apify-influencer-brand-collabs

Influencer–Brand Collaborations

Surface Instagram branded-content partnerships by chaining four Apify Actors against Meta's Ad Library. Distilled from the production influencer-brand-collabs mini-tool.

When to use

• "Who has Nike paid to promote them this quarter?"
• "What brands does @bellahadid do sponsored posts for?"
• Auditing an account's branded-content history
• Building a competitor's influencer roster

Don't use for: organic mentions or tags (use a hashtag/mentions scraper), TikTok or YouTube collabs (different platforms), generic competitor ads (query Meta Ad Library directly).

Inputs to gather

1. Instagram handle or URL — @adidas or https://www.instagram.com/adidas/
2. Lookback window — days; default 90
3. Enrichment toggles (each adds cost + time):
   • Content insights — likes, comments, views per collab
   • Profile enrichment — followers, bio, verified status of the other side

Direction (brand vs creator) is detected empirically. Do not ask.

The pipeline

Call each via mcp__claude_ai_Apify__call-actor. Use mcp__claude_ai_Apify__fetch-actor-details first if you've never run one of these and want the exact input schema.

Step 1 — Resolve the target

// actor: apify/instagram-profile-scraper
{ "usernames": ["adidas"] }

Grab fbid from the first item. No fbid → can't query Ad Library → stop and tell the user. Most common cause: private account.

Step 2 — Build the Meta Ad Library URL

https://www.facebook.com/ads/library/branded_content/?id={fbid}&query={username}&target=instagram&start_date={YYYY-MM-DD}&end_date={YYYY-MM-DD}

Date range = the lookback window (default 90 days, ending today).

Step 3 — Fetch collaborations

// actor: apify/brand-collaboration-scraper
{ "startUrls": ["<ad library url>"], "resultsLimit": 10 }

Schema is fixed: every item has creator (always the influencer side) and brandPartners[0] (always the brand side). Do not try to infer direction from these fields.

Step 4 — Detect direction empirically

Count how often the target username appears on each side of the results:

• target appears more on creator side → target is the influencer; results are the brands
• target appears more on brandPartners side → target is the brand; results are the creators

⚠️ Do not use isBusinessAccount to infer this. It's unreliable — e.g. @fifaworldcup is a business account but appears as the creator of its own branded content.

Step 5 — (optional) Content metrics

Split collab URLs by type:

• /reel/... → reel scraper
• /p/... or /tv/... → post scraper

Match back to collabs via shortcode in the URL: /(p|reel|tv)/([A-Za-z0-9_-]+).

Engagement formula: likesCount + commentsCount + (videoViewCount ?? videoPlayCount ?? 0).

Run the two scrapers in parallel — they're independent.

Step 6 — (optional) Enrich the result side

Collect unique usernames from the side that is not the target. Then:

Only enrich the side the user actually cares about. The input handle is already known.

What to present

After aggregation, surface:

• Headline counts: total collabs, unique partners, total engagement (if metrics enriched)
• Top 5 collabs by engagement — only meaningful when content insights were toggled on
• Content-type mix: Post vs Reel; Reels usually dominate engagement
• Weekly timeline across the date range — spikes reveal campaign launches
• Per-partner card (when profiles enriched): handle, full name, followers, verified, category, collabs in this run, avg engagement

For who-questions, the partner list alone is enough. Metrics only matter for which-was-best-questions.

URL parsing

Strip Instagram's _u/ and _n/ deep-link prefixes before extracting the handle:

These slugs are not usernames — skip them: explore, reels, stories, direct, accounts, about, p, reel, tv, tags, locations, _u, _n.

Pitfalls

• Target is private → profile scraper returns no fbid. Bail early with a clear message.
• No results → try in order: widen the date range, double-check the handle (strip _u/), confirm the account actually runs branded content. Meta only indexes ads they've classified as branded content.
• Engagement is all zeros → user skipped content enrichment. Offer to re-run with it on.
• Direction looks wrong in the output → re-check the empirical count. Don't trust isBusinessAccount.
• Brand collabs with no metrics are still answer-shaped for who questions — don't gate the whole flow on enrichment.

Cost & time

Full 4-actor run: ~3–5 minutes, a few cents of Apify compute. Order of magnitude:

If the user just needs a partner list, skip both toggles.

Reference implementation

Production route this skill was distilled from: mini-tools-main/src/app/api/tools/influencer-brand-collabs/route.ts — full transformation logic, error handling, and slimmed display shapes for each scraper's output.