Mockitt AI · Ask a better question of your data
AI Data Analysis
Find the change that matters in your spreadsheet.
Use the Mockitt AI desktop app to work through questions about local Excel and CSV files. Explain the metric and comparison you need, inspect the data, and develop a summary you can trace back to the underlying numbers.
File tasks use the desktop app for Windows or macOS.

Choose the task you need to complete
Sales changes
Compare periods, products, or regions. Define whether the figures include tax, canceled orders, and returns before asking where growth came from.
Channel performance
Compare visits and leads, then add costs or conversions if the decision calls for them. Traffic alone cannot establish return on ad spend.
Expense patterns
Group spending by category, team, or month. Separate one-time charges from recurring costs so a large purchase does not look like a continuing trend.
Project backlogs
Review task age, status, and completion dates. An empty completion field means the record is missing a date, not that the task finished today.
Understand the rows before calculating the totals
Start by asking which worksheets, columns, and records were read. Confirm what one row represents. An order-level table and an order-item table require different counting rules. Adding monthly totals to the daily rows that produced them would count the same activity twice.
Keep blank values separate from zeros. A blank amount may mean the information has not arrived; zero is a recorded value. Identify duplicates using a meaningful key, such as an order ID and item ID. Two rows with the same amount are not necessarily duplicates, and an unusually large value is not automatically wrong.
A small channel comparison you can check yourself
The illustrative data below compares two complete periods with the same measurement rules. Total visits rise from 2,410 to 2,628, an increase of 218. The overall growth rate is 218 divided by 2,410, or 9.0% rounded to one decimal place.
Organic search adds 240 visits, paid search loses 68, and partner referrals add 46. Those changes add up to 218. Organic search contributes more than the total net increase because the paid-search decline offsets part of that growth. The table describes where the change occurred; it does not explain its cause.
| Channel | Previous period | Current period | Change | Change rate |
|---|---|---|---|---|
| Organic search | 1,240 | 1,480 | +240 | +19.4% |
| Paid search | 860 | 792 | −68 | −7.9% |
| Partner referrals | 310 | 356 | +46 | +14.8% |
| Total | 2,410 | 2,628 | +218 | +9.0% |
On a small screen, swipe across the table to see every column.
Match the method to the question
Use grouped totals to understand composition, a two-period comparison to locate increases and decreases, and a complete time series to inspect whether a pattern persists. For unusual records, list the rule they triggered and their location in the source file. Investigating an exception is different from deleting it.
Recalculate ratios from their numerators and denominators. An overall conversion rate is total conversions divided by total visits, not usually the average of channel conversion rates. The same applies to growth: compare the two period totals rather than averaging each channel's percentage. If the baseline is zero, report the absolute change and explain why that percentage is not defined.
Separate a finding from a business decision
The channel example supports a specific finding: total visits increased while paid-search visits declined. It cannot establish whether advertising became less profitable. That decision also needs costs, conversions, attribution rules, and a suitable comparison period. Request those inputs before asking for a budget recommendation.
Choose a chart that supports the discussion. A bar chart can compare channels; a line chart can show a sequence of periods. Keep the table when readers need exact numbers. Include the metric, unit, and time range in the chart title, and make sure its values agree with the accompanying table.
From your materials to a useful result
Choose a file and a question
Identify the worksheet, time range, unit, and meaning of a row. Keep an original copy and specify a separate output file.
Check what was read
Review fields, missing values, duplicate candidates, and record counts. Resolve a failed file read before asking for numerical conclusions.
Build the analysis in stages
Summarize first, compare second, and interpret third. Request the results table, chart, and narrative separately so discrepancies are easier to spot.
Trace the answer to the input
Spot-check source rows, totals, and ratios. Add missing business context before moving from an observed pattern to a recommendation.
Illustrative analysis prompt
Compare two periods without losing the denominator
Continue with a focused revisionWrite three findings supported by the checked table. Separate observations from possible explanations. List the additional cost and conversion data needed before making a channel-budget decision.
The sample figures illustrate the calculation. Use comparable periods and your own records for a real analysis.
A few useful details
AI Data Analysis questions
Is this a BI platform?
This workflow starts with local files and a defined analysis task. Database connections, continuous refresh, governed business metrics, and shared live dashboards are separate capabilities; they should not be assumed from file-based analysis.
Can I start without knowing how to code?
Yes, you can describe the question in ordinary language. You still need to explain what the columns mean and how the business measures the result. Start with a small summary or comparison before adding complex conditions.
How do I verify the example's 9.0% growth?
Add the previous-period values to get 2,410 and the current-period values to get 2,628. Subtract to get 218, then calculate 218 ÷ 2,410 × 100 = 9.0456%, which rounds to 9.0%.
What if the file cannot be read?
Check that it opens locally and confirm the file location, worksheet, headers, and CSV encoding. Ask for the specific failure. A report should not fill in guessed numbers when an input or calculation is unavailable.
Can I begin before collecting the data?
You can define the question, required fields, and report structure. Use clearly labeled sample data to illustrate the format, then replace it with verified records before making a real business decision.
Bring your next task to Mockitt AI
Start with the materials you have and the result you need. Build a first version, review it, and keep working from there.