Ramp Vendor Spend Analysis

作者: Ramp

分析来自Ramp的供应商支出数据并导出至已连接的系统。当用户要求“分析供应商支出”、“建立供应商数据库”、“更新受管供应商”、“供应商排名报告”、“显示支出超过X的供应商”、“供应商支出分析”、“供应商续约”、“合同截止日期”,或询问供应商支出模式、供应商负责人、采购订单或部门级供应商数据时使用。

npx skills add https://github.com/ramp-public/mcp-skills --skill ramp-vendor-analysis

Ramp Vendor Spend Analysis

Overview

This skill extracts comprehensive vendor data from Ramp, enriches it with owner/department information and contract data from purchase orders, and outputs structured results. Data can be displayed as text/tables or exported to connected integrations (Notion, Google Sheets, etc.).

When to Use This Skill

  • User asks to analyze vendor spend or build a vendor database
  • User wants to see top vendors by spend (L365, L30, YTD, all-time)
  • User asks about vendor owners or which departments own which vendors
  • User asks about contract end dates, renewals, or purchase order status
  • User wants vendor data exported to Notion, spreadsheets, or other systems
  • User asks ad-hoc questions like "vendors with spend over $50k" or "upcoming renewals"

Prerequisites

  • Ramp MCP server must be connected
  • For exports: Target integration (Notion, Google Sheets, etc.) should be connected

Data Collection Workflow

Step 1: Load Core Vendor Data

Tool: ramp-demo:load_vendors

This is the primary data source. Call with default parameters to load all vendors.

Parameters: {}

Data extracted per vendor:

  • Vendor ID
  • Vendor Name
  • Vendor Owner ID
  • Vendor Contacts (email, phone)
  • Total spend L365 (last 365 days)
  • Total spend L30 (last 30 days)
  • Total spend YTD (year to date)
  • Total spend all-time
  • Billing frequency
  • Tax information (W-9 status, tax ID)
  • Active status

Step 2: Load Users for Owner Details

Tool: ramp-demo:load_users

Required to resolve Vendor Owner IDs to names and get their department assignments.

Parameters: {}

Data extracted per user:

  • User ID
  • First name, Last name
  • Email
  • Department ID
  • Location ID
  • Manager ID

Step 3: Load Departments

Tool: ramp-demo:load_departments

Required to map Department IDs to department names.

Parameters: {}

Data extracted:

  • Department ID
  • Department Name

Step 4: Load Bills for Payment Method Analysis

Tool: ramp-demo:load_spend_export

Query bills to determine preferred payment methods per vendor.

Parameters:
  spend_export_type: "bills"
  from_date: [365 days ago, YYYY-MM-DD format]
  to_date: [today, YYYY-MM-DD format]

Data extracted per bill:

  • Vendor ID (to join with vendor data)
  • Payment method used
  • Bill amount
  • Bill date

Step 5: Load Purchase Orders for Contract Data

Tool: ramp-demo:load_purchase_orders

Purchase orders contain contract/renewal information. The spend end date on a PO maps to the contract end date for that vendor relationship.

Parameters:
  from_date: [365 days ago, YYYY-MM-DD format]
  to_date: [today, YYYY-MM-DD format]

Data extracted per purchase order:

  • PO ID
  • PO Number
  • Vendor ID (to join with vendor data)
  • Spend end date → maps to contract end date
  • Total amount
  • Amount paid/billed
  • Receipt status (FULLY_RECEIVED, PARTIALLY_RECEIVED, OVER_RECEIVED, NOT_RECEIVED)
  • Three-way match enabled flag
  • Created date

Contract status logic:

  • If vendor has PO with spend_end_date → "Contract in place", use date as contract end
  • If vendor has multiple POs → "Multiple contracts", use earliest upcoming end date
  • If vendor has no POs → "No contract in place"
  • If PO is nearing full billing (amount_paid approaching total_amount) → flag for renewal attention

Step 6: Join and Enrich Data

After loading all data sources, perform the following joins using SQL queries:

