Marketing

Brand context

Fill in the brand profile, customer types, customer quotes, swipe ads and documents the AI works from.

View as Markdown

Marketing → Brand (/m/marketing/brand) is one page with everything the AI should know about one brand. All of it is optional; each part makes the scripts better. The page has three sections, What you sell, Who it is for and Your voice, and under More two more: Guardrails and Examples and documents (the swipe file and documents, as links with how many each holds). An empty section says what filling it in helps with. Use the brand switcher at the top right to work on another brand; the choice stays in the address bar (?brand=).

The profile fields (products, offers, selling points, both voice fields, banned claims and notes) share one bar at the bottom: change a field and Save and Discard changes appear; Profile saved confirms it. Customer types and quotes save per action. If you leave the page or switch brand with unsaved changes, Discard unsaved changes? asks first: Discard changes or Keep editing.

Brand used to have five tabs. Old links to them still work: /m/marketing/brand/profile, /m/marketing/brand/customer-types and /m/marketing/brand/voice open this page at the matching section.

What you sell

Each field takes one item at a time: type it and press Enter.

FieldWhat goes in
ProductsOne product or service per item
OffersCurrent deals and guarantees
Selling pointsWhat sets the brand apart

Logo sits under the list: Choose logo picks an image from your media library, Change and Remove replace or clear it. Scenes can attach it to the intro or the end.

Who it is for

Who it is for holds the brand's customer types: the kinds of people it sells to, in their own words. Click Add customer type and fill in Name, Pains, Desires, Objections (one per line each), Awareness and Notes.

Awareness is how much the person already knows: Unaware, Problem aware, Solution aware, Product aware or Most aware.

Deleting a customer type keeps the quotes linked to it, but they lose the link.

Your voice

Your voice has Voice: do (how the brand sounds) and Voice: don't (words and tones to avoid), saved with the bar at the bottom.

Voice of customer

Voice of customer, in the same section, is a list of real quotes from reviews, emails and calls, so the AI can reuse your customers' wording. Each quote has its Quote text, an optional Theme, an optional Customer type and a Source: Added by hand or Imported.

  • Paste quotes adds one or many at once: paste one quote per line; empty lines are skipped.
  • Import opens the shared import flow for a CSV, TSV, .xlsx or .txt file (one quote per line). Columns such as Review, Comment or Yorum are recognised as the quote and Theme or Topic as the theme; when a column is unclear the sheet asks you to choose. The same quote is never added twice. See Imports.

A quote can be up to 5,000 characters and a theme up to 255; a longer row is skipped with its row number, and the other rows land.

Guardrails

More → Guardrails holds Banned claims (claims that must never appear in a script, one item each) and Notes. They save with the same bar as the other profile fields.

Swipe file

Swipe file keeps ads worth learning from. Open it from More → Examples and documents → Swipe file (/m/marketing/brand/swipe-ads); Brand at the top takes you back. Click Add swipe ad and fill in Title, Link, Transcript, Notes and Video length (seconds, optional). Each link can be in a brand's swipe file once: saving a link that is already there shows "This link is already in your swipe file." under Link.

Tags are grouped by type: click Add tag group, name the group under Tag group (for example "hook" or "format") and type values under Values, pressing Enter after each.

Import brings in many swipe ads at once from a CSV, TSV or .xlsx file: one ad per row with a title, a link or a transcript. A link that is already in the swipe file is updated, not added twice. See Imports.

Video and audio

Video or audio file takes a recording of the ad (MP4, MOV, WebM, MP3, M4A, WAV, OGG, FLAC or AAC, up to 20 MB). Borkol writes the transcript for you on your OpenRouter key, with zero data retention (with it off, only providers that do not train on your data), using the Transcription model from Settings → AI pipeline → More. The file is kept privately with the ad.

