JB Analytica reference architecture — a published example. The business and every figure on it are invented.

Products

Revenue per product. This is the same realised revenue the performance page totals, read one grain down: realised_net_amount_eur on fct_order_items rather than on fct_orders, summed by product instead of by month. The column is not re-derived here — a cancelled line contributes €0 on this page for exactly the reason a cancelled order does on that one, because it is the same definition in the same mart.

Origin, roast and category are mart columns too. dim_products composes both the display labels and the product name from the source's lowercase enums, so a product reads the same way on every page and no page has to know that the warehouse stores light.

The revenue column is a bar drawn inside the table cell rather than a chart beside it. Product names run to nearly 40 characters, which a horizontal bar chart clips; contentType=bar puts the bar where the label cannot collide with it.

What this ranking cannot tell you is which products are worth selling. Margin is a mart column too — unit_margin_eur on dim_products — but the generator draws cost independently of demand, so ranking by it would be ranking noise. Architecture lists the other places the generated data runs out.