Migrating from Excel to a real BI tool

Most companies' first analytics tool is Excel. Then it is Excel with lookups across six tabs. Then it is Excel plus one person whose unofficial job is keeping the workbook alive. At some point that stops working, and the question becomes how to move without losing the numbers people rely on.

This is a plan for that move at a small or mid-size company, in three phases, with a list of what not to do.

Signs you have outgrown Excel

• Versions multiply: "Revenue_v2", "Revenue_final", "Revenue_final_JM". Nobody is sure which one the board saw.
• One person is the only one who can update certain cells without breaking formulas, and reporting stops when they are on holiday.
• Two reports show different numbers for the same thing, and someone spends Friday afternoon reconciling them.
• You cannot easily look at last quarter, because the workbook only holds the current state and older versions were overwritten.
• The data you actually need lives in Stripe, your CRM, Google Analytics or a database, and every week someone exports CSVs and pastes them in.

If three or more of these are true, the cost of staying is already higher than the cost of moving.

Phase 1: audit (3 to 5 days)

List every spreadsheet the team relies on for reporting. For each one, write down:

• who maintains it and who reads it
• the question it answers ("what did we sell last week, by product")
• where its data comes from (an export, a manual count, another spreadsheet)
• how often it is updated

You will usually find that a handful of workbooks carry most of the value, and the rest are one-off analyses nobody opens any more. Mark the handful. Those are the ones you will rebuild.

While you are at it, write down the definitions hiding in the formulas. "Revenue" in the finance workbook might exclude refunds and tax while "revenue" in the sales sheet includes both. The migration is a good moment to agree on one definition, and the post on building a single source of truth has a method for doing that without a months-long project.

Phase 2: connect the sources directly (5 to 10 days)

Whatever was being exported and pasted into Excel, connect it to the BI tool directly so it updates on a schedule instead of by hand.

In clariBI that usually means:

• Business apps. Stripe, HubSpot, Linear and 83 other apps through the MCP catalog (trial and Starter up), or one of 80 API connectors such as Zendesk, Mailchimp or Klaviyo that use an API key.
• Google and Meta. Google Analytics, Search Console, Google Ads and Meta Ads connect with a sign-in.
• A database. If you have an app database in PostgreSQL, connect it with a read-only user and a view that aggregates what you need.
• Spreadsheets that stay spreadsheets. Some data only exists in a sheet, such as a budget or a manual stock count. Connect it through Google Sheets or upload the file.

For files, a few habits save trouble: put the headers in the first row, one table per sheet, unmerge cells, and save CSVs as UTF-8. Formulas come in as their values, so a #REF! in the workbook arrives as text. The help article on uploading CSV, Excel and PDF files lists the formats and limits.

Every connected source gets an automatic dashboard built from its columns, without AI and without credits. That dashboard is often enough to replace a simple "export and chart it" spreadsheet on its own.

Phase 3: rebuild the reports that matter (5 to 10 days)

Take the handful of workbooks you marked in phase 1 and rebuild each as a dashboard or a saved report. Do not copy the layout of the workbook. Start from the question it answers and build the smallest view that answers it: a few metric cards, a trend line, a table.

Then run the old and new side by side for a month. Each week, compare the headline numbers. When they differ, find out why. Often the old spreadsheet was wrong (a filter left on, a row pasted twice), and sometimes the new one needs a definition fixed. After a month of agreement, archive the spreadsheet as read-only so it stays available for audits but nobody updates it.

In clariBI you can rebuild these three ways: the dashboard builder with AI-suggested metrics (trial and Starter up), manual mode from a template, where you pick the column for each chart yourself and no credits are used (on the trial, Lite and up, though Lite sees few templates today because most are marked for Starter), or by asking the question in plain English and saving the answer as a report. The dashboard builder guide explains each option.

What not to do

• Do not rebuild every spreadsheet. Many exist because making one was easier than asking someone. They will fade once the answers are easy to get elsewhere.
• Do not recreate Excel inside the BI tool. BI tools are dashboards and questions over data that updates itself, not cells and formulas. A worksheet rebuilt cell by cell in a BI tool is worse than both.
• Do not switch everyone on day one. Move one team's reporting first, fix what the side-by-side run shows, then move the next.
• Do not skip the definitions. Migrating three different meanings of "revenue" into a new tool gives you three wrong charts instead of three wrong cells.

What to keep doing in Excel

Excel is still the right tool for one-off calculations, quick what-if models and anything you need to hand to someone outside the company. The post on why modern BI has not killed the spreadsheet makes the case in more detail.

The rule of thumb: if a spreadsheet is updated on a schedule and read by more than one person, it belongs in the BI tool. If it is a scratchpad, leave it in Excel. When you need data out of the BI tool, dashboard data exports as CSV or JSON and reports download as PDF; export formats covers the details.

A worked example

Say a 15-person online shop runs on one workbook with three tabs: weekly sales (pasted from a store export), ad spend (pasted from Google Ads and Meta Ads) and a cost-per-order calculation that divides one by the other.

• Audit: one workbook, read by the two founders and the marketer, updated every Monday by the marketer in about an hour.
• Connect: Google Ads and Meta Ads by sign-in, and the weekly order export uploaded or kept in a Google Sheet the store app updates.
• Rebuild: one dashboard with weekly revenue, weekly ad spend and cost per order. Ask "What was cost per order by week for the last 12 weeks?" and check the first four weeks against the workbook.

If the numbers match for four Mondays in a row, the workbook can be archived and the marketer gets the hour back.

Next step

Pick the one spreadsheet that takes the most hours to update each week, connect its sources in a clariBI trial, and compare next Monday's numbers with the workbook. That single comparison tells you more about the migration than any feature list.