While it works the row shows Transcribing; when it is done the transcript replaces what is in the Transcript field, links removed. Other states:

StatusMeaning
Transcription failedThe file could not be transcribed. Hover the pill for the reason, for example that your OpenRouter account is out of credits
Waiting for monthly limitThe brand reached its monthly limit; raise it under Costs

For both, the row menu has Try again. A transcription costs a little and counts toward the brand's monthly limit; it shows under Costs → More → By model as Transcription.

Watch competitors

Watch at the top of the swipe file follows competitors in the Meta Ad Library. Add an entry with Type (Page or Search term), the Page id or Search term, and Countries (EU countries only: the Ad Library covers ads delivered in the EU; a new entry starts with the countries of your latest one), then click Add. A brand watches up to 20 entries; a second copy of one you already have says "You already follow this."

  • Page id is the number, not the page URL. Open the page in the Ad Library (facebook.com/ads/library), search the advertiser and open their page: the address ends in view_all_page_id= followed by the id (for example view_all_page_id=123456789012345).
  • Search term finds ads whose text contains the words; it compares without case or extra spaces.

Saving an entry starts an Ad Library check when the connection works, with the notice Checking the Ad Library: 1 of 3; after that Borkol checks every day. Without a working connection the dialog says so and the entries are read once it works. The X next to an entry removes it.

The ads land in the swipe file with the page name under the title and four extra columns: Winner score (0 to 100, always marked Estimate), Days running, Copies (active ads of the same text on that page) and Reach (EU reach). The score comes from days running, copies, reach and copies started in the last 14 days. It is an estimate: long-running means the advertiser keeps paying for the ad, not proven sales. Filter with Source (Added by you or Meta Ad Library) and Delivery (Running or Stopped); sort by Winner score or Reach.

Ad Library ads are read-only: they update with each sync, so they have no Edit or Delete. Click one for its page, score, ad text and Open in the Ad Library; you can upload its video under Video or audio file for a transcript and its length. To stop following an advertiser, remove its entry under Watch. An import never changes them (see Imports).

Once ads are scored, Winner rate by page (estimate) above the list shows per page how many of its ads score 70 or more ("2 of 3 winners · 67%"), with Too little data under 3 ads. Scripts, angles and scenes get the top 8 ads that ran 14 days or more, ranked by winner score, in a section labelled as an estimate, apart from your own results.

Documents

Documents holds your marketing plans, brand guides and research. Open it from More → Examples and documents → Documents (/m/marketing/brand/documents); Brand at the top takes you back. Click Upload document, give it a Title, pick a Kind (Marketing plan, Brand guide, Research or Other) and choose the File.

  • Accepted files are PDF, DOCX, Markdown and plain text, up to 20 MB.
  • Borkol extracts the text from the file and stores both. The list shows File name, Size and Extracted characters; opening a document shows the Extracted text.
  • At most the first 200,000 characters of extracted text are kept; anything after that is left out of the extracted text and the proposals.
  • Some files have no usable text: a scanned PDF or an encrypted PDF has no text layer, and a PDF over 10 MB is stored without text extraction. The file is still stored, and the page says "No text could be extracted. Scanned or encrypted PDFs have no text layer. The file is stored." Such a document proposes no knowledge.
  • Each heading in the text becomes a proposed knowledge entry (kind Brand fact) holding the text under that heading. A text without headings is split into paragraphs instead. A document proposes at most 40 entries. After the upload, a message such as 3 knowledge entries proposed for review appears with Review proposals, which opens the list of proposed entries. Nothing is used until you accept it. See Proposed entries.

Deleting a document removes the file and its extracted text. Knowledge entries created from it stay.

Setup score

Each brand has a setup score, shown per brand under Settings → Brands (for example "50% complete"). The Brand page no longer shows it: empty sections say what is missing instead. It counts six parts, each worth the same:

PartDone when
Brand profileProducts or Selling points is filled in
Customer typesThere is at least one customer type
Voice of customerThere is at least one quote
Swipe fileThere is at least one swipe ad
DocumentsThere is at least one document
KnowledgeThere is at least one knowledge entry, in any status