New Lead Generation Needs a Learning Budget

Define what is actually new

New lead generation may mean a new audience, source, signal, proposition, channel, region, or workflow. Those are different experiments. Name one novelty and keep the rest as close as practical to a known baseline; otherwise the team cannot tell which assumption produced the outcome. Run one bounded source test with a denominator, hour budget, and a prewritten stop rule. Novelty without those three objects is scaled ambiguity. What's the one new assumption you're actually testing? You'll waste the week if I can't name the novelty.

I run one novelty at a time. The bounded test I will keep: new source is a 17 August 2026 county food-license CSV; baseline is last quarter’s chamber list; denominator is 80 eligible manufacturers; budget is six hours; stop if two hours produce no reconstructable accept. SBA market-research and ICO direct-marketing pages date the research and the send duty; they are not the win. Unfamiliar names are not a win. What’s new, and what stays baseline?

• New audience: hold source and proposition stable.
• New source: test against the same segment rule.
• New signal: keep the channel constant.
• New channel: reuse an approved audience definition.
• New region: refresh market and data context.
• New workflow: preserve the acquisition hypothesis.
• Keep an assumption ledger with expected observation, actual observation, interpretation, and next action. Update it during the experiment so the final review cannot rewrite weak early evidence as an inevitable learning journey.
• Compare the new source with a deliberately sampled baseline set reviewed by the same people. Reviewer rotation, training changes, or a different mix of easy records can otherwise masquerade as source quality.

Write the baseline beside the experiment

The baseline should state audience, source, acceptance rule, action, cost, and downstream disposition for a recent comparable period. Market research from SBA and target-market guidance from Australia help establish demand, audience traits, limits, buying behavior, and discovery channels before the test. Can you show the stop rule before you spend the next hour? Can't we write the baseline first?

• Choose a comparable period.
• Lock eligibility definitions.
• Show raw counts and denominators.
• Record owner capacity.
• Keep known seasonal differences visible.
• State what cannot be compared.
• Keep an assumption ledger with expected observation, actual observation, interpretation, and next action. Update it during the experiment so the final review cannot rewrite weak early evidence as an inevitable learning journey.
• Compare the new source with a deliberately sampled baseline set reviewed by the same people. Reviewer rotation, training changes, or a different mix of easy records can otherwise masquerade as source quality.

Validate a new source with a small sample

Take a representative set containing obvious fits, borderline cases, and records expected to fail. Preserve provenance, apply the existing acceptance rule, and compare false positives, missing fields, corrections, and review time with the baseline source. Do not unlock or contact merely because a record was returned. If the baseline is missing, you're not running an experiment. I want a stop rule.

• Inspect source scope and update frequency.
• Review company identity and segment evidence.
• Record every rejection reason.
• Compare reviewer agreement.
• Use OKKI Go to refine search conditions before selective unlock.
• Stop if provenance or permission context is inadequate.
• Keep an assumption ledger with expected observation, actual observation, interpretation, and next action. Update it during the experiment so the final review cannot rewrite weak early evidence as an inevitable learning journey.
• Compare the new source with a deliberately sampled baseline set reviewed by the same people. Reviewer rotation, training changes, or a different mix of easy records can otherwise masquerade as source quality.

Spend a learning budget deliberately

Cap records, money, operator hours, and calendar time. Predefine the observation that warrants extension, revision, or closure. A small experiment is not permission to ignore objections, opt-outs, fairness, accuracy, or other applicable duties described in ICO guidance. Don't scale a hunch.

• Budget the review effort, not only media spend.
• Reserve time for correction and analysis.
• Assign a stop authority.
• Route objections immediately.
• Reject scope expansion without a new decision.
• Archive the failed hypothesis.
• Keep an assumption ledger with expected observation, actual observation, interpretation, and next action. Update it during the experiment so the final review cannot rewrite weak early evidence as an inevitable learning journey.
• Compare the new source with a deliberately sampled baseline set reviewed by the same people. Reviewer rotation, training changes, or a different mix of easy records can otherwise masquerade as source quality.

Compare learning with the baseline

At review, examine accepted candidates, correct referrals, qualified conversations, disqualifications, suppressions, source corrections, cost, and elapsed owner time. Use OKKI Go sending status only to distinguish delivery observations; the team must classify whether a reply supports the hypothesis. You'll waste the week if I can't name the novelty.

• Keep baseline and test definitions aligned.
• Show counts beside every rate.
• Inspect differences in segment mix.
• Separate new-source errors from message errors.
• Document what the novelty taught.
• Decide whether the operating contract must change.
• Keep an assumption ledger with expected observation, actual observation, interpretation, and next action. Update it during the experiment so the final review cannot rewrite weak early evidence as an inevitable learning journey.
• Compare the new source with a deliberately sampled baseline set reviewed by the same people. Reviewer rotation, training changes, or a different mix of easy records can otherwise masquerade as source quality.

I keep one novelty and freeze the rest against a written baseline.

What's new in your test, and what stays comparable? Can't we write the baseline first?

If you can't show denominator, budget, and stop rule, you're scaling ambiguity.

Don't expand a source just because the names look unfamiliar. I want a stop rule.

A bounded source test compares false positives, missing fields, and review time with the baseline source on a representative mix of fits and expected fails.

Closing a weak novelty saves later research and follow-up; that is a successful experiment. Don't scale a hunch.