What lead-source attribution records can measure

Research question and scope

What does a lead-source field measure when a person can encounter several channels before making contact? The unit of analysis is one received inquiry with a declared attribution rule, timestamp, source value, and correction history. The method basis is a first-touch or last-touch rule chosen before analysis, with unknown and multi-source categories retained instead of silently reassigned.

Supported findings

Attribution records can describe how inquiries were labeled under a stated rule. They cannot demonstrate that a channel caused a purchase, that an untracked channel had no influence, or that a small source difference is commercially meaningful.

Evidence note 1

A referral may also click a search result before submitting a form.

Evidence note 2

A campaign code can identify a link without identifying the reason a person trusted it.

Evidence note 3

A missing source is a data state that should remain visible.

Evidence note 4

Changing the form question breaks comparisons across months.

Evidence note 5

A lead count and a qualified-lead count answer different questions.

Evidence note 6

A referral may also click a search result before submitting a form. For this question, note 1 should be recorded before any summary is written. The reviewer should preserve the original observation, identify who or what was eligible to appear in it, and state which part of the claim the note can support. A later reader should not have to infer whether the item was measured directly, reported by a participant, copied from a schedule, or supplied by an external publication.

Evidence note 7

A campaign code can identify a link without identifying the reason a person trusted it. For this question, note 2 should be recorded before any summary is written. The reviewer should preserve the original observation, identify who or what was eligible to appear in it, and state which part of the claim the note can support. A later reader should not have to infer whether the item was measured directly, reported by a participant, copied from a schedule, or supplied by an external publication.

Evidence note 8

A missing source is a data state that should remain visible. For this question, note 3 should be recorded before any summary is written. The reviewer should preserve the original observation, identify who or what was eligible to appear in it, and state which part of the claim the note can support. A later reader should not have to infer whether the item was measured directly, reported by a participant, copied from a schedule, or supplied by an external publication.

Evidence note 9

Changing the form question breaks comparisons across months. For this question, note 4 should be recorded before any summary is written. The reviewer should preserve the original observation, identify who or what was eligible to appear in it, and state which part of the claim the note can support. A later reader should not have to infer whether the item was measured directly, reported by a participant, copied from a schedule, or supplied by an external publication.

Evidence note 10

A lead count and a qualified-lead count answer different questions. For this question, note 5 should be recorded before any summary is written. The reviewer should preserve the original observation, identify who or what was eligible to appear in it, and state which part of the claim the note can support. A later reader should not have to infer whether the item was measured directly, reported by a participant, copied from a schedule, or supplied by an external publication.

Evidence note 11

A second-pass check for research-question-fitness-lead-source-attribution asks a different question about item 1: A referral may also click a search result before submitting a form. If this condition changes the denominator, timing, or meaning of the record, annotate the break rather than smoothing it away. If it does not change the stated measure, explain why that judgment is defensible. This makes the evidence auditable without implying that the local record represents everyone who was absent from it.

Evidence note 12

A second-pass check for research-question-fitness-lead-source-attribution asks a different question about item 2: A campaign code can identify a link without identifying the reason a person trusted it. If this condition changes the denominator, timing, or meaning of the record, annotate the break rather than smoothing it away. If it does not change the stated measure, explain why that judgment is defensible. This makes the evidence auditable without implying that the local record represents everyone who was absent from it.

Evidence note 13

A second-pass check for research-question-fitness-lead-source-attribution asks a different question about item 3: A missing source is a data state that should remain visible. If this condition changes the denominator, timing, or meaning of the record, annotate the break rather than smoothing it away. If it does not change the stated measure, explain why that judgment is defensible. This makes the evidence auditable without implying that the local record represents everyone who was absent from it.

Evidence note 14

A second-pass check for research-question-fitness-lead-source-attribution asks a different question about item 4: Changing the form question breaks comparisons across months. If this condition changes the denominator, timing, or meaning of the record, annotate the break rather than smoothing it away. If it does not change the stated measure, explain why that judgment is defensible. This makes the evidence auditable without implying that the local record represents everyone who was absent from it.

Evidence note 15

A second-pass check for research-question-fitness-lead-source-attribution asks a different question about item 5: A lead count and a qualified-lead count answer different questions. If this condition changes the denominator, timing, or meaning of the record, annotate the break rather than smoothing it away. If it does not change the stated measure, explain why that judgment is defensible. This makes the evidence auditable without implying that the local record represents everyone who was absent from it.

Evidence note 16

Interpretive memo 1: A referral may also click a search result before submitting a form. Compare this observation only with records collected under the same rule. A changed instrument, revised question, different class or service boundary, or new follow-up interval can create a discontinuity. Mark that discontinuity in the table and describe the likely effect on interpretation. The honest result may be a smaller claim, an explicit unknown, or a request for a qualified reviewer.

Evidence note 17

Interpretive memo 2: A campaign code can identify a link without identifying the reason a person trusted it. Compare this observation only with records collected under the same rule. A changed instrument, revised question, different class or service boundary, or new follow-up interval can create a discontinuity. Mark that discontinuity in the table and describe the likely effect on interpretation. The honest result may be a smaller claim, an explicit unknown, or a request for a qualified reviewer.

Evidence note 18

Interpretive memo 3: A missing source is a data state that should remain visible. Compare this observation only with records collected under the same rule. A changed instrument, revised question, different class or service boundary, or new follow-up interval can create a discontinuity. Mark that discontinuity in the table and describe the likely effect on interpretation. The honest result may be a smaller claim, an explicit unknown, or a request for a qualified reviewer.

Evidence note 19

Interpretive memo 4: Changing the form question breaks comparisons across months. Compare this observation only with records collected under the same rule. A changed instrument, revised question, different class or service boundary, or new follow-up interval can create a discontinuity. Mark that discontinuity in the table and describe the likely effect on interpretation. The honest result may be a smaller claim, an explicit unknown, or a request for a qualified reviewer.

Evidence note 20

Interpretive memo 5: A lead count and a qualified-lead count answer different questions. Compare this observation only with records collected under the same rule. A changed instrument, revised question, different class or service boundary, or new follow-up interval can create a discontinuity. Mark that discontinuity in the table and describe the likely effect on interpretation. The honest result may be a smaller claim, an explicit unknown, or a request for a qualified reviewer.

Bounded conclusion

What does a lead-source field measure when a person can encounter several channels before making contact? The defensible endpoint is a documented observation with its source context and uncertainty attached; any stronger claim would exceed the stated unit or method.

Sources and verification

1. https://www.ftc.gov/business-guidance/resources/advertising-marketing-internet-rules
2. https://www.nist.gov/privacy-framework
3. https://www.bls.gov/ooh/personal-care-and-service/fitness-trainers-and-instructors.htm