Skip to content

gooseworks-ai/goose-skills / skills/marketing/research-audience

research-audience

Verified

Audience research: who the buyers are, in their own words, where they gather and how to reach them — plus named prospect lists when asked. Use when the marketer says "research our audience", "who should we target", "mine our reviews", "why do people buy us or leave", "where does our audience hang out"; or for lists: "find 50 right-fit leads with emails", "find retailers or accounts to pitch", "score these leads / who do we call first", "who's looking for an alternative to X", "tell me when a past champion changes jobs". Reads the brand's own customer data first and delivers a readable report saved to the brand. Never contacts anyone. Not for market size or trends — use research-market; not for writing the messages — use write-outreach.

references/watches.md

View on GitHub
# Standing watches — tell me when it happens

Load when the marketer asks to be told when something happens ("tell me when a past champion changes jobs", "watch these accounts for hiring", "flag anyone asking for an alternative to <rival>"), or when a list run ends and a watch would clearly help. A watch is not a repeated list: it checks a fixed set of people or accounts and reports only what changed.

## The three watches

| Watch | Checks | Signal | Cost per run |
|---|---|---|---|
| **Champion moved** | The champions list — people who praised the product publicly under their own name (with the link they published), or the marketer's own list of users and contacts | A new employer that passes the fit gates | **Paid:** one LinkedIn profile read per person checked (`/v1/linkedin/profile`), skipping anyone read in the last 7 days |
| **Account hiring** | A target-account list (up to ~100 accounts) | A new post for a role the product replaces or supports (list-sources.md § Hiring) | Free: careers pages and job-board searches |
| **Switch intent** | The rival names and pain phrases from the last list | New "alternative to <rival>", "moving off <rival>" posts by identifiable people at fitting companies | Free: Hacker News search and web search; paid only for Reddit or LinkedIn post search at volume |

## Champion moved — snapshot, then compare

1. **Baseline (first run):** for each champion, the profile URL, the company and title seen, and the date (`research/audience/watch/champions.csv` — people's data, never imported into the brain).
2. **Each check:** re-read profiles not read in the last 7 days. Compare the current company with the baseline after normalising both (drop legal and group suffixes such as inc, llc, ltd, gmbh, group, holdings; strip punctuation; lower-case). Different after normalising → a move. Skip a row when either side is blank or the read failed (say how many).
3. **Days since the move** from the current role's start date (year only → assume mid-year, say so). Fresh = under 120 days; best under 30.
4. **Does the new company fit?** Run the fit gates (fit-and-scoring.md). Pass → an alert row with the angle (write-outreach § Champion moved); fail → logged, no alert.
5. Write the new baseline with the check date. Alerts are keyed on **person + new employer**, so a champion who moves again alerts again.

## Account hiring and switch intent

State: the saved queries in `research/audience/watch/queries.md` and what was already seen in `watch/seen.csv` (post URL or job URL, date) — a post is alerted once. Account hiring: per account, search its careers page and the job boards for the mapped roles; a new post since the last check → alert with the post's URL and date, the tier (act today / this week / queue) and who to contact (one level above the role). Switch intent: run the saved queries over the period since the last check; an identifiable author at a fitting company → alert; anonymous posts → quotes for the bank, no alert.

## Setting one up

- Offer it only when it would help this brand, with the reason, a cadence and the **cost per run** (champion watch: about one paid profile read per champion; account hiring: about 2 web searches per account, free but run time; switch intent: about 18 searches): champions monthly; account hiring weekly or every two weeks; switch intent weekly.
- **Created paused**, through `automation_upsert`, per the `spend-gate` skill (`../../spend-gate/SKILL.md`) § C (scheduled runs): the marketer's yes names a **standing per-run credit ceiling** (champion watch: about one profile read per champion per run, from the credits log), and says the run itself uses credits. Free watches still need the yes; they cost only the run.
- Each run starts and ends with run-memory ([run-memory.md](run-memory.md), job `champion watch` / `account hiring watch` / `switch-intent watch`): the last record tells it what was already alerted and what each alert led to (from the ledger); the new record holds the counts (checked, moves or posts found, alerts, credits) — never names or profile URLs. A quiet run still writes its record.
- Each run writes `research/audience/watch/<D>.md` — what changed, nothing else — and delivers it through write-report only when something fired ("No moves among 42 champions this month" is one line in chat, not a report).
- Every alert row is ledgered (outcomes-ledger); the same person at the same employer is never alerted twice; suppressed people are never watched.

## Never

**Retention:** watch rows not alerted or contacted in 12 months are deleted, with a one-line note. Never watch private individuals, consumers or anyone who asked not to be contacted; never store emails or phones in a watch file; never scrape who liked or commented on posts; never run a paid watch without a standing ceiling.
add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks.Stars: 115kUpdated Claude CodeCodexVerifiedInstall