Every B2B outbound team is fighting over the same 5,000 companies.
Apollo pulled them. Clay enriched them. Trigify caught the LinkedIn posts. ZoomInfo tagged the intent. And every competitor in your category got the exact same list, 60 days ago, from the exact same vendors.
The hiring signal is commoditized. "They just posted a Head of Growth role" used to be edge. Now it is the default input for every outbound AI agent on the market, which means the same 10 roles on any given Tuesday get pitched by 200 vendors inside 48 hours. Reply rates on hiring-signal outbound are dropping roughly 8% per quarter. If that is all you are running on, the math is against you.
The fix is not to buy better commercial data. It is to watch signals your competitors are not watching.
I spent 3 weekends mapping 7 of them. Every signal below is free or nearly free, publicly available, and under-used because most teams do not know where to look. Each one opens a buying window that commercial aggregators miss by 30 to 90 days, or drops entirely because it is too unstructured to ingest.
Below is the concept in full. The companion blueprint (Claude Code skill files for all 7 signals, the scoring layer, the schema) is free at the end of the post if you want to ship one this week.
Why custom beats commercial
Every commercial signal provider is a rollup of the same underlying public data. Apollo and ZoomInfo are LinkedIn scrapes plus domain enrichment. Bombora and 6sense are third-party cookie networks. Trigify and Parma watch LinkedIn post engagement. Crustdata watches headcount deltas.
These tools are not wrong. They are table stakes. The problem is that they are floors: everyone in your category pays for the same data, on the same cadence, and watches the same 5,000 accounts trigger every week. By the time your SDR sends the email, six competitors have already pitched the same CEO on the same event.
A custom signal wins on one of three axes:
Earlier. You catch the event before the aggregators do. SEC Form D filings show private funding rounds 15 to 90 days before Apollo's "recently funded" filter updates.
Deeper. You capture signal the aggregators drop because it is unstructured. The exact quote a CFO gave on the last earnings call about which tooling budget is growing.
Broader. You combine two or more public signals into a compound trigger no single tool surfaces. An 8-K announcing a new CFO plus a LinkedIn post about "rebuilding the finance stack" equals a buying window that neither signal alone would justify.
The 7 signals below all win on at least one of these axes.
The 7 signals
Most of these are unfashionable. Patent filings and court records sound like compliance work, not growth work. Commercial real estate is a category most SDRs ignore. That is the point. If the signal were obvious, it would already be commoditized.
Let me go deep on one, and then show how they stack.
Deep dive: SEC filings (EDGAR)
EDGAR is the most underrated public signal source in B2B. The SEC requires publicly traded companies, and any private company raising institutional capital, to file structured disclosures. Four of them matter for outbound:
Form D catches private placement filings. Due within 15 days of close. This is your earliest signal for a funding round, often 30 to 90 days before the press release and well before Apollo's filter catches up. Fresh cash is in the bank, the board wants to see spend, and the 30-120 day window after a Form D is the highest-intent buying period in B2B.
8-K catches material events. Item 2.01 is an acquisition. Item 5.02 is an executive departure or appointment. Item 1.01 is a material contract. When a new CFO or CTO is filed under 5.02, expect a full vendor stack review in the first 90 days. The "new-leader vendor flip" is one of the most reliable triggers in the business.
S-1 is IPO registration, 6-12 months before the IPO. Pre-IPO companies over-invest in governance, attribution, security, and financial tooling to pass their audit.
10-Q is the quarterly earnings filing. The managed-discussion section is where executives literally tell you which budgets are growing. Running an extraction pass over 10-Q text turns the filing into a list of stated priorities you can open with.
The data is free. The full-text search endpoint is https://efts.sec.gov/LATEST/search-index. Rate limit is 10 requests per second. You have to set a User-Agent header with a real email, or EDGAR will 403 you. That is the entire friction.
The Claude Code recipe is a single skill file. Configure it with your User-Agent email and an ICP CSV (company name, 10-digit CIK, domain). Run /edgar-signal 8-K 30 ./icp.csv and you get a CSV back: filing date, company, item codes, category (funding, acquisition, exec change, IPO prep, strategic comment), summary, source URL, and a suggested outreach angle per signal type.
