Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.signalark.app/llms.txt

Use this file to discover all available pages before exploring further.

Most of Signal Ark’s signal detection is passive — the platform continuously monitors the web and surfaces activity as it happens. Targeted Discovery is different. It’s an on-demand, account-first scanning pipeline that actively hunts for specific signals on the live web for the accounts you care about most. Use it when you need to go deeper than the standard monitoring cadence allows.
Targeted Discovery is a credit-intensive feature currently gated behind a feature flag and requires the signals:create permission. Contact your workspace Admin if you don’t see it in the Discovery menu.

​
How a scan works

When you launch a Targeted Discovery scan, Signal Ark submits your selected accounts to an asynchronous pipeline that handles large account batches, respects external provider rate limits, and resumes gracefully if interrupted — no scan is lost due to a timeout or transient error.

​
Progressive query tiers

For each account, the engine runs three tiers of Exa AI query packs, starting with the most precise:
  • Tier 1 — High precision, lower recall: Tight queries that look for definitive signal evidence. If a strong match is found, the scan stops here.
  • Tier 2 — Balanced: Broader queries that cast a wider net when Tier 1 comes up empty.
  • Tier 3 — Maximum coverage: Catch-all queries designed to surface signals in edge-case sources. Used only when Tiers 1 and 2 produce no usable candidates.
This progressive approach maximizes the signal-to-noise ratio while keeping token costs under control. The engine scans across 9 core signal families, including hiring activity, funding events, executive changes, tech stack shifts, and competitive mentions.

​
Candidate evaluation

Every URL returned by Exa AI goes through a 4-dimension LLM evaluation before it’s accepted as a valid signal:
DimensionWhat it checks
Name matchDoes the content actually refer to the target company?
Domain matchDoes the context align with the company’s known domain?
Context consistencyIs the described event logically coherent?
Signal qualityIs this a genuine signal or SEO noise?
A candidate must score at least 60 out of 100 across these dimensions to proceed to insertion. Anything below that threshold is discarded.

​
Deduplication

Before a new signal is written to your account record, it passes through a 4-layer deduplication check:
1

Exact URL match

Rejects any candidate whose source URL already exists in the signal index.
2

Semantic title similarity

Compares the candidate’s headline against existing signals using semantic similarity to catch near-duplicates with slightly different URLs.
3

Payload fingerprinting

Hashes the full signal payload with SHA-256. Identical content from different sources is deduplicated.
4

Recent time-window exclusion

Prevents the same underlying event from being logged more than once within a rolling 7-day window.

​
Credit cost

Targeted Discovery charges 5 credits per account scanned in a batch. Before you confirm a scan, Signal Ark shows you the total credit cost based on the accounts you’ve selected.
Your credit balance is shown in the top navigation bar. If a scan would exceed your available credits, Signal Ark will warn you and let you reduce the account selection before proceeding.

​
Running a scan

1

Open Targeted Discovery

Navigate to Discovery > Targeted Discovery in the sidebar.
2

Select accounts

Use the filters to narrow your account list by ICP score, industry, signal recency, or account owner. Select the accounts you want to scan.
3

Review the credit estimate

Signal Ark shows a summary of how many accounts are selected and the total credit cost. Adjust your selection if needed.
4

Launch the scan

Click Run Discovery. The pipeline runs asynchronously — you’ll receive an in-app notification when it completes.
5

Review new signals

New signals appear on each account’s signal timeline, tagged with a targeted_discovery source label so you can distinguish them from passively detected signals.

​
Saving and rerunning searches

You can save any account filter configuration as a Saved Search for quick reuse. Saved searches appear in the Targeted Discovery sidebar under My Searches. To rerun a saved search, open it and click Run Again — the same account selection and filters are applied, and you’ll see a fresh credit estimate before confirming.
Create a saved search for your highest-priority ICP accounts and run it on a weekly cadence to keep their signal records current without manually rebuilding the filter each time.