Matt Cox — Growth profile

Intent

I’m building a LinkedIn network of marketing and growth leaders at industrial and manufacturing companies. The people I want to connect with run demand generation, content, and revenue marketing inside complex-sale B2B operations — usually small-to-mid-market manufacturers, machinery and equipment makers, industrial chemicals, and adjacent heavy-industry verticals.

They tend to be people who think publicly about how industrial buying actually works (long cycles, engineering buyers, channel partners, trade shows, technical content) rather than generic SaaS-style growth advice. I want connections that lead to substantive conversations and mutual engagement on the platform, not connection-count vanity.

Audience search

These fields feed harvestapi/linkedin-post-search directly on every run. Edit freely — the audience tool re-reads this section on each invocation.

Search queries (`searchQueries`)

Topics whose posts should surface candidate authors. Mix narrow phrases with broader topic terms so we capture both keyword hits and topical adjacencies.

“industrial marketing”
“manufacturing marketing”
“B2B industrial”
“industrial growth”
“industrial demand generation”
“industrial content marketing”
“OEM marketing”
“engineering buyer”

Author keyword filter (`authorKeywords`)

Boolean expression applied client-side via substring match against author headlines/job titles. Restricts results to authors whose role lives in marketing/growth, regardless of what they happened to post about. The actor’s own authorKeywords field is too strict to be useful — see commit history for why we filter client-side.

marketing OR growth OR "demand gen" OR "demand generation" OR GTM

Industry IDs (`authorsIndustryId`)

LinkedIn industry codes from harvestapi’s reference table. The set below covers the core of industrial / manufacturing. Trim entries that surface noise; add adjacent codes if recall is too low.

25 — Manufacturing (parent)
55 — Machinery Manufacturing
135 — Industrial Machinery Manufacturing
147 — Automation Machinery Manufacturing
901 — Agriculture, Construction, Mining Machinery Manufacturing
935 — Engines and Power Transmission Equipment Manufacturing
923 — HVAC and Refrigeration Equipment Manufacturing
928 — Metalworking Machinery Manufacturing
54 — Chemical Manufacturing
840 — Fabricated Metal Products
807 — Primary Metal Manufacturing
1029 — Transportation Equipment Manufacturing
146 — Packaging and Containers Manufacturing

Reference table: https://github.com/HarvestAPI/linkedin-industry-codes-v2/blob/main/linkedin_industry_code_v2_all_eng.csv

Geographic policy

Best-effort, applied client-side after Apify returns. The actor doesn’t expose a location filter and only returns the author’s headline (not their profile location), so this filter is a substring match against the headline.

Include keywords:

United States
USA
Canada
North America

Exclude keywords:

India
Bangalore
Mumbai
Delhi
Hyderabad
Chennai
Pune
Pakistan
Bangladesh
Philippines
Manila
Nigeria
Lagos

Strict mode: false

(With strict: false, profiles whose headline matches no include keyword are kept by default unless they hit an exclude keyword. Flip to true to require a positive include match. The current setting is permissive because most US/Canada profiles don’t put their country in the headline.)

Time window (`postedLimit`)

week

(Daily runs want fresh signal. Bump to month if the candidate pool feels too thin.)

Sort (`sortBy`)

relevance

Content type (`contentType`)

all

Max posts per query (`maxPosts`)

(Caps each search query at 30 results. 8 queries × 30 = 240 posts max per run, ~$0.36/run at the actor’s $1.50/1000 rate. Bump up if recall feels thin after a few daily runs.)

Growth cadence

Operational policies that govern the daily workflow — read by the pipeline tools and the daily-brief composer.

Daily caps

Connection requests: 15/day
DMs: 3/day

(Hard caps in code, not just a guideline. Spread the requests across the day, never burst.)

Active hours

8am - 6pm Mountain Time, weekdays only

(Bounds when Fitzy can DM the rep with prompts and, in Phase 2, when Unipile actions can fire.)

Pre-connect warmup

Default preference: yes. Suggest 1 engagement (comment or reaction) on a target’s content before surfacing the connection request. Override per-person when there’s a reason to skip the warmup (e.g., we already know each other from elsewhere).

Connection note policy

Send a 200-character note with each request. Tone per voice.md §3 Outreach row. Structure: specific reference to their content → one genuine observation → soft opener that invites response without demanding one.

Follow-up DM policy

After acceptance, wait 1-2 days before sending an initial DM. 2-4 sentences. Same Outreach tone, no pitch, no Calendly, no ask for a call in the first message.

Pipeline status reference

Documentation of the 5 states. Records flow cold → engaged → requested → connected, with dropped as a terminal off-ramp.

cold — surfaced via audience scan, no action yet. Default state when a new record lands.
engaged — touched their content (comment, reaction). Pre-connect warmup. They’re now aware of you (or at least, we know we noticed them with intent).
requested — connection request sent. Awaiting their response.
connected — they accepted. In your network. Active vs. quiet is a computed property of last_touch_at, not a stored sub-status.
dropped — explicitly removed. Won’t be re-surfaced by future audience scans. Tombstoned with a reason field so we remember why. Use cases: declined request, expired request, profile turned out to be wrong fit, manual skip.

stale is a derived flag, not a status: status == connected AND last_touch_at > 14 days ago. The daily brief surfaces stale connections as “people to re-engage” but their stored status stays connected.

Disqualifiers

Drop a candidate during the judgment pass even if they hit the search filters, when any of these apply:

Profile is clearly automation / repost-only with no original writing
Crypto, web3, or DeFi content as the primary output
AI-bro productivity content (“I asked ChatGPT to…” style) as the dominant output
Connection collectors (10K+ connections with low original-post engagement ratio)
The author themselves isn’t in a marketing, growth, demand-gen, revenue, or related commercial role. The most common form of this is HR / Talent Acquisition / recruiters at industrial companies posting hiring announcements for marketing roles. The company is operationally interesting (active vacancy) but the author is not a connection target — they’re hiring for the person we’d want to connect with, they’re not that person.
LinkedIn influencers whose feed is mostly generic leadership platitudes
Posts are entirely about their own company’s product without a substantive perspective beyond the pitch
Cold-outreach service vendors hawking “I’ll book you 50 meetings with manufacturing VPs”

Tone notes

What an obviously-good pick looks like:

Posts substantively about industrial buying behavior, complex sales, or technical content marketing
Original thinking — not just aggregating other people’s takes
Engagement quality is decent (comments suggest real conversation, not emoji noise)
Title is real — VP / Director / Head of Marketing or Growth at an actual industrial company
Bonus: small-to-mid-market firms (50–1,000 employees). They’re more reachable than F500 brand teams.

Borderline cases:

Posts good content but only once a month → still worth surfacing; flag the cadence so I know it’s a low-volume connection
Title is Founder or CEO of a niche industrial firm → fine if their writing lives in the marketing/growth zone
Works at an agency serving industrial — fine if their content is in-category; less interesting if they’re brand-agnostic generalists

Clear nos:

Anyone in the disqualifier list above
People whose only “industrial” signal is one repost; primary content is elsewhere