The full skill definition (with step-by-step logic, the 10-Q priority-extraction prompt, and the ICP scoring output) is in the blueprint.
How you tune it for your ICP: edit the buying-window category table inside the skill to match what actually indicates a sale for your product. If you sell security tooling, weight 5.02 exec change lower but weight 1.05 costs associated with exit activities (restructurings) higher, because the new CISO usually rebuilds the stack first. The tuning is one table inside the skill, not a code change.
Trigger logic: only route to outbound when ICP match confidence is above 0.85, category is one of FUNDING, EXEC_CHANGE, ACQUISITION, or IPO_PREP, and the filing is inside the last 45 days. Send 10-Q strategic-comment matches to a manual review queue. The quotes are rich but need human judgment before you use them.
Stacking is where the math gets interesting
One signal with 15% precision means you waste 85% of outreach. Two independent signals firing in the same 30-day window pushes compound precision to 70-80%. Three firing at once is functionally a warm introduction.
The art is in picking signals that measure different things, not the same thing two ways. Form D plus 10-Q is redundant (both financial). Form D plus speaker-list is orthogonal (financial plus stated priorities).
Three example stacks:
Funding-Plus-Intent. Form D fires, plus the company's head of growth engages on your topic on LinkedIn, plus the website adds a tag in an adjacent category. Three different axes, same 30-day window. Route to a senior AE with a founder-signed email.
Exec-Change-Plus-Priority. 8-K 5.02 fires, and within 45 days the new exec shows up on a conference speaker list. Read their talk abstract, open with their stated agenda. This one closes.
Competitive-Displacement. Stack-diff detects the company removed your competitor's tag, and someone at the company posted publicly about dissatisfaction. You have about 7 days before every competitor sees the same post.
The scoring weights in the blueprint give each signal a base weight and a decay half-life. Form D scores 10 for 45 days, then halves. 8-K exec change scores 8 for 30 days. Patent R&D acceleration scores 6 for 90 days. Weighted sum per company per week. Anything scoring above 12 enters the outbound queue.
How to start: pick 2, ship in a week
Do not try to build all 7 at once. I have watched two clients attempt that and stall by week 4. Pick 2 signals and ship them.
For most B2B SaaS selling to $10M+ ARR targets, the starting pair is Form D and conference speaker lists. Form D because it is the most reliable single buying-window signal and covers 80% of institutionally-funded targets. Speaker lists because executives literally tell you their priorities in the abstracts, and the scraping surface is small.
Day 1: install Claude Code, drop in the edgar-signal skill, configure your User-Agent email, point it at your ICP CSV.
Day 2: triage 10-20 Form D matches. Feed the top 5 into your outbound sequencer.
Day 3: pick 3 conferences your ICP attends in the next 90 days. Run the speaker-scan skill.
Day 4: cross-match, triage, add to pilot cohort.
Day 5: measure. If the pilot reply rate is above 2x your baseline, you know the stack works for your specific market.
Add one signal per week after that. By week 7 you are running all 7 with a unified scoring layer. By the end of the year you are the only vendor in your category with a signal stack this deep, and your outbound reply rates will show it.
Get the full blueprint
This post is the concept. The blueprint is the implementation: all 7 signals with full Claude Code skill files, the unified signals schema, the weighted scoring logic, the MVP and full-stack architecture diagrams, and a "what not to do" list I wrote after watching teams mis-wire this.
Comment "STACK" on the LinkedIn post below and I will DM it to you, or if you landed here from the newsletter, grab it directly below.
If you want help applying it
Building the stack is the easy part. Choosing which 2 signals matter for your ICP, wiring them against your existing CRM and sequencer, and writing the sequence that fires when each signal hits is where most teams stall. Vertical SaaS selling into hospitals does not run the same stack as a horizontal dev tool.
If you want a 30-minute second pair of eyes, book below:
I run The GTM Architects. We build outbound engines for B2B software companies and signal stacking is core to every engagement. If the stack maps to your market, I will tell you which 2 signals to build first and what 90-day reply-rate lift looks like. If it does not map, I will tell you that too.
Either way, you walk out sharper than you walked in.
Until next week,
The GTM Architects