FactorPrism®
Snowflake Native App

Why did my metric change in Snowflake?

FactorPrism finds the business locations driving the movement, ranks their contributions, and reconciles the explanation to the total change—inside your Snowflake account.

See it on one of your metrics Watch the 73-second demo
The short answer

Connect the metric, define the business structure, and get a ranked explanation

A useful answer has to show where the movement originated, how much each location contributed, and whether gains and declines offset. FactorPrism performs that analysis as a Snowflake Native App—no export pipeline and no manual slice-by-slice investigation.

+
A complete explanation that ties out Every contribution sums exactly to the observed change — zero unexplained residual, no balancing plug (verified: residual 0.00e+00 on a real engine run, 13/13 automated verification checks). Small drivers roll into a labeled "Other smaller factors" bar that is itself a sum of real attributed drivers.
+
Down to factor intersections Attributes the move across dimensions and their intersections (region × product × channel) — so the localized driver that single-dimension slicing smears away shows up where it actually acts.
+
In seconds A complete, reconciled intersection-level analysis in seconds—thousands of segment intersections evaluated in under approximately four seconds in the core engine. Very large or deep datasets, the warehouse scan, and optional narration can take longer.
+
Your data never leaves Snowflake Runs in-account as a Native App with SELECT-only access to a single table or view and zero external network calls. The AI-written narrative runs via Snowflake Cortex, inside your account, under your existing roles and governance. (Metadata-only error telemetry, never row data, is sent to the provider; Cortex may use cross-region inference within Snowflake.)
Three steps

From a changed KPI to an answer you can present

Connect a table or view, select the date, metric, and business dimensions, then review the ranked drivers and waterfall. You can use the visual app, call the same analysis from SQL, or invoke it from an AI-agent workflow.

Connect a table or view → Run the analysis from SQL → Use it from Cortex Code → Review the published benchmark →

See whether it works on one of your metrics

Start with a 30-minute guided evaluation using a sanitized extract, representative sample, or a Snowflake table or view you choose. No installation is required before the conversation.

Request a guided evaluation Open the Marketplace listing