Definition
Root cause analysis is a structured investigation that asks not "what are customers saying?" but "what changed to make them say it?" In feedback work, the symptom is usually visible in the data: a theme that was 4% of mentions last month is now 18%; star average on a SKU dropped half a point in two weeks; return-reason "does not hold charge" tripled on one batch. RCA is the work that comes after detection.
The techniques are borrowed from manufacturing and incident response. Theme drill-down isolates the feedback subset driving the symptom. Cohort comparison splits that subset by retailer, region, manufacturing date, firmware version, or accessory bundle. Timeline alignment overlays the symptom against product events — a packaging revision, a supplier change, a listing update, a firmware push, a price change, a marketing campaign. The goal is a testable hypothesis about cause.
Why it matters
Without RCA, feedback work stops at reporting. A dashboard that shows sentiment dropped is useful exactly once: it tells QA something is wrong. What QA needs next is a named cause so it can route action to the responsible team — design, packaging, a supplier, a factory line, a listing owner. A feedback system that cannot do that forces a second investigation off-platform, which is where most programs stall.
RCA also shortens the distance between signal and fix. A brand that detects a packaging issue three weeks after launch and routes the cause to the packaging supplier the same week ships a corrective change months before warranty data would have forced it. In Indellia, the Anomaly Agent flags the spike, and a Theme Agent drill-down plus cohort comparison narrows it to the likely cause. The investigation stays inside the feedback data.
Example
A home-appliance brand sees a 23% week-over-week jump in the theme "loud motor" on one vacuum SKU. The Anomaly Agent flags it. A Consumer Insights analyst uses Indellia to drill down: 71% of the spike is on units purchased in the last 30 days, concentrated at Home Depot and Lowe's, not Amazon. Timeline alignment shows a supplier change for the motor assembly shipped to retail warehouses 36 days earlier; Amazon inventory is still on the prior supplier. QA opens a factory ticket against the new supplier, not a general motor review. The whole investigation takes a morning, not three weeks, and the RCA artifact — symptom, cohort, hypothesis, evidence — is attached to the factory ticket so the supplier has the data in one place.