Quick answer: Keep four questions separate: how the person first found you, how they most recently returned, how the CRM record was created and which campaign link was used. Store unknown when evidence is missing. One field called “Lead source” is usually asked to do four jobs, then gets blamed when the report becomes interpretive dance.
Jump to: Four source questions · Minimum fields · Worked example · Reporting rules · Copyable record
For: founders, owners and lean service teams connecting campaign links, forms and sales follow-up. This is a vendor-neutral operating framework. Product behaviour and plan availability should be checked in the selected CRM. Examples are fictional and are not attribution benchmarks.
Decision gate: If the team cannot explain which event updates a field, do not use it to rank channels.
- Four source questions that should not share one answer
- The minimum source fields for a lean team
- Worked example: one enquiry, four valid source facts
- Pass campaign detail into the CRM without inventing precision
- Protect history during imports and manual cleanup
- Report lead quality with a cohort and an unknown row
- Run five acceptance tests
- Copyable lead-source record
- Useful detail: Keep drill-down evidence before a manual source edit
- Sources and editorial basis
Four source questions that should not share one answer
Original or first source answers how a person first became known through tracked activity. Latest source answers how that person most recently returned through known activity. Record source answers how the CRM record itself was created, such as a form, import, integration or user entry. Campaign detail records the tagged initiative or placement associated with a link.
These facts can all be different without any data being wrong. HubSpot’s current documentation, for example, separates Original Traffic Source and Latest Traffic Source from Record Source. It describes Record Source as the tool or process that created a record, not the marketing interaction that introduced the person. Exact labels vary by platform, but the distinction is useful even in a spreadsheet.
On a small screen, swipe the table sideways to see all columns.
| Question | Example value | Suggested update rule |
|---|---|---|
| How did they first find us? | Email marketing | Preserve the earliest supported value; do not overwrite it with a later visit. |
| How did they most recently return? | Direct traffic | Update when the measurement system identifies a newer qualifying interaction. |
| How was the record created? | Website form | System-set creation fact; keep it separate from acquisition source. |
| Which tagged link was used? | service_review_2026q4 | Store the captured raw campaign value and its date when available. |
| What can we not prove? | Unknown | Keep unknown visible; do not distribute it across known sources. |
Do not ask a salesperson to overwrite an automatically collected first source because a customer mentioned another channel on a call. That comment may be useful self-reported context, but it is a different field with a different evidence standard.
The minimum source fields for a lean team
Start with the fields that answer recurring decisions. A practical first version is:
- Original source group: a controlled category such as organic search, paid search, email, referral, social, direct, offline or unknown.
- Original source detail: the supporting campaign, site, referrer or note when captured.
- Original source date: when the earliest supported interaction occurred.
- Latest source group and date: the most recent supported interaction, kept separate from the first.
- Record creation source: form, import, API, integration or manual entry.
- Campaign fields: raw source, medium, campaign and content values when they exist.
- Self-reported source: the customer’s own answer, labelled as self-reported rather than silently replacing measured history.
- Evidence status: system observed, customer reported, manually verified or unknown.
Add each item to the CRM data dictionary with its object, owner, allowed values and blank rule. Avoid making campaign fields mandatory for every enquiry. A phone call from a referral may be valid business even when no campaign parameter exists.
Worked example: one enquiry, four valid source facts
Imagine a fictional three-person web studio. Priya first visits a service page from a tagged customer newsletter. Two weeks later she returns by typing the site address, submits a form and becomes a CRM contact.
- Original source: email marketing.
- Original campaign detail: service_review_2026q4.
- Latest source at conversion: direct traffic.
- Record source: website form.
If the team stores only the form creation event, email appears to have contributed nothing. If it overwrites the first source with the later direct visit, the newsletter disappears. If it calls the newsletter the record source, it confuses acquisition with the mechanism that created the CRM row. Keeping the four facts separate preserves a usable story without pretending one label explains the whole journey.
Now change the facts. Priya phones after a colleague recommends the studio, and an employee creates the record manually. Record source is manual entry. Self-reported source may be referral. Original tracked web source may remain unknown. That is not a failure to complete the form; it is an honest description of the evidence.
