# Imagine AI Benchmark

> LinkedIn content benchmark for B2B companies: who posts, what lands, and how firms compare. Built by Imagine AI (useimagine.ai) as shared infrastructure between automated firms: your agent reads it the same way ours does.

## What is in it
- 800 tracked B2B companies, 2,008 founders and executives, 90,690 original LinkedIn posts across 18 industries, data to 2026-10-08, updated weekly.
- Every account gets a log-normal fit of its reactions: typical post (geometric mean), spread, a 95% interval and a position vs the research baseline.
- Posts are scored only against their own account (multiple of typical, z). Companies and people are compared at the account level.
- Baseline: research baseline (Imagine AI study of YC W24/S24 founders, 20,000+ posts). Study: https://useimagine.ai/blog/the-distribution-of-linkedin-engagement
- Method: lognormal-v1. Details: https://benchmark.imagineai.me/dashboard/methodology

## Access tiers
- Public (no key): bands and aggregates only. Where an account sits vs the research baseline (p50 / p75 / p90), industry percentiles, post-type shares, cadence as words, trends as percent. No typical-post numbers, no post text or links.
- Customer (sign in, or a workspace agent key from /app/settings/agents): everything, including post-level data, for the companies in your workspace landscape.
- Public MCP tools (no Authorization header): methodology, list_companies, company_fit_summary, category_breakdown.

## Connect an agent
- MCP (Streamable HTTP, JSON-RPC): POST https://benchmark.imagineai.me/api/mcp with Authorization: Bearer <token>. Server card: https://benchmark.imagineai.me/.well-known/mcp.json
- REST (read): https://benchmark.imagineai.me/api/openapi.json
- Tokens: sign in at https://benchmark.imagineai.me/login (customers) or ask Imagine AI.
- Each result includes meta: asOf (data edge), generatedAt, method, baseline, sample, sourceUrl, freshness. Quote asOf when the data is older than 30 days.

## Rules for interpreting numbers
1. Compare accounts by typical post (geometric mean reactions), never by arithmetic averages or totals.
2. Judge a post against its own account only (multiple of typical, z >= 2 is a breakout).
3. A company page and its people are different accounts.
4. "insufficient" = under 18 original posts at least 14 days old; do not conclude.

## Read tools
- list_companies (public: bands only): List companies in the benchmark catalog (customers, competitors, industry peers).
- get_company: One company: profile, tracked people (with post counts and active-poster flag), and when the page and each person were last scraped.
- search_people: Find tracked people (founders, execs, employees) by name, company, or role.
- get_posts: Fetch LinkedIn posts with reactions/comments/shares, category, media type, url and a text excerpt.
- account_fits: Fit each account (a person, or a company page, never pooled) to a log-normal on reactions: geometric mean (typical post), x/÷ spread, baseline percentile/band.
- company_fit_summary (public: bands only): A company's page fit plus the distribution of its people's geometric means (median, baseline band), top people and who is under the 18-post floor.
- industry_landscape: Every account in an industry (or a custom company set): per-company rollup (page fit + people median), population vs the research baseline, and category lift with the author effect removed (which post types beat each author's own typical post).
- posts_vs_account: One account's posts scored against that account's own log-normal: z-score, multiple of its geometric mean, label (breakout/above/typical/below/flop).
- competitor_moves: Recent competitor posts (company pages + their people, originals) grouped into the 10 B2B post types (hiring, product, company announcement, event, customer highlight...), each scored vs the author's own typical post, plus gaps: types competitors posted in the window that the primary company has not posted lately, and types that worked for them.
- share_of_voice: Weekly original posts (and reactions) per company in a landscape, plus each company's share of posts in the last 4 weeks vs the prior 4.
- content_matrix: Companies x the 10 B2B post types: share of each company's posts and the within-account lift (geometric mean of multiples vs the author's typical post).
- market_map: Each company as a point: posts/week (volume) vs team typical post (geometric mean, reactions), followers, quadrant (loud & landing, quiet but strong, loud & flat, quiet & flat) relative to the landscape median volume and the research baseline p50..
- momentum: Rolling 8-week geometric mean of each company's posts relative to their own long-run typical (1.0 = usual), weekly, with the landscape median.
- breakout_posts: Competitor and own posts in the last 30 days that scored z >= 2 against their own account (the data behind breakout alerts), with type and multiple..
- methodology (public: bands only): Imagine AI research rules (log-normal engagement, geometric means, research baseline 49/77/130).
- company_benchmark: Per-company volume table: posts, posts/week, people posting, median and outlier-robust engagement, top category, momentum.
- compare_competitors: A customer vs its competitors: presence (volume, share of voice), content (category mix vs each company's own median) and people (who carries each company).
- category_breakdown (public: bands only): Post counts and engagement per content category (product update, hiring, personal story, industry insight, ...) for companies or people..
- cadence: Posting cadence: posts and engagement by day of week and by month for companies or people..
- pipeline_status: Scraping pipeline health: active run (and whether it is stuck), recent runs, unrecovered failed/partial scrapes with errors, people whose scrape returned 0 posts, stale company pages / active posters, unclassified backlog..
- request_competitor: Track a competitor company for your workspace: verifies the LinkedIn company page, adds it if new, attaches it as your competitor and runs the first pull.
- list_events: What happened in a landscape since a date: breakout_post (a post at z>=2 vs its own account), new_monumental_post (funding/launch/big-hire style post), share_shift (share of voice moved >=5 pts in 4 weeks), new_company_tracked.
- connections: Firm-to-firm ties derived from public LinkedIn activity between tracked companies: reshares of each other's posts, mentions of the other firm or its people, shared events.

## Write tools (write scope, spend scraping credits)
- add_company: Add a company by its LinkedIn company URL (linkedin.com/company/<slug>).
- add_people: Add people (LinkedIn /in/ URLs) to a company.
- queue_scrape: Queue a pipeline run that scrapes company pages (plus their active posters) and/or specific people, then classifies new posts.
- cancel_run: Cancel a queued/running pipeline run (frees the one-active-run slot).
- subscribe_events: Store a subscription for this token: companies + event types (+ optional webhook_url for future push).
- unsubscribe_events: Deactivate one of this token's subscriptions by id.

## Humans
- Leaderboard: https://benchmark.imagineai.me/arena
- Full dashboard: https://benchmark.imagineai.me/dashboard