Marketplace

Buyer

The demand side of the marketplace: one row per registered buyer, carrying how they were acquired, where they are, and how much they have transacted. Lifetime order count and a repeat flag make demand-side retention a first-class field — most buyers place a single order and never return, and the share who come back for a second is the health number every marketplace watches.

Fields

SAFE_DIVIDE(COUNTIF(is_repeat), COUNTIF(lifetime_orders >= 1))

Cancellations

One row for every order that was cancelled — and only those. Each row records who pulled the plug (buyer, seller or platform), at which stage (pre-payment, pre-fulfilment or in-transit), the stated reason, and how much was refunded. Because refund magnitude scales with stage — near zero pre-payment, close to full once fulfilment is underway — this is where an ops or trust-and-safety lead finds the fixable concentration of fill-rate failures.

Fields

COUNT(DISTINCT orders.seller_id)

Relationships

Category

The category tree every listing and order is filed under, and — crucially — the standard take rate the platform charges on sales in each category. Take rate is the biggest lever a category manager holds: raise it and revenue per order rises, but the sellers in that category feel the squeeze. Rates follow real marketplace norms — physical goods sit lower (8–15%), while services and premium or handmade categories carry a higher cut (15–22%).

Fields

Listings

One row per listing, the unit of supply a buyer actually orders. Each listing belongs to a seller and a category and carries a price and an availability state. The count of live listings per seller is exactly that seller's active inventory, which makes this the backbone behind every order's price and category — and the first place to look when searches arrive and find nothing to buy.

Fields

Relationships

Orders

The match at the heart of the marketplace: one row per order placed, tying a buyer to a seller's listing. It carries the full transaction economics — gross merchandise value, the category take rate, the platform's cut and the reciprocal seller payout. Revenue is recognised only on orders that settle: a cancelled order keeps its booked GMV for funnel analysis while its net GMV, platform revenue and seller payout all fall to zero, so gross and net are never confused.

Fields

SAFE_DIVIDE(SUM(platform_revenue), SUM(net_gmv))

COUNTIF(status = 'cancelled')

SAFE_DIVIDE(cancelled_orders, COUNT(order_id))

SAFE_DIVIDE(SUM(net_gmv), COUNT(order_id) - cancelled_orders)

COUNT(DISTINCT IF(status != 'cancelled', seller_id, NULL))

CASE WHEN gmv < 50 THEN 'Small (under $50)' WHEN gmv < 100 THEN 'Medium ($50-99)' WHEN gmv < 250 THEN 'Large ($100-249)' ELSE 'Premium ($250+)' END

SAFE_DIVIDE(SUM(net_gmv), active_sellers)

Relationships

Reviews

One row per review left after a fulfilled order — the marketplace's trust signal. Only a minority of completed orders are ever reviewed, since reviewing is voluntary, and ratings follow the familiar J-shape: mostly five stars, a hard bump at one star, little in between. The reviewer role separates the buyer's rating of the seller from the seller's rating of the buyer, and the complaint flag — concentrated in the low ratings — marks where trust is breaking down.

Fields

Relationships

Search Requests

One row per search a buyer runs — the top of the demand funnel and the cleanest read on marketplace liquidity. Each request records the category searched, how many results came back, whether the buyer clicked, and whether the search converted into an order, along with the time it took to match. Searches that return nothing, or return results but never convert, are the "demand with no fill" signal a liquidity analyst hunts for; the live supply they could have found is in Listings.

Fields

Relationships

Seller

The supply side: one row per seller, with the primary category they sell in, their buyer rating, how many listings they keep live, and — critically for supply-health work — when they last sold and whether they are still active. Supply is heavily concentrated: a small share of sellers drives most of the platform's GMV, so lifecycle status (active, dormant or churned, driven by the recency of their orders, or unactivated until the first sale) is where a supply manager looks first when GMV softens.

Fields

Relationships