Signal-based outbound playbooks: a GTM engineering guide with Clay and Unify

Signal-based outbound is the operating model where every send is triggered by a time-decaying buying signal, not by a static list. The engineering work sits in Clay and Unify: signal capture, entity resolution, sequencer hand-off, and sunset rules. This is a technical playbook for GTM engineers building production signal pipelines for B2B SaaS motions.

The architecture in one paragraph

Every signal is a table with a company (or person) primary key, a captured-at timestamp, and a signal-specific payload. Tables join in Clay on normalized domain (Unify handles the equivalent inside its own graph). A master table flags accounts with two or more active signals in a rolling 30 day window. That master table feeds Outreach, Salesloft, or Instantly through a sequencer webhook. Everything sunsets on a signal-type-specific TTL.

Signal capture in Clay

Clay's HTTP API and 100+ waterfalls are the workhorse. The signals worth capturing first for a technical B2B SaaS ICP:

• Tech installs and removals - BuiltWith, Wappalyzer, or HG Insights waterfall keyed on domain. Store install date so you can trigger on < 30 day installs only.
• Hiring surges - LinkedIn job scrape or LinkUp for role-specific counts. Trigger when open roles in a target team exceed a per-segment threshold.
• Funding events - Crunchbase or Harmonic. Filter to Series A through C for mid-market motions.
• Champion movement - UserGems or Champify write back into a Clay table keyed on prior_company and new_company.
• Product and pricing changes - Diffbot on target pricing pages, refreshed weekly. Emit a signal on non-trivial diffs.

Intent signals with Unify

Unify captures first-party and third-party intent that Clay alone cannot: G2 category views, warm accounts from your own site with reverse IP resolution, and the topic clusters the account is researching. Push the raw event stream from Unify into a dedicated Clay table so intent joins the same master table as your firmographic signals. Never sequence on raw Unify intent alone. Compound it with a job change, a hiring surge, or a champion move, or reply rates crater.

Entity resolution: the hard part

Signals arrive keyed by name, domain, LinkedIn URL, or company ID. Normalize on the registrable domain (etld+1) before joining. Handle subsidiary and parent mappings explicitly with a manual override table. Skip the join on any row where domain is missing or a public-inbox provider (gmail, outlook). This is where 60 percent of signal-based outbound programs quietly fail: two systems fire on the same account and dedupe never happens, so the prospect gets three touches in a day.

The sequencer hand-off

Master table rows fire a webhook into the sequencer with signal_type, signal_captured_at, and a personalization payload (funded_amount, new_role_title, competitor_removed). Sequence selection is deterministic: a lookup table maps signal_type to a specific sequence and first-line template. Never let the sequencer pick the play. Log the assignment back into Clay so reporting closes the loop.

Sunset rules

Every signal decays. Suggested TTLs from production motions:

• Job changes: 60 days from start date
• Funding: 120 days from announcement
• Hiring surges: 30 days from last scrape
• Tech installs: 45 days from install date
• Intent surges: 14 days from last event

Without sunsets the master table becomes a static list wearing a signal costume. Reply rates fall to firmographic-baseline within a quarter.

Reporting that matters

Instrument three metrics per signal_type: reply rate, meeting conversion, and weighted pipeline per contact touched. Kill any signal running below the firmographic baseline after 200 touches. Double budget on signals running 2x baseline or better.

Where this fits

Signal-based outbound is the highest-leverage output of GTM engineering and the operating substrate under a modern outbound prospecting motion. For the taxonomy of signals themselves, see 12 buying signals that beat firmographic outbound. For the buying decision framing, see revenue engine vs buying meetings.