How to Use Finzite MCP with AI Assistants
The goal
Use natural-language questions to explore your Finzite data without manually building reports. The more specific your question is about time period, sales channel, store, metric, or comparison, the more useful the answer can be.
1. A simple formula for better questions
A strong MCP question usually contains four things:
- Metric or topic - what you want to understand (for example, contribution, COGS, advertising spend, revenue).
- Time period - such as the last 3 months, this month, or a specific date range.
- Scope - such as a specific store, online sales, POS sales, or an advertising channel.
- Analysis request - what you want the AI to do with the data: compare, explain, identify trends, rank, or summarise.
2. Understanding expenses and costs
Use MCP to investigate where money is being spent and which cost categories are contributing most to your variable costs.
Example: Find the biggest expense
More specific versions can give you a more actionable answer:
- What are my 5 biggest expense categories over the last 3 months, and what percentage of total variable costs does each represent?
- Which cost category increased the most over the last 3 months?
- Break down my total variable costs for the last 3 months into COGS, fulfilment fees, transaction and platform fees, and advertising spend.
- Which expense is growing fastest month over month?
3. Finding negative trends in contribution
For trend questions, always state the period and the scope you care about. This is especially useful when you need to isolate online sales from POS or a particular store.
Example: Contribution trend for online sales
Useful follow-up questions:
- Show me the monthly contribution and contribution margin for all online sales for the last 3 months.
- What changed most over the period that could explain the decline in contribution?
- Which cost category contributed most to the change in contribution for online sales?
- Compare online contribution this month with the previous month and explain the main drivers of the change.
Tip: ask for the driver, not only the trend
Instead of stopping at “Is contribution declining?”, follow with “What are the main drivers of the change?” This turns the question from a status check into an investigation.
4. Comparing periods
MCP is especially useful when you ask it to compare two clearly defined periods.
| Topic | Example prompt |
|---|---|
| Revenue | Did revenue increase or decrease this month compared with last month? |
| Contribution | Compare contribution for the last 30 days with the previous 30 days. What changed? |
| Advertising | How has advertising spend changed over the last 3 months, by channel? |
| Costs | Which variable cost category increased the most compared with the previous quarter? |
5. Channel and store-specific questions
Use scope words such as “online sales”, “POS”, “all stores”, or a specific store name to prevent mixing different parts of the business.
- What is my contribution for online sales only this month?
- Compare POS and online contribution for the last 90 days.
- Which store had the highest contribution last month?
- Which store had the largest increase in COGS over the last 3 months?
- Show me advertising spend for all stores, broken down by channel.
6. Questions about advertising efficiency
- How much did I spend on Meta, Google, and TikTok ads over the last 30 days?
- Which advertising channel had the highest spend last month?
- How did advertising spend affect contribution over the last 3 months?
- Compare advertising spend and sales revenue by month for the last 6 months.
7. Questions about sales performance
- What were my gross sales, discounts, returns, net sales, and total sales last month?
- Which month had the highest revenue over the last 12 months?
- What is my average order value for the last 30 days?
- How have orders and revenue changed over the last 3 months?
8. Questions about profitability
- What was my gross profit for the last 3 months?
- What is driving changes in my contribution?
- Break down my contribution for the last month into COGS, fulfilment fees, transaction and platform fees, and advertising spend.
- Which month had the highest contribution margin over the last 6 months?
9. Ask follow-up questions
You do not need to ask one perfect question. Start broad, then drill into the answer with follow-up questions.
10. Avoid vague questions
A vague question can produce a less useful answer because the scope is unclear. Add the missing context.
| Instead of… | Try… |
|---|---|
| How are sales? | How did revenue and total sales perform over the last 30 days compared with the previous 30 days? |
| What are my costs? | Break down my variable costs for the last 3 months and identify the largest category. |
| Is contribution bad? | Is contribution trending down for all online sales over the last 3 months? |
| What is my ad spend? | What was my total ad spend over the last 30 days, broken down by Meta, Google, and TikTok? |
11. Prompt checklist
- What metric am I asking about?
- What time period should be analysed?
- What store, channel, or sales scope should be included?
- Do I need a comparison?
- Do I want a trend, ranking, breakdown, or explanation?
- Should the answer identify the main drivers behind the result?
Example of a strong Finzite MCP question
“Compare my contribution for all online sales over the last 3 months. Show the monthly contribution and contribution margin, identify whether there is a downward trend, and explain which cost categories were the main drivers of any change.”
12. Getting more useful answers
- Use exact periods when precision matters: “1 July to 30 September” is clearer than “recently.”
- Specify scope: “online sales only”, “all stores”, or a named store.
- Ask for both the result and the reason: “What changed?” followed by “Why did it change?”
- Use follow-up questions to move from summary → trend → driver → action.
- When validating an important number, ask MCP to show the supporting breakdown or comparison used in the answer.
Was this article helpful?
That’s Great!
Thank you for your feedback
Sorry! We couldn't be helpful
Thank you for your feedback
Feedback sent
We appreciate your effort and will try to fix the article