Separate lead source fields that answer different capture questions
Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.
An enquiry arrives through a website form. Its URL contains campaign tags. The visitor says a colleague recommended the business. Which belongs in the source field: website, campaign or recommendation?
All three can be correct. Store submission channel, observed arrival context and self-reported discovery separately. Record missing information explicitly, distinguish rejected values from absent ones, and attach a collection-rule version so the record's meaning remains recoverable.
This design preserves what was known when an enquiry was created. It does not decide which interaction deserves credit for a sale. The practical outcome is an enquiry that staff can explain without guessing how its source fields were populated.
1. Define three questions, not one source label
A field named source invites incompatible answers. One employee selects the form location, another selects a campaign, and another selects a referral. Define each question before choosing field names.
- Submission channel: Through which route was this enquiry submitted or recorded?
- Observed arrival context: What technical arrival evidence was accepted at capture?
- Self-reported discovery: What does the person say first introduced them to the business?
If an employee records a phone enquiry in a CRM, the submission channel is phone. Manual entry describes the recording method, not the customer's entry route.
A tagged URL describes observed context, not motivation. It can coexist with a recommendation. Likewise, missing tags do not prove that no earlier interaction influenced the enquiry.
Discovery is the person's account. Do not replace it with technical evidence, even when the two appear inconsistent. They answer different questions.
2. Define submission and discovery combinations precisely
The following hypothetical dictionary is a reusable design, not a statement about existing product controls. The examples throughout are fictional. Adapt the channel and category lists to the business, but preserve their meanings.
submission_channel
Type: controlled choice
Values: website_form, inbound_email, phone,
business_messaging, in_person, unknown
Filled example: website_form
Rule: unknown means the entry route cannot be established.
submission_detail
Type: bounded text or controlled identifier
Filled example: project_enquiry_form
Rule: identify the actual intake point; empty if unavailable.
capture_rule_version
Type: identifier linked to a maintained rule register
Filled example: website_capture_v1
Rule: save the version actually used for this enquiry.
Use unknown if it cannot be established.
discovery_status
Values: answered, not_asked, unanswered, unknown
answered: the person supplied a response, including uncertainty.
not_asked: the question was not presented or asked.
unanswered: it was asked, but no response was supplied.
unknown: whether a response was obtained is unavailable.
discovery_category
Values when answered: personal_recommendation, online_search,
social_media, publication_or_newsletter, event, other, unknown
Otherwise: empty
Filled example: personal_recommendation
discovery_detail
Type: optional bounded text
Filled example: A colleague recommended your team.
Rule: no unnecessary personal details.Apply these combination rules consistently:
answered + unknownmeans the person explicitly selected uncertainty, such as “I don't know.” It does not mean a missing answer.answered + othermeans a supplied answer falls outside the listed categories. Optional detail can clarify it; uncertainty belongs underunknown.not_asked,unansweredand statusunknownall require an empty category and empty discovery detail.- If the capture result is unavailable, use status
unknown, notunanswered. If an answer is recoverable but cannot be mapped confidently, review it before assigning a category.
For example, a returned form with an untouched optional question is unanswered. A missing submission payload that prevents determining whether the person answered is unknown. Neither justifies populating the category from campaign tags.
3. Separate accepted evidence from capture quality
One status cannot adequately describe both retained evidence and rejected input. Use an evidence status alongside capture quality and bounded rejection reasons.
arrival_context_status
Values:
tags_present: at least one recognised tag value was accepted
referrer_only: accepted referrer, no accepted tag values
no_context_observed: successful capture, no accepted evidence
not_collected: route was configured not to collect context
not_applicable: web arrival context does not apply
unknown: evidence result cannot be established
arrival_capture_quality
Values:
clean: capture completed without rejecting recognised values
partially_rejected: some candidate values accepted, some rejected
rejected: candidate values rejected; no arrival values accepted
failed: capture failed; evidence result is unavailable
not_collected: collection intentionally disabled for this route
not_applicable: web arrival context does not apply
unknown: capture quality cannot be established
arrival_rejection_reasons
Type: list of controlled values
Values: length_exceeded, invalid_format, disallowed_value, unknown
Rule: nonempty for partially_rejected or rejected; empty otherwise.
Use unknown when rejection is known but its reason is not.
arrival_utm_source, arrival_utm_medium, arrival_utm_campaign
Type: bounded text
Rule: retain accepted values only; absent or rejected values empty.
arrival_referrer
Type: validated URL or documented reduced representation
Rule: retain only the detail needed; empty if unavailable.Here, referrer_only means only a referrer was accepted, not that no tags were supplied. Quality and rejection reasons disclose whether supplied tags were rejected. Similarly, no_context_observed + rejected means no usable evidence remained; it must not be displayed as “no tags supplied.”
Enforce these combinations: failed quality requires unknown evidence status and empty evidence fields. The not_collected and not_applicable values pair with their matching evidence statuses. partially_rejected requires at least one accepted arrival value; rejected requires none. A clean capture can still have no evidence.
Validate expected structure and length, and display retained values safely. Do not retain rejected raw input merely to explain the rejection. The W3C validation guidance also distinguishes accessible error handling from validation itself: explain actionable errors understandably, and do not rely on client-side checks alone.
4. Preserve collection provenance with a rule register
Evidence types do not identify collection methods. Tags read from the submitted page and tags retained from an earlier page can produce identical stored values while describing different moments.
Save capture_rule_version on each enquiry and link it to a maintained, immutable definition. A method change needs a new version, rather than an edit that changes the interpretation of older records.
