Skip to main content

​
Classify Company Signals

To detect growth signals across target accounts programmatically, use Deepline’s 2-pass play: web search (Claude Code’s native web agent, Exa, or Parallel AI) queries the web for specific signal types (expansion, acquisition, hiring surges, regulatory changes), then call_ai classifies each company against your criteria with evidence and confidence scores. Define all your signal types in one natural-language prompt — adding a new signal type is one line, not a new workflow.
B2B buyers are 57% through the purchase decision before engaging a sales rep, according to CEB/Gartner research. Signal-based selling detects buying intent early, letting your team engage before competitors do.
The play monitors target accounts for specific business events — facility expansions, acquisitions, hiring surges, regulatory exposure, leadership changes — that indicate buying intent or trigger a sales motion. Static firmographic filters miss these changes. B2B data decays at roughly 22.5% per year (HubSpot via Cognism), which is why real-time signal detection matters.

​
How do I classify company signals with Claude Code?

Tell Claude Code what signal types you care about and point it at your target list. Deepline searches for evidence of each signal type per company and classifies matches with confidence scores.
“Classify these companies by growth signals: expansion, acquisition, new funding, regulatory pressure”
“Which of these companies opened new manufacturing facilities in the last 12 months?”
“Flag companies in targets.csv that are affected by new cybersecurity regulations”
With Codex: 
codex "Classify companies in targets.csv by signals: expansion, acquisition, hiring surge, regulatory pressure"

​
What does the signal classification workflow do step by step?

Two passes: targeted web search for each signal type, then AI classification with evidence extraction. Every detected signal includes a confidence score, evidence snippet, and source URL.
1

Search for signals (Pass 1)

For each company, web search queries for the specific signal types you requested: expansion news, acquisitions, regulatory filings, hiring surges.
2

Classify and tag (Pass 2)

call_ai reviews the search results and classifies each company against your signal criteria, tagging matches with evidence.
3

Write tagged output

Results include signal type, confidence level, evidence snippet, and source URL for each detected signal.

​
Which providers power the signal classification?

Two tools run in sequence. Signal types are defined in your prompt, not hardcoded — classify on any signal without changing templates. 25+ providers let you cross-reference signals against firmographic data for higher confidence.
  1. Web search — Claude Code’s native web agent, Exa, or Parallel AI runs targeted searches for specific signal types per company
  2. call_ai — Classification and evidence extraction from raw search results
You can classify on any signal: facility expansion, M&A activity, leadership changes, regulatory exposure, hiring in specific departments, product launches, or custom signals relevant to your industry. 
For each company in targets.csv, classify by these signals:
- EXPANSION: new office, facility, or market entry
- ACQUISITION: acquired or was acquired
- HIRING_SURGE: 20%+ headcount growth in last 6 months
- REGULATORY: affected by new compliance requirements

Write to signals.csv with columns: company, signal_type, confidence, evidence, source_url
Only include companies that match at least one signal.
Signal classification works best when you’re specific about what you’re looking for. “Growth signals” is too broad. “Opened a new manufacturing facility in North America in the last 12 months” is actionable.
Related: Company Research Brief | Competitive Landscape | Qualify & Score Leads

​
Frequently Asked Questions

​
How do I detect buying signals across target accounts?

Tell Claude Code: “Classify companies in targets.csv by signals: expansion, acquisition, hiring surge, regulatory pressure.” Deepline uses web search to find evidence for each signal type, then call_ai classifies matches with confidence scores and source URLs. Results land in a CSV with one row per detected signal.

​
How do I add or change signal types?

Signal types are defined in your prompt, not hardcoded into a configuration. Add a new signal type by adding one line to your prompt. Change classification criteria by editing the prompt and rerunning. This makes it easy to iterate on what signals matter as your GTM strategy evolves.

​
Can I classify signals in bulk from a CSV?

Yes. Pass a CSV of target companies to Claude Code. Each company goes through the search-then-classify pipeline, checked for every signal type you specified. Output includes only companies that matched at least one signal, with evidence and source URLs for each match.

​
What kinds of signals can Deepline detect?

Any signal with a public web footprint: facility expansions, M&A activity, leadership changes, IPO filings, product launches, regulatory exposure, hiring surges in specific departments, office openings, partnership announcements, and custom signals relevant to your industry. If it shows up in news, press releases, or SEC filings, Deepline can detect it.

​
How much does signal classification cost?

Each company uses 1 web search credit per signal type searched and 1 call_ai credit for classification. Searching 50 companies across 4 signal types costs approximately 200-250 credits.