Better Analyst

formulabot.com now redirects to betteranalyst.com, and the brand update page says Formula Bot is now Better Analyst with the same account and work and no migration needed. The scope has widened: Formula Bot began as help with Excel formulas, while Better Analyst covers connecting and preparing data, enrichment, questions in chat, dashboards, reports and scheduled delivery. The Terms of Service name Datasetmatch LLC as the company behind the service.

The original tools remain. The Formula Generator writes Excel and Google Sheets formulas from a plain-English description or explains a pasted formula step by step, for one AI Action each. The plan documentation lists the Excel and Google Sheets add-ons under Starter, so a free account works in the web app.

Better Analyst counts two things. Each prompt in chat is one message, whether it is a simple question or a complex analysis. AI Actions pay for specialised tools, and the usage documentation prices them: formula generation or explanation 1 action, enrichment 1, 3 or 5 actions depending on whether the table has under 100, up to 1,000 or more rows, PDF Table Extract 5 actions for any size, and AI Extract 10 actions per selected page, plus 2 for scanned-document mode and 2 for Row QA.

Editorial arithmetic: the free plan’s 25 actions cover 25 formulas, five Table Extracts, or two pages of AI Extract, but not a 5-page AI Extract, which costs 50. Free also allows one file per chat, five uploads in total, files up to 5 MB and a 5-minute limit on each code run, so a test file should be small.

On the pricing page, the “Save 50%” label sits on the monthly toggle, where every card shows a half-price first month. Editorial arithmetic for Starter over a year: monthly billing costs $10 + 11 x $20 = $230, yearly billing $168, which is $62 less. Against twelve regular months ($240), yearly saves $72, or 30%. The same 30% applies to Max ($336 against $480 per user) and Business ($672 against $960 per user).

Yearly billing is charged once, so the refund rule matters before choosing it: a refund needs a request within 30 days of the payment and fewer than 50 billable uses since then. A yearly plan that is used for a few weeks will usually be past that threshold.

Data Flows have row allowances, and the Fair-Use Policy explains how they behave. Rows are counted as returned by the source, before filters and joins, once per successful run; retries do not count twice, and usage falls out of a rolling 30-day window rather than resetting on a set day. Workspace owners see warnings at 70% and 90%.

At 100%, new manual runs are rejected with the time when capacity frees up, and scheduled runs are skipped until usage falls, then resume. A run that would exceed its ceiling fails before writing, so a destination sheet is never partly replaced. There are no automatic overage charges, but connector, source API and spreadsheet cell limits can impose lower limits than the plan.

Section 10 of the terms sets retention by plan. On the free plan, chats, messages and uploaded files inactive for more than 60 days are deleted permanently. On paid plans, data is kept for as long as the subscription is active, and the subscription guide says data is kept for 30 days after a cancelled plan expires. Connection credentials stay until you delete them.

The Privacy Policy adds that disconnecting a Google account removes the stored OAuth connection, while existing chats, query logs and reports follow the normal retention rules. Airtable is connected with read-only scopes, and Better Analyst does not modify records or download attachment files.

The privacy policy names OpenAI, Google and Anthropic as AI providers. Better Analyst says it does not train its own models on your messages, files or outputs, and that it does not opt in to provider training where a no-training setting is available. Voice dictation audio is sent to OpenAI for transcription and is not saved by Better Analyst. Data is stored on AWS in the US East (Ohio) region.

The security page says analysis code runs in an isolated sandbox for each chat and that connection credentials are encrypted at rest. Owners and admins can limit which tables, columns and rows the AI reads; per-member row access, SSO, SCIM and an exportable audit log are Business features.

1. Prepare a CSV under 5 MB with a year of monthly sales for three products and calculate the yearly totals yourself.
2. Upload it and ask for revenue by product with a chart; open the SQL or Python shown with the answer.
3. Compare each total with your own figures and note how many of the 10 monthly messages the follow-ups used.
4. Run Table Extract on a one-page sample invoice (5 AI Actions) and count the rows and amounts against the PDF.
5. Ask the Formula Generator for one SUMIFS formula and test it in your own spreadsheet.

Accept if the totals match the shown code and the extracted rows match the PDF. Reject if figures differ without explanation or the messages run out before the check is complete. No result is recorded here.