This filled hypothetical register entry supplies the provenance for the worked examples below:
Rule version: website_capture_v1
Route: project enquiry website form
Submission: fixed website_form; detail project_enquiry_form
Arrival method: read recognised tags from the submission-page URL
Earlier-session retention: none
Recognised tags: utm_source, utm_medium, utm_campaign
Tag validation: hypothetical maximum 100 characters per value;
reject control characters; reject duplicate keys
Reason mapping: excess length = length_exceeded;
control characters = invalid_format;
duplicate keys = disallowed_value
Referrer method: use referring-page information when available;
retain origin only, without path or query
Discovery method: optional first-heard question on this form;
no preselected answer
Writers: capture logic writes arrival evidence;
the person's response supplies discovery
Failure rule: unavailable arrival result becomes unknown + failedThis starting method has a limitation: tags may disappear if a visitor moves to an untagged page before submitting. Retaining earlier context is a separate design choice. If introduced, document the retention and selection rules under a new version.
The version identifies the documented collection method, not an employee's identity or a correction history. Keep those separately where needed. An unknown version should trigger review, not an invented explanation.
5. Work through a complete enquiry and a rejected tag
Fictional worked example: Morgan hears about a supplier from a colleague, later follows a tagged newsletter link, and submits the project form. Morgan selects personal recommendation.
Enquiry: fictional Morgan project enquiry submission_channel: website_form submission_detail: project_enquiry_form capture_rule_version: website_capture_v1 arrival_context_status: tags_present arrival_capture_quality: clean arrival_rejection_reasons: [] arrival_utm_source: trade_digest arrival_utm_medium: email arrival_utm_campaign: autumn_planning arrival_referrer: empty discovery_status: answered discovery_category: personal_recommendation discovery_detail: A colleague recommended your team.
No value needs replacing. A faithful summary is: Submitted through the project form; newsletter tags accepted from the submission URL; the enquirer reports a colleague's recommendation. “Source: email” would discard two useful answers.
Now consider a second fictional record. A recognised source tag exceeds the hypothetical limit in the rule register, but a valid referring origin is available. The person explicitly selects uncertainty.
Enquiry: fictional rejected-tag enquiry submission_channel: website_form submission_detail: project_enquiry_form capture_rule_version: website_capture_v1 arrival_context_status: referrer_only arrival_capture_quality: partially_rejected arrival_rejection_reasons: [length_exceeded] arrival_utm_source: empty arrival_utm_medium: empty arrival_utm_campaign: empty arrival_referrer: https://newsletter.example discovery_status: answered discovery_category: unknown discovery_detail: empty
The correct summary is: Referrer retained; a supplied tag was rejected for length; the person does not know how they first heard about the business. If no referrer had been accepted, the arrival pair would instead be no_context_observed + rejected. Truly absent tags and referrer after successful collection would be no_context_observed + clean.
6. Ask people only what capture cannot observe
Do not ask website visitors to identify the submission route or decode tags. Ask the question only they can answer: How did you first hear about us? Offer a short category list, an uncertainty option and optional detail.
If the business wants to know what prompted today's enquiry, that is a different question. Give it a separate definition rather than mixing first discovery with the immediate trigger.
Keep discovery optional unless there is a concrete operational reason to require it. A forced choice may produce a convenient answer rather than useful information. Do not use a default selection to make records appear complete.
Give each control a clear, programmatically associated label, as described in W3C guidance on form labels. For staff-recorded enquiries, use the same category meanings. An employee who never asked should record not_asked, not infer a response from the conversation.
7. Protect meanings when records change
Attach capture values to the enquiry, not only to a mutable contact-level source field. A returning contact can submit through another route; the newer enquiry should not rewrite the earlier one's context.
- Known intake routes set submission channel. An unidentified import should not default to website.
- Capture logic writes arrival evidence. A discovery response must not replace it.
- People supply discovery. Staff may record an answer on their behalf, but must not substitute an impression.
- Corrections need an explanation. Where feasible, retain the prior value, correction reason, editor and time.
For phone enquiries, web context is normally not_applicable. A caller's statement that they found the business through search belongs in discovery, not observed arrival evidence.
Preserve old mixed source fields as historical context unless supporting evidence permits a reliable split. Check field types, form behaviour and write controls in the actual workspace and current plan before implementation. These are recommended operating rules, not claims that any Alactic product provides every control described.
8. Test interpretations as well as saved values
Build a fictional test set covering the two worked records, absent context, partial rejection, total rejection, failed capture, a phone enquiry, a skipped question and explicit uncertainty. Check stored fields and staff-facing summaries.
- Verify that each channel matches the actual intake route.
- Check every discovery status against its permitted category combination.
- Confirm that absent, rejected and unavailable evidence remain distinguishable.
- Verify that the saved rule version resolves to the method actually used.
- Submit another enquiry for the same fictional contact and inspect the earlier record.
- Make a correction and check whether its reason remains recoverable.
Ask someone who did not configure the fields to explain the records. If they interpret an empty referrer as proof of a direct visit, or a rejected tag as an absent tag, improve the summary and definitions.
Start with one intake route. During review, count invalid field combinations, unexplained capture failures and discovery values populated without an answer. These observable defects are more useful than measuring how many source fields look complete. Expand only when records preserve the three answers without silently converting uncertainty into certainty.
Put the next conversation to work.
Explore Alactic Communications for WhatsApp conversations, campaigns and AI-assisted workflows. Choose the product and plan that fit your team.
Explore Communications