Tool: ramp-demo:execute_query

-- Join vendors with owner names, departments, and contract info from POs
SELECT 
  v.vendor_name,
  v.total_spend_l365,
  v.total_spend_l30,
  v.total_spend_ytd,
  v.total_spend_all_time,
  v.billing_frequency,
  v.tax_status,
  u.first_name || ' ' || u.last_name AS vendor_owner_name,
  u.email AS vendor_owner_email,
  d.name AS owner_department,
  po.spend_end_date AS contract_end_date,
  po.total_amount AS po_total,
  po.amount_paid AS po_paid,
  po.receipt_status,
  CASE 
    WHEN po.id IS NULL THEN 'No contract in place'
    WHEN po_count.cnt > 1 THEN 'Multiple contracts'
    ELSE 'Contract'
  END AS contract_status
FROM vendors v
LEFT JOIN users u ON v.vendor_owner_id = u.id
LEFT JOIN departments d ON u.department_id = d.id
LEFT JOIN purchase_orders po ON v.id = po.vendor_id
LEFT JOIN (
  SELECT vendor_id, COUNT(*) as cnt 
  FROM purchase_orders 
  GROUP BY vendor_id
) po_count ON v.id = po_count.vendor_id
ORDER BY v.total_spend_l365 DESC;

Note: Actual column names may vary. After loading data, query the schema:

PRAGMA table_info(vendors);
PRAGMA table_info(users);
PRAGMA table_info(departments);
PRAGMA table_info(purchase_orders);
PRAGMA table_info(bills);

Output Schema

The final enriched dataset should include these fields per vendor:

FieldSourceDescription
Vendor NamevendorsCompany/vendor name
Vendor Ownerusers (joined)Full name of owner
Owner Emailusers (joined)Email of vendor owner
Owner Departmentdepartments (joined)Department name
Spend L365vendorsLast 365 days spend
Spend L30vendorsLast 30 days spend
Spend YTDvendorsYear-to-date spend
Spend All-TimevendorsTotal historical spend
Billing FrequencyvendorsMonthly, annual, etc.
Tax StatusvendorsW-9 on file, tax ID
Preferred Payment Methodbills (aggregated)Most common payment method
Vendor ContactsvendorsContact email/phone
Contract Statuspurchase_ordersContract / No contract / Multiple contracts
Contract End Datepurchase_orders.spend_end_dateDate contract/PO expires
PO Total Amountpurchase_ordersTotal value of purchase order
PO Amount Paidpurchase_ordersAmount billed against PO
PO Receipt Statuspurchase_ordersFulfillment status

Handling Ad-Hoc Queries

For filtered queries like "show vendors over $50k spend":

  1. Load data using Steps 1-5 above
  2. Apply filters in the SQL query:
SELECT * FROM vendors 
WHERE total_spend_l365 > 50000
ORDER BY total_spend_l365 DESC;

Common filter patterns:

  • Spend thresholds: WHERE total_spend_l365 > [amount]
  • Department filter: WHERE owner_department = '[dept_name]'
  • Active only: WHERE is_active = 1
  • Top N vendors: LIMIT [n]
  • Upcoming renewals: WHERE contract_end_date BETWEEN date('now') AND date('now', '+90 days')
  • Missing contracts: WHERE contract_status = 'No contract in place' AND total_spend_l365 > 25000
  • POs nearing completion: WHERE po_paid / po_total > 0.8

Output Options

After data collection, ask the user how they want to receive the results:

Option 1: Text/Table Display (Default)

Display results as a formatted markdown table directly in the conversation.

Option 2: Export to Connected Integration

Check for available integrations and offer export:

For Notion:

  • Create or update a database with the output schema
  • Map fields to Notion properties:
    • Vendor Name → Title
    • Spend fields → Number ($ format)
    • Contract Status → Select (Contract | No contract in place | Multiple contracts)
    • Contract End Date → Date
    • Tax Status → Select (Tax details verified by Ramp | Missing | N/A)
    • Payment Method → Select (Pay by card | Pay by Bill pay (ACH) | Mixed)

For Google Sheets:

  • Create a new sheet or append to existing
  • Include headers matching the output schema

For other integrations:

  • Adapt the output schema to the target system's format

Example Usage

User: "Build me a vendor database with our top vendors"

Workflow:

  1. Call load_vendors → loads vendor table
  2. Call load_users → loads users table
  3. Call load_departments → loads departments table
  4. Call load_spend_export with type="bills" → loads bills table
  5. Call load_purchase_orders → loads purchase orders table
  6. Execute join query to create enriched dataset with contract info
  7. Ask user: "I've compiled data on [X] vendors. Would you like me to display this as a table, or export to a connected system like Notion?"
  8. Output based on user preference

