did not drill into. Add intersections and the number of possible cuts explodes. Checking them one by one is not a workable search.
You guess where to cut next
Each filter reveals another plausible pattern—but only among the levels and combinations you thought to ask for.
The search explodes
Every added dimension multiplies the possible levels and intersections. Most will never fit into a manual investigation.
You cannot know what you missed
A defensible answer requires ruling out cuts you never had time to make.
A one-dimensional pivot cannot isolate an effect that exists only at an intersection. Region can look normal. Product can look normal. One region × product combination can still drive the entire change.
The missing layer
A complete search—not another cut.
It does not wait for you to guess the next cut. FactorPrism examines the relevant dimensions and intersections together, separates overlapping movement, locates each factor where it acts, and measures its contribution to the reported change.
- Find the factor hidden in a combination. See a region × product × channel movement that looks harmless in every individual view.
- Separate broad movement from local exceptions. Distinguish the rising tide from the segment quietly falling behind it.
- Replace several partial stories with one measured answer. Contributions reconcile to the metric change so the team can inspect the result instead of debating the arithmetic.
Product demo
See FactorPrism get to the answer in 70 seconds.
From “why did sales drop?” to ranked factors, supporting evidence, and the next questions worth investigating.
The questions teams get stuck on
Recognize the investigation?
FactorPrism is built for the questions that survive the first dashboard, pivot, and drill-down.
Hidden under the headline
The total looks healthy. One part of the business is not.
A broad lift masks a regional drag that the rolled-up number makes look positive.
See the hidden region → Offsetting movementA beat and a miss cancel each other out.
“On plan” conceals two large, opposing movements that demand different actions.
See the variance question → Conflicting cutsThree dashboards tell three different stories.
Payer, denial code, and service line each look responsible because the same records appear in all three.
See the hidden intersection → The combination mattersNo individual slice explains the movement.
Region looks normal. Product looks normal. Their intersection contains the decisive factor.
See the product-region surprise →From uncertainty to a Finding
Bring the question. FactorPrism does the searching.
Choose what changed
Use one governed Snowflake table or view, then define the result, time window, and business dimensions that matter.
Analyze the dimensions together
FactorPrism separates overlapping signals and searches meaningful intersections without asking you to preselect a hypothesis.
Open the measured answer
Review the ranked factors, their contributions, linked evidence, and the focused next checks in a reusable Finding.
Under the hood, FactorPrism draws on methods from signal processing to separate overlapping business signals before AI helps explain the measured result.
Measured against known answersReview the benchmark →
Built to produce an answer you can check.
FactorPrism was tested against controlled scenarios where the underlying factors were known in advance—not judged only by whether the output sounded plausible.
280blind benchmark runs
100%known factors recovered in the operating regime
0unexplained residual in the published benchmark
Native to Snowflake
Your data stays where it is governed.
- Runs inside your Snowflake account. No external analytics service receives the source data.
- Read-only access. You choose the table or view and keep your existing roles and controls.
- Works for people and software. Investigate in the app or call the same engine from SQL, tasks, pipelines, and approved agents.
Start with a real business surprise
Bring the number your team is still debating.
Use a sanitized extract, a representative sample, or a Snowflake table or view you choose. In a focused session, we will see whether FactorPrism can automatically surface the factors buried across the dimensions.
Free during early access · No card required · No sensitive data required for a guided evaluation
FAQ
Questions before you investigate
Why can’t a dashboard or pivot table explain the whole change?
A dashboard or pivot answers only the cut you ask for. Even when the effect is simple, you have to guess the right dimension and level. As dimensions multiply, the search explodes. And if an effect exists only at an intersection, separate one-dimensional cuts cannot isolate it: region can look normal and product can look normal while one region × product combination drives the change. FactorPrism searches the relevant levels and intersections together, separates overlaps, and measures how much each factor contributed to the whole change.
Does my data leave Snowflake?
No. The Native App runs inside your Snowflake account and needs read-only access only to the tables or views you choose.
Can I try it before installing?
Yes. Watch the 70-second demo or request a guided evaluation using a sanitized extract, representative sample, or source you choose.
Can I use the same analysis from SQL or an agent?
Yes. The API returns structured factor results to SQL, Snowflake tasks, pipelines, and approved agent workflows.