Pass campaign detail into the CRM without inventing precision
Google Analytics documents campaign URL parameters including source, medium and campaign for identifying referral links and campaigns. Use the UTM naming guide to keep those values consistent. Test the final link and the capture path before distributing it.
A campaign parameter identifies the tagged link that was used. It does not prove that the campaign alone caused the enquiry. Forwarded emails, copied links, blocked measurement and later visits can all change what is visible. Store raw values when available, but group them for reporting only through a documented mapping.
When the CRM receives a form submission, write the raw campaign values into dedicated fields or a linked interaction record according to the product’s supported method. Do not concatenate every value into one free-text “source” field. Keep the destination, capture date and consent or measurement limitations in the test record. This article does not change the site’s consent or analytics configuration.
Protect history during imports and manual cleanup
Before an import, map each source column to the intended target field. Treat a blank as unknown unless the source system defines it differently. Do not turn every imported contact into “offline” merely because the import created the new CRM record; record creation and prior acquisition are different facts.
For manual corrections, require a reason, date and operator. Preserve the previous value where the platform supports history, or export a dated snapshot before a material bulk change. A correction should improve traceability, not make the chart line up with a preferred explanation.
If multiple records represent the same person, resolve the duplicate before comparing source performance. Use the duplicate cleanup checklist and decide which history survives the merge. A merge that keeps the newest email but drops the earlier acquisition event can change reported first source.
Report lead quality with a cohort and an unknown row
Choose a cohort of valid enquiries received during the same period. For each original source group, show the number received, qualified, won, lost and still open at a stated review date. Keep unknown as a visible row. Then label every rate with its denominator.
Fictional example: 30 valid enquiries arrive in a month. Twelve are email, eight organic search, four referrals and six unknown. At the review date, three email, two organic and one referral enquiry are won; the rest include both lost and still-open work. Report the counts and unresolved share before comparing win rates. Do not allocate the six unknown enquiries in proportion to the known rows.
Use the pipeline metrics guide to separate workload from outcomes. Small counts can swing sharply, so avoid ranking channels or people from a handful of cases. A source report should help decide what to investigate next, not produce confidence that the sample has not earned.
Run five acceptance tests
- First then latest: a later visit updates the latest field without changing the supported first field.
- Form creation: record source identifies the form while campaign detail remains available separately.
- Manual record: manual creation does not fabricate an acquisition source.
- Unknown path: a user can save and route an enquiry without guessing a channel.
- Report trace: a counted enquiry can be traced back to its record, cohort rule and source evidence.
Record pass, fail and not tested separately. Repeat the tests after changing a form, integration, import mapping or source taxonomy. If a test needs administrator access that daily users do not have, it has not proved the normal workflow.
Copyable lead-source record
Reporting question:
Contact or enquiry reference:
Original source group, detail and date:
Latest source group, detail and date:
Record creation source:
Raw source / medium / campaign / content:
Self-reported source, if collected:
Evidence status and known limitations:
Correction reason, owner and date:
Cohort period and review cutoff:
Qualified / won / lost / still-open status:
Unknowns and next investigation:
Decision rule: preserve a source field when it answers a distinct question with a repeatable update rule. Combine fields only when the underlying events are genuinely the same. Add detail when it changes a decision; otherwise keep the first version small enough that the team will maintain it.
Useful detail: Keep drill-down evidence before a manual source edit
HubSpot clears traffic-source drill-down properties when a record’s traffic-source value is changed manually.
A tidy category can remove useful context if someone edits it without recording why. Prefer a separate correction note for disputed attribution until the team has checked the original evidence and approved the intended change.
Try this:
- Inspect the source category and both drill-down fields on a harmless sample.
- Record the evidence and reason before any permitted correction.
- After the edit, check which details remain; do not treat blank drill-downs as proof that no campaign was involved.
Source: HubSpot: traffic source properties and drill-downs. Checked October 3, 2026. The practical check is a PSL suggestion.
Sources and editorial basis
Primary references checked September 30, 2026: HubSpot’s Original and Latest traffic source properties, Record Source properties and Google’s campaign URL documentation. Product terminology, automatic updates, permissions and plan availability can change; verify them in the intended account.
The field set, update rules, examples, tests and reporting framework are Practical SaaS Lab editorial proposals, not measured customer results or a complete attribution model.
