The data was everywhere, and nowhere together
Each system answered its own small question. None could answer the big ones. Hover or tap a system to see what it held and what it cost the team.
Years of sales and customer history
- Many separate tools, each with its own reports
- Nobody had one view of a customer
- Reports were rebuilt by hand, every time
One nightly pipe per source, into one warehouse
Click a source to follow its data from the raw copy, through cleaning, into the shared marts everyone uses.
Sources
Raw
Land it as it comes
Full record kept, nothing edited
Staging
Clean it once
One customer key, typed dates, money in dollars
Marts
Share one view
Orders, customers, channels, cohorts
Nightly job per sourceHistory first, then only what is newEvery load writes an audit rowRead-only copy for heavy analysisNothing written back to a source
The three layers shown are the demo design. Row counts and sample rows are read from the synthetic warehouse.
One shared view of every order and customer
Change the dates, channels or customer type. Every chart recomputes from the same marts.
Date range
Customer
Channel (last touch)
Revenue trend
Net revenue by month (after discounts, before shipping and tax)
Acquisition vs retention
Revenue from first orders vs repeat orders
New customersRepeat customers
Cohort retention curves
Share of first-time buyers with a second order, by days since their first order
Channel contribution
Share of net revenue by last-touch channel
Sample kit to full order
Customers whose first order was the sample kit, and how many came back for a full-size product
Questions that took days, now one saved query
Pick a question. You see the exact SQL, the answer table and a chart, all computed from the synthetic warehouse.
Build vs buy, at public list price
Move the sliders to size the team and the marketing list. The yearly subscription cost of each option recomputes from the same list-price assumptions.
Illustrative estimate from public list prices; not client costs. Limits: a warehouse is not a CRM. It has no sales pipeline screens, sends no campaigns and has no vendor support. It takes builder time, which is not priced here. List prices are not what most companies pay.
Yearly subscriptions, cheapest full suite vs this approach
Assumptions
- Public list prices in USD, checked Oct 20, 2026
- No discounts, taxes, add-ons or overages
- Existing web store, email and ads tools kept in every option
- Builder time not priced in any option
- Suite A seats: half core seats, half sales seats
- Suite B: one Sales Cloud seat per user
- This approach held flat across the slider range
- Suite A = HubSpot, Suite B = Salesforce. Price references only, not the tools used in the real project