Most retail analytics setups are over-built and under-trusted.
Walk into the average mid-market retailer and you will find 30-60 dashboards across Shopify, Klaviyo, Meta Ads Manager, Google Ads, GA4, the email tool, and a handful of homegrown spreadsheets. Each dashboard tells a slightly different story. None of them are trusted at executive level. The team falls back on the CEO asking specific questions on Slack.
This is not a tooling problem. It is a discipline problem. The cure is fewer dashboards, one source of truth, and a strict weekly cadence.
The retail analytics stack we actually recommend.
Three layers, in order:
Layer 1: warehouse
Snowflake, BigQuery, or Postgres-based warehouse (Supabase for very small retailers). All commerce, marketing, CRM, and POS data lands here. This is the only source of truth for retail revenue.
Layer 2: transformation
dbt or similar. Define metrics once, in version control. Stop the "my CAC is different from your CAC" debate by defining CAC in one SQL model used by every dashboard.
Layer 3: dashboards
Looker, Metabase, Hex, or Lightdash. We recommend retailers maintain fewer than 12 dashboards total — one for executive, one per channel, one for CRM/lifecycle, one for inventory, one for finance reconciliation.
The eight retail KPIs that should appear every Monday.
- Revenue (week over week, year over year).
- Orders + AOV.
- Blended CAC and channel CAC.
- Cost per qualified lead by channel.
- Repeat purchase rate (rolling 90 days).
- Active inventory weeks of supply.
- Customer support response time and CSAT.
- AI search citation share (the new top-funnel proxy — see retail AEO).
Attribution — the honest version.
Last-click attribution remains the working default for retail reporting because it is simple and stable. Multi-touch attribution models add accuracy but require trust in the model itself.
What actually works in practice: use last-click for tactical channel reporting, validate strategic decisions with periodic incrementality tests (geo-holdouts, channel pauses), and treat blended ROAS as the only honest top-line number.
Reporting cadence — weekly beats monthly.
Monthly business reviews are too slow for retail. By the time the report is ready, the month is over and the levers are gone. Weekly retail business reviews — every Monday, same format, same KPIs, decision-orientated — are what consistently distinguish disciplined retail operators.
What a weekly retail business review looks like:
- One page, eight KPIs, traffic light against plan.
- Five-minute readout of variance.
- Decisions documented in writing, not implied.
- Owner + due date for every action item.
AI in retail analytics — where it helps now.
Three concrete AI-in-analytics patterns we deploy with retail clients:
- Anomaly detection on daily revenue, returns, and ad spend. Flag deviations beyond expected variance before humans notice.
- Natural-language query on the warehouse — letting non-technical retail managers ask "why is luxury watch revenue down 12% week over week" and get a real answer.
- Demand and inventory forecasting for grocery, beauty, and seasonal fashion (see AI for retail).