Northpeak NeedShift Intelligence

Know when a company's buying window opens — and why

Northpeak NeedShift Intelligence monitors public company evidence and detects emerging B2B needs before they become obvious. It now goes beyond intent scoring by maintaining a Buying Window over time: when the opportunity opens, strengthens, cools, or closes.

Track signals across hiring, security, AI/data, sales expansion, customer support, operations and procurement — then get the likely need, confidence, buying-window state, role-level buying committee, evidence gaps and recommended next action.

Built for: B2B sales teams, lead-intelligence workflows, prospecting systems, agencies and sales automation.

Instead of: "Company X has an intent score."

You get: "Company X's Security / Identity buying window just opened. Confidence 83. New evidence explains why, the likely buying roles are identified, and the engine tells you what still needs confirmation before outreach."

Buying Window Intelligence

NeedShift turns repeated evidence checks into an explicit opportunity lifecycle:

• OPENED — evidence crossed the configured confidence threshold.
• STRENGTHENING — the need remains qualified and meaningful new evidence or confidence growth appeared.
• OPEN — the need remains qualified without a major directional change.
• COOLING — the need remains visible but evidence is weakening.
• CLOSED — a previously open buying window fell below the confidence threshold.
• BASELINE / NOT_OPEN — there is not yet a qualified buying window.

Every result can include buyingWindowDrivers, previousBuyingWindowState, confirmationNeeded, and nextCheckDays. This makes the output useful for deciding not only what a company may need, but when an account deserves action and what evidence is still missing.

Buying committee without invented people

For a qualified need, buyingCommittee provides role-level hypotheses for the economic buyer, champion, technical evaluator and procurement role. These are explicitly marked ROLE_HYPOTHESIS; NeedShift does not invent named contacts when direct public evidence is absent.

What it does

For each company, the Actor fetches user-supplied public URLs such as careers, security/compliance, product, newsroom, operations, procurement, or technology pages. It creates a repeatable evidence snapshot, classifies likely need domains, checks counter-evidence, compares the current evidence with the previous run, and returns decision-ready intelligence.

The engine tracks both evidence hashes and detected signals. Repeated runs can therefore distinguish newly observed signals from evidence already known to the monitor.\n\n### Multi-source causal evidence\n\nNeedShift also classifies public evidence into strategic event classes: Hiring, Funding, Executive Change, Tech/Vendor Change, Expansion, and Contraction. These events are not treated as standalone intent scores. They are connected to the detected need and Buying Window through an evidenceGraph, with supporting or counter direction, temporal novelty, and source traceability.\n\npredictiveSignals highlights newly changed strategic events, while anomalyScore measures how unusual the current evidence change is relative to the persisted baseline. This lets downstream workflows distinguish routine evidence from a meaningful shift without paid enrichment, browser automation, or an external LLM.

Core output

Each company analysis can include:

• detectedNeed and transitionStatus.
• buyingWindowState, previousBuyingWindowState, and buyingWindowOpened.
• buyingWindowDrivers — the new signals and changed sources driving the window.
• buyingCommittee — role-level economic buyer, champion, technical evaluator and procurement hypotheses.
• confirmationNeeded — evidence gaps and cautions to verify before action.
• nextCheckDays — suggested monitoring interval based on window state.
• confidence, previousConfidence, and confidenceDelta.
• newSignals and temporalEvidence — what is new or changed relative to persisted state.
• supportingEvidence and counterEvidence with source URLs and evidence hashes.
• alternativeHypotheses, likelyBuyer, urgency, causalChain, and recommendedAction.
• evidenceGraph — causal links from strategic events to the detected need, including support/counter direction.\n- predictiveSignals — newly changed Hiring, Funding, Executive, Tech/Vendor, Expansion, or Contraction evidence.\n- anomalyScore — a 0–100 measure of how much fresh evidence changed in the current run.\n- eventSignals — normalized strategic event classes detected from the supplied public sources.\n- evidenceSources — fetch status, HTTP status, hashes, and latency for traceability.

Example input

{
  "companies": [
    {
      "name": "Example company",
      "domain": "example.com",
      "evidenceUrls": [
        "https://example.com/careers",
        "https://example.com/security"
      ]
    }
  ],
  "monitorId": "weekly-b2b-watch",
  "outputMode": "all",
  "minConfidence": 45,
  "maxEvidenceUrlsPerCompany": 8
}

Use a stable monitorId for repeated runs. State is persisted per monitor and company. Changing the monitor ID intentionally creates a fresh baseline.

Recommended workflow

Supply 2–8 stable public URLs per company and run the same monitor on a schedule. The first run establishes state; later runs compare current signals and page hashes with stored evidence. nextCheckDays suggests a tighter cadence for newly opened or strengthening windows and a slower cadence when no window is open.

Use outputMode: "all" when you want every analyzed company returned. transitions_only returns qualifying needs and also preserves COOLING or CLOSED windows so downstream systems can react when an opportunity deteriorates.

Interpreting results

A high confidence score means the supplied evidence strongly matches one of the supported need domains. It is not proof that a purchase will occur. Verify important evidence before sales outreach or operational decisions. counterEvidence, confirmationNeeded, alternativeHypotheses, newSignals, and source hashes make that verification easier.

A null detectedNeed is a valid result: the supplied evidence did not clear the requested confidence threshold.

Supported need domains

The deterministic engine covers Security / Identity infrastructure, Data / AI infrastructure, Sales / GTM expansion, Customer support infrastructure, Logistics / Operations expansion, and Finance / Procurement infrastructure.

Reliability and limits

The Actor is HTTP-first and does not require browser automation, residential proxies, paid enrichment APIs, LinkedIn cookies, or an external LLM API. Public pages can still block automated requests, change markup, disappear, or contain ambiguous language. Failed evidence URLs are reported in evidenceSources rather than silently treated as positive evidence.

The Actor accepts up to 100 companies per run and up to 20 evidence URLs per company; the configurable default is 8 URLs per company.

Pricing

NeedShift uses pay-per-result pricing. The current Store price is $4.90 per 1,000 results, plus the Store-displayed infrequent Actor-start event where applicable. Platform usage is included for the user under the current Store pricing configuration.