Lead-Generation Sources: Rank Evidence Before Volume

Define a source by its evidence

A lead-generation source is the origin of an observation that may support a commercial next step. The source can preserve explicit requests, referrals, event context, content interactions, partner introductions, account research, or social participation. These events differ in identity confidence, fit, intent, timing, and permission; one lead count erases the differences that matter. A trade directory entry, an event attendee list, a referral, a website visit, and a direct request do not carry the same meaning. The directory may support company existence and category. The attendee list may show participation in a topic. A referral may supply relationship context. A visit may show an interaction without identity or authority. A direct request may reveal a question while leaving account fit unresolved. Source comparison begins by preserving those meanings instead of translating every event into the same lead status. Start with five distinct events: a directory listing, event registration, partner referral, anonymous website visit, and direct specification request. Record what each event directly observes, its date, identity level, permission context, and the action it can support. The listing may establish company category; the request may establish a question. Neither inherits meaning from the other. A source with no reconstructable origin is held before qualification. Keep ITA primary versus secondary as a dated distinction. Drop leftover rank language that is not a scored ranking.

Build a source taxonomy

Create a taxonomy around evidence: first-party requests, customer referrals, partner routes, events, inbound content, search, paid acquisition, marketplaces, social context, and outbound research. Define what each can and cannot imply. A webinar registration may show topic interest; a direct demo request may show desire for a conversation; neither automatically proves budget or authority. The ranking criteria need a direct taxonomy. The U.S. International Trade Administration distinguishes primary research gathered directly from secondary research obtained through existing statistics or reports, then points exporters to country- and industry-specific resources. Extend that distinction for lead operations without claiming it predicts revenue: an owned form preserves a submitted action, a marketplace request preserves an expressed requirement, an observed social event preserves public context, a provider file preserves whatever provenance the contract discloses, and a referral preserves a human introduction. Link each source event to later qualification and opportunity disposition by a stable source ID; otherwise a ranking by volume cannot show which evidence survived. Can you name what your source observed? Can you show when your source observed it? Can your reviewer trace your source to a later opportunity state? Which permission did your source preserve for you? What would make you retire your source instead of buying more volume? Assign stable event codes to first-party requests, referrals, partner routes, events, content interactions, paid acquisition, marketplaces, social context, and outbound research. Keep the original code when qualification adds an interpretation. A webinar registration remains attendance or topic interest until new evidence changes it; a demo request remains an expressed request but may still fail account fit. Taxonomy prevents unlike events from becoming one lead total. OKKI Go should retain its own source code and observation limits when used for B2B company research, rather than being merged into an undifferentiated lead total.

• Lead-generation sources should be compared by the evidence they preserve about intent and eligibility, not by how many contacts they contribute to the database.
• The International Trade Administration distinguishes primary market research conducted directly from secondary research obtained from existing statistics or reports and lists country- and industry-specific sources.
• A trade directory entry, an event attendee list, a referral, a website visit, and a direct request do not carry the same meaning.
• Record what each event directly observes, its date, identity level, permission context, and the action it can support.
• A source with no reconstructable origin is held before qualification.

Score context, confidence, and permissions

Score sources on identity, account eligibility, role relevance, event specificity, recency, provenance, permission or objection status, cost to verify, and correction handling. Make missing data visible. Do not let a source award itself confidence through a proprietary score the buyer cannot reconstruct. A cheaper record can be expensive if research and misrouting costs are hidden. Use a review rubric that separates identity, account eligibility, role relevance, event specificity, recency, provenance, permission or objection status, verification effort, and correction handling. Missing values stay visible. A proprietary score cannot award itself confidence. For one hundred illustrative records, report accepted, rejected, held, corrected, and suppressed counts separately, with no conversion benchmark implied.

Connect source events to qualification

Preserve the original event when qualification adds an interpretation. The seller should be able to distinguish “downloaded,” “referred,” “researched,” and “requested contact” after the record enters a CRM. Define action bands, owners, and reversal rules. A later objection or identity correction should update eligibility without deleting the historical observation. Follow one account found in a directory, represented at an event, and later tied to a direct specification question. The evidence strengthens because each source answers a different question, not because there are three copies. Verify the responsible role, request scope, permitted reply, and next review date. If the product cannot meet the specification, disqualify the account while retaining the source trail and the reason.

Allocate budget through marginal learning

Compare sources through accepted leads, qualified conversations, progression, negative signals, verification time, and incremental learning. Use source-specific denominators and cohort windows. Budget should move when a source repeatedly provides decision-useful evidence at an acceptable total cost, not merely because its top-of-funnel number looks efficient. Compare sources using accepted records, qualified conversations, progression, negative signals, verification time, correction burden, and source-specific denominators over a stated cohort window. Move budget only when repeated cohorts show decision-useful evidence at an acceptable total cost. A cheap file that requires extensive identity and role repair may lose to a smaller referral source even when its top-line cost per record looks lower.

Audit source overlap and duplication

Map when several vendors, ads, or platforms describe the same person or company event. Deduplicate identity while retaining source history. Multiple copies of one announcement do not create corroboration, and several paid touches preceding one request do not automatically deserve equal causal credit. Resolve several copies of the same company or announcement into one identity while retaining every source event and timestamp. Repeated syndication is not independent corroboration. Where sources conflict on role, status, or location, create a correction task instead of selecting the most convenient value. The audit records which observation survived, who decided, what downstream states changed, and how future imports avoid recreating the duplicate.

Write a source retirement rule

Set thresholds for unverifiable identity, stale data, objections, duplicate burden, weak qualification, or hidden operating cost. A familiar source should not survive solely because the team has always reported it. Preserve the evaluation period and decision so future teams can understand why spending changed. Consider one account found in three places. Its directory profile fits the market, an employee attended a relevant event, and a contact later asks for a specification. The evidence becomes stronger because the pieces answer different questions, not because three sources automatically equal intent. The team still verifies the responsible role, the request's scope, the permitted reply, and the next review date. If the specification cannot be served, the account is disqualified despite the rich trail. If the contact asks only for educational material, the response should match that request rather than force an opportunity stage. Ask the source a narrower question before you score the record. What did the directory actually observe? What did the event registration require? Who made the referral, and what context did they provide? Does the website visit identify a person or only a session? What did the direct request ask for? You should be able to answer without borrowing meaning from another source. Then combine evidence deliberately. Does the second source confirm identity, add role context, or reveal timing? Could it simply repeat the same stale field? Which conflict deserves investigation? When the account advances, record the decisive observation. When it stops, retain the reason. That discipline lets you compare sources by the decisions they support rather than by the number of names they deliver. You can test the record quickly. Can you name the event? Can you show its date? Do you know whether the source observed a company, a person, or an action? What are you still inferring? Which second source could correct you? Would you advance if the sources conflict? Who owns the correction? What will make you stop? OKKI Go can preserve the final source rationale beside company research and candidate review; the team still decides whether the evidence supports advance, return, retirement, or no contact. Retire or pause a source when identity cannot be verified, freshness breaches the agreed window, duplicate or objection burden rises, accepted progression remains weak, or the provider cannot explain correction and permitted use. Preserve the evaluation period, denominators, disputed cases, decision owner, and exit duties. The next team should see why spending changed and what new evidence would justify a controlled retest rather than reviving the source by habit.