User: "Show me vendors with contracts expiring in the next 90 days"

Workflow:

  1. Load all data sources (Steps 1-5)
  2. Execute filtered query:
SELECT vendor_name, vendor_owner_name, owner_department, 
       total_spend_l365, contract_end_date,
       julianday(contract_end_date) - julianday('now') AS days_until_expiry
FROM enriched_vendors
WHERE contract_end_date BETWEEN date('now') AND date('now', '+90 days')
ORDER BY contract_end_date ASC;
  1. Display results as table

User: "Which high-spend vendors don't have contracts?"

Workflow:

  1. Load all data sources (Steps 1-5)
  2. Execute filtered query:
SELECT vendor_name, vendor_owner_name, total_spend_l365
FROM enriched_vendors
WHERE contract_status = 'No contract in place'
AND total_spend_l365 > 25000
ORDER BY total_spend_l365 DESC;
  1. Display results with recommendation to establish contracts

User: "Show me purchase orders that are almost fully billed"

Workflow:

  1. Load purchase orders data
  2. Execute query:
SELECT vendor_name, po_number, po_total, po_paid,
       ROUND(po_paid * 100.0 / po_total, 1) AS percent_used,
       contract_end_date
FROM enriched_vendors
WHERE po_paid / po_total > 0.8
ORDER BY percent_used DESC;
  1. Display results - these vendors may need renewal attention

Error Handling

MCP Connection Failed

  • Verify Ramp MCP is connected in Settings > Extensions
  • Check API credentials are valid
  • Try reconnecting the integration

No Vendors Returned

  • Confirm the Ramp account has vendor data
  • Check if filters are too restrictive
  • Try loading without filters first

No Purchase Orders Found

  • Not all vendors will have purchase orders
  • Mark these vendors as "No contract in place"
  • This is expected for many card-based or self-serve SaaS vendors

Join Failures

  • Query table schemas first to verify column names
  • Check for NULL values in join keys
  • Use LEFT JOIN to preserve vendors without owners or POs

Export Failures

  • Verify target integration is connected
  • Check permissions on target database/sheet
  • Confirm field mapping is valid for target system

相关技能

m365-agents-ts
microsoft
使用 Microsoft 365 Agents SDK 构建适用于 Microsoft 365、Teams 和 Copilot Studio 的企业级代理,支持 Express 托管、AgentApplication 路由、流式响应以及 Copilot Studio 客户端集成。
official
golang-samber-mo
samber
Monadic types for Golang using samber/mo — Option, Result, Either, Future, IO, Task, and State types for type-safe nullable values, error handling, and functional composition with pipeline sub-packages. Apply when using or adopting samber/mo, when the codebase imports `github.com/samber/mo`, or when considering functional programming patterns as a safety design for Golang.
developmentcode-review
tinybird-cli-guidelines
tinybirdco
Tinybird CLI 命令、工作流程及操作。适用于运行 tb 命令、管理本地开发、部署或处理数据操作时使用。
official
azure-resource-visualizer
microsoft
分析Azure资源组并生成详细的Mermaid架构图,展示各个资源之间的关系。当:创建…
official
sentry-cloudflare-sdk
sentry
完整的 Sentry SDK 配置,适用于 Cloudflare Workers 和 Pages。当被要求“为 Cloudflare Workers 添加 Sentry”、“安装 @sentry/cloudflare”或配置错误…时使用。
official
initializing-warehouse
astronomer
初始化仓库模式发现。生成包含所有表元数据的 .astro/warehouse.md 文件,用于即时查询。每个项目运行一次,在模式变更时刷新…
official
agents-v2-py
microsoft
使用 Azure AI Projects SDK 中的 ImageBasedHostedAgentDefinition 构建基于容器的托管代理。
official
Music
ElevenLabs
使用ElevenLabs Music API生成音乐。适用于创作器乐、带歌词的歌曲、背景音乐、广告曲或任何AI生成的音乐作品。支持基于提示词的生成、用于精细控制的作曲方案,以及包含元数据的详细输出。
developmentofficial