Statistics you can check, not a black box
Every number comes from one analytics engine with fixed definitions. Every finding comes from a transparent statistical test. Every recommendation comes with the evidence behind it — and none of it depends on an outside AI service.
Three layers, one direction
Data flows up through layers with clear jobs. Each layer only uses what the layer below it computed.
Every metric — revenue, gross margin, AOV, units per transaction, discount rate, days of supply, sell-through — has exactly one definition. Queries are parameterized and run against daily summary tables, so answers are fast and consistent everywhere in the product.
Statistical detectors look for changes that are both unusual (robust z-scores against peers or history) and material (rupee impact). Findings carry their evidence, calculations and a recommended action.
Findings are ranked by rupee impact and delivered as a short brief each morning, with the full evidence one click away and PDF reports for your team.
How the analyst stays honest
These rules are built into the engine — not left to judgement.
Revenue, gross margin, AOV, discount rate and days of supply are each calculated one way, everywhere — the brief, the pages and the reports always agree.
Weekdays are compared with the same weekdays, stores with their own history and their peers, and the chain-wide trend is removed first.
A finding needs both statistical strength (robust z-scores) and a meaningful rupee impact before it reaches your brief.
Observed facts, likely explanations and recommendations are kept apart. Causes are stated as “likely” unless the data shows them.
Each insight lists its method, the formulas, the exact values and the underlying data in a “Why am I seeing this?” panel.
The engine runs inside the platform and your database. Your sales and inventory data isn't sent to an outside AI service.
Only flag what's unusual and material
A 5% dip isn't news if every store dipped 5%. The detectors compare each store with its own weekday pattern and with its peers, using the median and median absolute deviation so one extreme store can't hide another.
- Weekday-adjusted baselines remove the weekend effect
- The chain-wide trend is removed, so seasonality and festivals don't trigger alerts
- Robust z-scores for peer comparison; materiality floors in rupees and percentages
- Thresholds are visible and tunable in Settings
From export to first brief
An isolated organization for your data, with you as owner.
Your store list and product catalog, by CSV.
Line-level sales and stock snapshots. Summary tables build automatically.
The first analysis runs as soon as sales arrive. Invite regional and store managers.
See what your analyst finds in your data.
Book a 30-minute walkthrough, or explore the live demo with a full year of sample data from a 30-store chain.