- average_order_value
- 609
- Rows
- 48
- Version
- v1
- Refreshed
- 12 d ago
saved metric: Average amount of a paid order.
5 columns
every change makes a new one
uploaded by hand
Import report
Nothing is changed silently. Fixes create a new version.
- All 48 rows imported. No issues found.
Cleaning rules
Rules a person approved. They run on every refresh and sync, and each change makes a new version.
AI suggestions are off for this workspace. Turning them on sends the column profile and 50 sample rows to Mistral (EU) when someone asks for suggestions. Workspace settings
Columns
Agents read what each column means and its unit. Change a type and the dataset rebuilds as a new version.| Column | Meaning | Type |
|---|---|---|
order_idunique on every row | Order identifier, unique per row iddrafted Edit | |
order_date2 Jul 2026 to 28 Sept 2026 | Order date datedrafted Edit | |
customer_id11 values | Customer identifier, 11 distinct iddrafted Edit | |
amount120 to 1,500 · total 30,400 | Amount metricmoney (currency not stated)adds updrafted Edit | |
status2 values | Status, 2 values such as paid, refunded dimensiondrafted Edit |
Sample rows
The first 10 rows, exactly as agents receive them.
| order_id | order_date | customer_id | amount | status |
|---|---|---|---|---|
| 1 | 2026-07-13 | C-006 | 120 | paid |
| 2 | 2026-07-12 | C-009 | 1500 | paid |
| 3 | 2026-07-02 | C-009 | 180 | refunded |
| 4 | 2026-07-08 | C-007 | 180 | refunded |
| 5 | 2026-09-04 | C-001 | 320 | paid |
| 6 | 2026-09-13 | C-010 | 120 | paid |
| 7 | 2026-09-28 | C-001 | 250 | paid |
| 8 | 2026-07-18 | C-007 | 180 | paid |
| 9 | 2026-09-06 | C-009 | 180 | paid |
| 10 | 2026-07-18 | C-006 | 180 | paid |
Who can see this dataset
Everyone in this workspace sees it, in the app and through their agents.
Agent activity
Open in audit12:50
Paid_revenue, orders: count, by regions.region, where status is "paid"maya via Claude · v1 · aggregate · 4 rows · 152 ms · metric paid_revenue · 1 row unmatched on customer_id → customers.customer_id
same as ChatGPTCheck
12:48
Describe the datasetmaya via Claude · describe_dataset · 190 ms
12:46
Paid_revenue, orders: count, by regions.region, where status is "paid"maya via Claude · v1 · aggregate · 4 rows · 228 ms · metric paid_revenue · 1 row unmatched on customer_id → customers.customer_id
same as ChatGPTCheck
12:44
Rows: order_id, order_date, customer_id, amount, where status is "refunded", first 50northwind via Cursor · v1 · query_rows · 8 rows · 96 ms
12:44
Average_order_value, paid_orders, by order_datemaya via Claude · v1 · aggregate · 3 rows · 266 ms · metric average_order_value · metric paid_orders
12:42
Paid_revenue, where order_date ≥ "2026-07-01"northwind via Cursor · v1 · aggregate · 1 row · 120 ms · metric paid_revenue
same as ClaudeCheck
12:42
Paid_revenue, by customers.segmentmaya via Claude · v1 · aggregate · 4 rows · 133 ms · metric paid_revenue · 1 row unmatched on customer_id → customers.customer_id
same as CursorCheck
Thu 13:19
Paid_revenue, orders: count, by regions.region, where status is "paid"sam via ChatGPT · v1 · aggregate · 4 rows · 224 ms · metric paid_revenue · 1 row unmatched on customer_id → customers.customer_id
same as ClaudeCheck
Thu 12:42
Paid_revenue, where order_date ≥ "2026-07-01"maya via Claude · v1 · aggregate · 1 row · 171 ms · metric paid_revenue
same as CursorCheck
Wed 13:19
Paid_revenue, orders: count, by regions.region, where status is "paid"sam via ChatGPT · v1 · aggregate · 4 rows · 208 ms · metric paid_revenue · 1 row unmatched on customer_id → customers.customer_id
same as ClaudeCheck
Wed 12:56
Paid_revenue, orders: count, by regions.region, where status is "paid"maya via Claude · v1 · aggregate · 4 rows · 209 ms · metric paid_revenue · 1 row unmatched on customer_id → customers.customer_id
same as ChatGPTCheck
Wed 12:54
Describe the datasetmaya via Claude · describe_dataset · 247 ms
Use it
Same engine, same numbers, whichever door you use.
- Add datatomcp from the Claude, ChatGPT or Cursor directory, or add this URL as a custom connector in any agent (Claude Code, n8n…). One connector covers all your datasets.
https://app.datatomcp.com/mcp - Sign in when asked and allow access. Then ask about orders in plain language.
Full client-by-client setup on Connect.
curl https://app.datatomcp.com/v1/datasets/ds_AkKUKkPC1GpJ/query \
-H "Authorization: Bearer $DATATOMCP_KEY" \
-H "Content-Type: application/json" \
-d '{"limit": 10}'Create an API key in Account. Filters, metrics and paging are in the dataset docs.
Use the datatomcp dataset "orders" (id ds_AkKUKkPC1GpJ) to answer questions about this data.
Call describe_dataset first to read the columns and today's date.
Before filtering on a text column such as customer_id, call distinct_values to get the exact spelling.
Resolve relative dates ("last month") against the server date, using order_date.
It is linked to customers (customer_id = customers.customer_id), regions (customers.region_id = regions.region_id). Group or filter by their columns in the same query, for example customers.region_id.
Use aggregate for totals, counts and averages instead of reading rows. Never sum a column marked summable: false (prices, rates). Quote numbers exactly as returned with the unit from meta.units, mention the dataset version, and pass on meta.warnings that affect the number.Paste this into your agent's instructions so it queries this dataset the right way.