A data analysis AI assistant is only as useful as its ability to actually run numbers, not just talk about them. Ask Mio’s “Analyse my data” job — one of the 24 ready-made jobs — does exactly that: paste a table or attach a CSV/XLSX and Mio finds patterns and builds a summary, backed by the code runner tool that actually executes Python against your file instead of guessing at what it probably contains.
That distinction is the whole reason this feature is more reliable than asking a generic chatbot to “look at this data” — a model reading numbers out of a pasted table and describing what it notices is prone to arithmetic slips and pattern hallucination on anything beyond a small table. A model that writes and runs actual pandas code against the file, then reports what the code returned, isn’t guessing.
What “finds patterns and builds a summary” means in practice
Once you attach a file or paste a table, Mio can compute real statistics — totals, averages, growth rates, outliers, correlations between columns — using the code runner, which runs Python or Node in a sandbox specifically to “analyse a table (pandas), draw a chart (matplotlib), or convert a file,” as the features page puts it. The output is typically a written summary of what the data shows, plus a chart when a visual makes the pattern clearer than a paragraph would — a trend line, a bar comparison, a distribution. Because the numbers come from code actually running against your file rather than the model eyeballing a pasted snippet, totals and percentages in the summary should match what you’d get running the same calculation yourself.
What to expect from the written summary
A typical summary opens with the shape of the data — row and column counts, date range covered, which columns are numeric versus categorical — before moving into the findings themselves. That opening matters more than it looks: it’s your first check that the file was read correctly, and it’s the fastest way to notice if a column got mis-typed (a price column read as text, for instance) before you trust anything downstream of it.
What this is good at
- Summarising a large table quickly — row counts, column types, missing values, basic distributions.
- Finding trends over time — month-over-month change, seasonality, growth or decline.
- Spotting outliers — rows or values that don’t fit the pattern of the rest.
- Comparing groups — performance by region, category, customer segment.
- Producing a chart alongside the written summary, generated from the actual data rather than described in prose.
Pairing it with the code runner directly
“Analyse my data” and the code runner aren’t two separate things — the ready-made job is essentially a guided way of using the code runner for tabular analysis. That connection matters because it means you’re not limited to the summary Mio decides to write by default: you can ask follow-up questions that require new calculations — “what’s the median order value for customers who bought more than once,” “show me this broken down by quarter instead of by month,” “recalculate that excluding refunds” — and each one runs fresh code against the real file rather than trying to extrapolate from the first summary. This is a meaningfully different experience from a chatbot that answered your first question by reading the table once and is now working from memory of what it saw.
File size limits by plan
Because this job depends on an uploaded file, the plan you’re on sets a hard ceiling on how large that file can be. The limits, straight from the pricing page: 20 MB on the Chat plan, 100 MB on the Coding and Design plans, and 200 MB on the Business plan. The Free plan has no file uploads at all, so this job needs at least Chat. For most spreadsheets and CSV exports — even fairly large ones, since CSV is a compact text format — 20 MB covers tens of thousands of rows comfortably. If you’re regularly working with larger exports (a full year of transaction-level data, a wide table with many columns), the Coding or Business plan’s 100–200 MB ceiling gives real headroom, and those plans also unlock the full Code mode and execution sandbox rather than just the analysis job.
A worked example
Say you export a year of e-commerce order data as a CSV — order ID, date, customer, product, quantity, price, region. A request like “find the top 5 products by revenue, show monthly revenue trend as a chart, and flag any month where revenue dropped more than 15% from the previous month” gives Mio a concrete, checkable task. The code runner opens the file, computes revenue per product (price × quantity, grouped and summed), ranks the top five, builds a monthly trend, computes month-over-month percentage change, and returns both the numbers and a chart. Because it’s real code, you can ask it to show its work — “what exact numbers did you use for March” — and get the actual computed values, not a paraphrase.
Comparison: ways to analyse a spreadsheet
| Method | Runs real calculations | Handles large files | Produces a chart | Time to first insight |
|---|---|---|---|---|
| Ask Mio’s data analysis job | Yes, via the code runner | Up to plan limit (20–200 MB) | Yes, on request | Under a minute |
| Pasting a table into a generic chatbot | No, often estimates | No, practically limited | No | Fast, but less reliable |
| Excel pivot tables by hand | Yes | Yes, within Excel’s limits | Yes, manually built | 10–30 minutes |
| A dedicated BI tool (Power BI, Tableau) | Yes | Yes, built for scale | Yes, extensive | Setup overhead, ongoing |
Getting a useful summary instead of a generic one
The single biggest lever is telling Mio what question you actually care about instead of just attaching a file and saying “analyse this.” A file with fifteen columns could be summarised a dozen reasonable ways — Mio will pick sensible defaults (an overview of each column, obvious trends), but a specific business question gets a specific, useful answer instead of a general-purpose overview you then have to mine for what you needed. “What’s driving the drop in Q3 revenue” is a much better prompt than “tell me about this data.”
