Who is scaling deodorant on Meta now: Ad Library census, 2026-11-05

Research data. Ad lines quoted are competitors' published text.

How it was pulled, and what the numbers can and cannot say

• Tool. research/adlib/adlib_graphql.py driven on the AdsPower research profile (logged out), through a local copy raw/adlib/adlib_count.py that only adds a log line. Active ads, country US, keyword search on the domain or brand phrase, one query per row (the exact query is in the table). Queue scripts, logs and rows: raw/adlib/queue2.sh, queue3.sh, queue4.sh, *.new30.{log,jsonl}, summary summary.json / summary.txt (summarize.py).
• Why not adlib_hunt.py. Its --help and source show it runs the paid Apify actor apify~facebook-ads-scraper with the token in ecom-loop/.env. The brief and the standing rule both forbid paid Apify, so the free GraphQL scraper on the same research profile was used instead.
• "active" = Meta's own "~N results" figure printed on the Ad Library page at load (rounded by Meta). "?" means the page printed no count; where the first results page then returned 0 ads the brand has no matching active US ad for that query, and that is written as 0.
• "new 30d". The Ad Library returns no total for a date window. The start-date filter was sent, but the rows came back unfiltered and in relevance order (the Salt & Stone pull reached 200 rows: 70 started on or after Oct 6, oldest Feb 19). So: where every active ad was paged (small brands, and Dr. Squatch's deodorant query), "new 30d" is an exact count of active ads that started on or after 2026-10-06; where paging stopped at the cap (20 or 40 rows), it is "n of N sampled", a share and not a count.
• "vs Oct 7". The same query's "~N results" from the Oct 7 runs (../ads/adlib/census-20261007/, ../ads/adlib/20261007/, logged-in pool profile). Two runs of one query on Oct 7 sometimes disagreed (nativecos.com 21 vs 150, wearewild.com 640 vs 540); both are shown. The Oct 6 ScrollStash and Oct 1 page counts are in ../competitors/COMPETITORS.md.
• Pace. Four sessions shared the research profile this morning; paging ran at about one 10-ad page a minute, which is why big advertisers were sampled rather than paged.

Scaling table (sorted by active ads now)

Zero rows: each is the named query, active, US, on 2026-11-05, with the bootstrap loaded (lsd=ok) and the first results page empty (logs raw/adlib/<label>.new30.log). A zero means no matching active US ad for that query, not proof the brand runs nothing on Meta under another name.

Top 8 by evidence of scaling

1. Carpe: ~1,500 active, up from ~920; the most engaged deodorant creatives in AdSpy.
2. Dr. Squatch: ~590 active brand-wide; its deodorant inventory is almost wholly new (91 of 92 since Oct 6), rotating holiday and licensed drops.
3. Mountaineer Brand: ~640 active, up from ~170; a creator-whitelisted "ditch the chemicals" video at 12,262 engagement.
4. TONE: new; ~110 active, every sampled ad started Oct 13 or later; at Target; streamer-founded (AMP: "Where is AMP wearing TONE Vanilla Smoke?").
5. Salt & Stone: ~270 on its deodorant query, up from ~220.
6. Wild: ~500 active, every sampled ad new (seasonal gifting pages), down a little from Oct 7.
7. Lume: ~370 active, half of Oct 7's ~780; still churning (25 of 40 sampled new).
8. Old Spice aluminum-free via #OldSpicePartner creators to Walmart (~81 on the keyword, 18 of 20 new) and Native to Walmart, Amazon and click2cart (~100, 36 of 40 new).