Docs

Likely adds formulas to Google Sheets and Excel that answer a question about each cell with a number you can sort on.

Install for Google Sheets

Install Likely from the Google Workspace Marketplace. Then open any spreadsheet, choose Extensions → Likely → Open, and enter your email to get a free key or paste one you already have.

If your Workspace admin blocks Marketplace installs, the add-on also installs by hand in about two minutes and works the same:

  1. Open your spreadsheet, then Extensions → Apps Script.
  2. Replace the contents of Code.gs with Code.gs.
  3. Add a file with + → HTML, name it Sidebar, and paste Sidebar.html into it.
  4. In Project Settings, tick Show "appsscript.json" manifest file in editor, open that file and paste appsscript.json.
  5. Save, reload the spreadsheet, then open Extensions → Likely → Open and allow the script when Google asks. It is your own copy, so Google shows an "unverified app" warning: open Advanced and continue.
  6. Enter your email to get a free key, or paste a key you already have. The key is saved in that spreadsheet.

Install for Excel

Get Likely from Microsoft Marketplace, or in Excel go to Home → Add-ins → More Add-ins, search for Likely and choose Add. It works in Excel on the web, Windows and Mac, and once added it stays. Click the Likely button on the Home tab to open the pane, then get a free key or paste one you already have.

If your organisation blocks store add-ins, load Likely from its manifest instead. It takes about a minute and works the same:

  1. Download likely.xml.
  2. In Excel (web, Windows or Mac), go to Home → Add-ins → More Add-ins.
  3. Open My Add-ins, then Manage My Add-ins → Upload My Add-in, and pick the file you downloaded.
  4. Click the Likely button on the Home tab to open the pane, then get a free key or paste yours. It's saved for Likely on that device.

On Excel for the web a manually uploaded add-in belongs to that workbook and can drop out when the session ends, so you may have to upload it again. On Windows and Mac it stays. Installing from Microsoft Marketplace avoids that.

Roll it out to a team

A Microsoft 365 admin can install Likely for other people, so nobody has to upload anything themselves. This works whether or not your organisation allows store add-ins, and unlike a manual upload it survives sessions.

  1. Sign in to the Microsoft 365 admin centre as a Global Administrator.
  2. Go to Settings → Integrated apps → Upload custom apps.
  3. Choose Provide link to manifest file and paste https://getlikely.app/excel/manifest.xml, or upload a copy of that file.
  4. Assign it to yourself for a trial, to a group, or to the whole organisation, then accept the permissions and finish.

Likely then appears on the Home tab in Excel for those people, on the web, Windows and Mac. Allow up to 24 hours for it to reach everyone.

Keys for a team

Deployment installs the add-in; it does not hand out keys. Each person still pastes one into the pane once, and it is stored for Likely on that device.

Give the whole team the same key rather than one key each. Rows are counted against the key, so a shared key means one pool and one bill: Team is 2,000,000 rows a month, which is what a team of any normal size will share. Separate free keys give everyone 2,000 rows and no way to see the total.

The trade-off is that a key is a credential. Anyone holding it can spend the quota, and anyone who leaves keeps it until you rotate. Keep it where you keep other shared secrets, and mail support@simbertech.com to have it replaced if it leaks. Proper team accounts are coming; if you need them now, tell us and we will prioritise it.

Formulas

Google SheetsExcelReturns
=LIKELY(range, question)=LIKELY.IS(range, question)Probability from 0 to 1 that the question is true for each cell
=LIKELY_CHOICE(range, question, labels, [with_confidence])=LIKELY.CHOICE(range, question, labels, [with_confidence])The best label for each cell. With TRUE, adds a confidence column
=LIKELY_SCORE(range, question, levels)=LIKELY.SCORE(range, question, levels)A score from 0 (first level) to n−1 (last level)

labels and levels can be a range (e.g. Labels!A1:A6) or a comma-separated list ("billing, technical, sales"). Levels go from lowest to highest.

=LIKELY(A2:A500, "Is the customer asking for a refund?")
=LIKELY_CHOICE(A2:A500, "Which team should handle this?", "billing, technical, sales")
=LIKELY_SCORE(A2:A500, "How upset is the customer?", "Calm, Annoyed, Furious")

Pass a whole column range rather than filling a formula down: one formula over a range is one request, and the results spill into the cells below. In Excel, filled-down single-cell formulas also work, since Likely groups them automatically.

Bulk run

Formulas take up to 5,000 cells per range. For bigger columns, select the column and use Bulk run in the Likely panel. It writes plain values into the first empty column on the right, so nothing recalculates and there are no timeouts. Choice runs also write a confidence column.

Tools

Writing good questions

What counts as a row

Each non-empty cell judged for a question is one row. Blank cells are free. If Likely has already answered the same text for the same question for your key, the answer comes from cache and is also free. The cache holds a one-way fingerprint of the question and cell text next to the answer, never the text itself, it is never read across accounts, and entries are deleted after 30 days. Cells longer than 8,000 characters are judged on their first 8,000.

Limits

Rows per monthFree 2,000 · Pro 250,000 · Team 2,000,000
Cells per formula5,000 (use Bulk run for more)
Cells per bulk run200,000
Labels per choice2 to 50
Levels per scale2 to 10

Errors

API

The same engine is available over HTTP with your key:

curl https://getlikely.app/v1/judge \
  -H "X-Likely-Key: lk_…" -H "Content-Type: application/json" \
  -d '{"kind":"probability","question":"Is this a refund request?","values":["Refund me please","Love it"]}'

kind is probability (0 to 1), choice (with labels) or score (with levels). The response has one result per value, or null for empty values.