It also helps to describe the data briefly if the column names aren’t self-explanatory — “column C is the discount percentage applied, column F is whether the customer used a referral code” saves a round trip and avoids a summary built on a wrong guess about what a column means.
Where this differs from the spreadsheet job
It’s worth distinguishing this from the spreadsheet generator covered earlier in this series: that job builds a new multi-sheet .xlsx from a description — a budget, tracker or comparison you didn’t have yet. This job does the opposite: it takes data you already have and extracts insight, a summary and a chart from it. If your workflow needs both — analyse existing data, then hand the findings to someone as a clean spreadsheet — you can ask for that in the same conversation, moving from analysis into a formatted output.
Common mistakes that produce a weaker analysis
The most common one is uploading a file with no context and expecting Mio to know which column matters most to you. A table of order data could be analysed for revenue trends, customer retention, product performance, regional differences or seasonal patterns — all valid, all different. Without a steer, Mio picks a reasonable default overview, but that default is rarely the exact angle you actually needed, so you end up asking a follow-up question anyway. Leading with your actual question saves that round trip.
The second is treating a correlation the analysis surfaces as a cause. If the summary notes that revenue rose in months with more ad spend, that’s a pattern in the data, not proof that the ad spend caused the increase — other factors (seasonality, a product launch, a competitor’s stock-out) could explain the same pattern. Ask Mio to check for obvious confounding factors if the distinction matters for a decision you’re about to make, and treat any causal claim in the summary as a hypothesis worth testing further, not a settled fact.
The third is not checking the code runner’s assumptions when a column is ambiguous — a “status” column with values like “1” and “0,” or a date column mixing two formats, can be interpreted differently than you intended. Ask Mio to show you how it interpreted an ambiguous column before trusting a summary built on that interpretation, especially the first time you analyse a new type of export.
Privacy considerations for business data
Because this job usually involves uploading real business data — sales figures, customer records, financial numbers — it’s worth being deliberate about what you include. Strip or anonymise columns that aren’t needed for the analysis you’re asking for, particularly personal data like customer names, emails or addresses, if the question you’re asking (revenue trends, product performance) doesn’t actually require them. Ask Mio’s own handling is consistent across the platform — files you upload are yours, not used to train models, and stored on EU servers in Germany — but minimising what you upload in the first place is good practice regardless of the platform, the same discipline you’d apply before emailing a spreadsheet to anyone else.
Where this fits
The data analysis job is one of 24 ready-made jobs on the ready-made jobs hub, and it leans directly on the code runner covered in the first run of this series. Every guide here, including this one, sits under the “What Mio can do” tag if you want to browse the rest of the feature set.
Frequently Asked Questions
What file formats work for this job?
CSV and XLSX are the primary formats named on the features page; Mio can also work with data pasted directly as text if it’s small enough to fit in a message.
How large a file can I upload?
Up to your plan’s file size limit: 20 MB on Chat, 100 MB on Coding and Design, 200 MB on Business. The Free plan doesn’t support file uploads.
Can it handle messy data — missing values, inconsistent formatting?
Yes, to a reasonable degree — the code runner can clean and standardise a column before analysing it, though very inconsistent data (mixed date formats, free-text fields meant to be categories) benefits from you flagging the mess up front so Mio handles it deliberately rather than guessing.
Does the chart come as an editable file or just an image?
It’s generated as an image (via matplotlib) embedded in the response. If you need an editable chart inside a spreadsheet, ask for the analysis paired with a generated .xlsx that includes a native Excel chart instead.
Can I ask it to build a forecast, not just describe the past?
Yes, within reason — simple trend-based projections are something the code runner can compute, though anything requiring a sophisticated statistical model should be treated as an estimate, not a guarantee, and reviewed with that in mind.
Is my uploaded data used to train the model?
No — chats and files are yours, not used for training, and stored on EU servers in Germany.
What’s the difference between this and Code mode?
This ready-made job is a guided entry point into the same underlying code runner that powers Code mode — Code mode gives you the full sandbox for general programming tasks, while this job is framed specifically around tabular analysis with a summary and chart as the expected output.
The Bottom Line
“Analyse my data” is useful specifically because it computes rather than guesses — the code runner behind it executes real pandas operations against your actual file, so the numbers in the summary are numbers you can trust and re-derive. Ask a specific question, mind your plan’s file size limit, and use the same conversation to drill into follow-up questions rather than starting over. Try it on the Chat plan with your next export, and lead with the specific question you actually need answered rather than a bare attachment.