Compare CRM integration and lead attribution services using a field map, matching rules, failure handling, and traceable test records. Agree on the meaning of each report before connecting its totals to budget decisions.
The checklist below is a suggested purchasing and acceptance process. Use it to compare providers on the same assignment, from website intake through the customer record and final report. No vendor capability, customer result, or improvement percentage is assumed.
Write the integration brief around a record journey.
List each intake route: website forms, campaign landing pages, calls, referrals, imports, and manual entry. For each route, identify the sending system, receiving system, owner, authentication method, fields, and expected record type.
Ask the provider to demonstrate a test inquiry from its source through the resulting CRM record. Require an explanation of what happens when that person returns, submits another form, or already exists in the database.
Define where the integration ends. Specify whether the scope includes source capture, identity matching, activity history, reporting, or reconciliation with sales outcomes. Assign ownership of any step that remains manual.
Require proposed production writes to be listed explicitly. Review the test plan and rollback procedure before granting access that can change customer data. Keep the acceptance exercise separate from live campaigns.
Separate source evidence from attribution policy.
Store the source values received at intake separately from the label used in reporting. For this suggested design, keep an original observed source, a latest observed source, a submitted campaign identifier, a landing page, and an intake timestamp. Include how each value was obtained.
Google documents campaign URL parameters such as utm_source, utm_medium, and utm_campaign. It also notes that parameter values are case sensitive. Use a written naming convention and preserve the submitted values when applying a reporting label.
Ask the provider to define its attribution policy in plain language. For example, your chosen first-observed policy could retain the earliest source captured by this integration. A separate latest-observed field could record a later inquiry. These are proposed field rules, not proof of a person's complete history.
Keep a self-reported answer to “How did you hear about us?” in its own field. Do not silently replace observed source data with that answer. Show both when investigating a disagreement.
Approve the field map and matching rules.
Request a mapping table before implementation. Give every field a definition, destination, allowed values, and rule for missing data. Identify fields that the integration may update and fields it must preserve.
| Field group | Proposed contents | Acceptance check |
|---|---|---|
| Event identity | External submission ID and source system | Replaying the same event does not create another inquiry. |
| Contact identity | Approved matching keys and match outcome | A matching contact receives the activity according to the agreed rule. |
| Source evidence | Submitted campaign values, landing page, timestamp | The test record preserves the values supplied at intake. |
| Reporting label | Normalized source, medium, and campaign | The label follows the documented mapping. |
| Delivery status | Attempt time, outcome, and error reference | A failed attempt is visible and can be reconciled. |
Require uncertain identity matches to follow an explicit review path. Test shared email addresses, changed contact details, and incomplete submissions with synthetic records. Agree which records require human review before merging or overwriting anything.
Ask for a dated mapping version on each deployment. Keep the mapping and its change history with the project handover. Reject undocumented field renaming during implementation.
Test repeated submissions and failures.
Request a foreground demonstration using synthetic contacts and a small set of events. Observe both the sending system and CRM. Do not accept a success message from the integration alone as proof that the receiving record is correct.
Test these scenarios against the agreed requirements:
- A new inquiry supplies the expected source values.
- An existing contact submits a later inquiry with a different campaign.
- The same event is delivered again after a retry.
- Source values are absent, malformed, or outside the naming convention.
- A credential expires or the receiving service rejects the request.
- A partial failure occurs after the contact exists but before its activity is recorded.
For every scenario, record the input, expected outcome, actual CRM record, and delivery status. Require the provider to explain how an administrator finds and resolves failures without repeatedly creating contacts.
Use an unknown-source value when evidence is missing. Keep corrections traceable. Do not assign a campaign merely because its launch happened near the inquiry date.
Define reporting dimensions before comparing totals.
Google Analytics distinguishes first-user, session, and event channel dimensions. Its channel groups follow defined classification rules. Name the dimension and scope when comparing an analytics report with a CRM report.
Choose a reporting period, timezone, counting unit, and duplicate policy. State whether a row represents a person, inquiry, opportunity, or customer. Specify which date determines inclusion, such as inquiry creation or a recorded sales outcome.
Ask the provider to reconcile a small sample manually. Trace each row to its underlying event and record. Keep excluded and unresolved records visible in the reconciliation rather than removing them to make totals agree.
If you need a cost worksheet, define its inputs explicitly: selected-period spend divided by the counted inquiries or customers in the agreed group. State what spending is included. Label that result as a calculation under your chosen rules, rather than evidence that a channel caused a sale.
Compare service proposals and handover terms.
Compare discovery, implementation, testing, monitoring, support, and documentation separately. Ask which systems and permissions are required. Require a written owner for credentials, failure alerts, field definitions, and future changes.
Request examples of deliverables with private information removed. Review the field map and acceptance record rather than accepting an unsupported claim about a provider's accuracy. Ask a permissioned reference how errors were handled and whether the handover was usable.
Before launch, require the test results, mapping version, administrator instructions, and a recovery procedure. Keep configuration and documentation in accounts your business controls. Set a review point after the initial release to inspect unresolved events and source labels.
For the website measurement side, see the SEO measurement guide. To discuss the boundaries of an integration project, contact Victor.
Questions to settle before implementation.
Can a CRM integration establish every lead's true origin?
Define the claim more narrowly. Ask which intake routes are covered and which source evidence each one supplies. Require the report to show missing information rather than promising complete knowledge of a person's journey.
Should a returning inquiry overwrite its original source?
Choose that policy before implementation. In the proposed design above, retain the original observed source and add the latest inquiry as a separate observation. Document any correction with its reason and date.
Do we need a multi-touch attribution platform?
List the reporting decisions you need to make. Ask a candidate to demonstrate those decisions using your required fields and synthetic records. Compare an additional platform only when the agreed reporting requirements call for it.
What should the provider deliver before payment milestones are accepted?
Put the deliverables in the agreement: field mapping, matching rules, tested scenarios, reconciled records, failure visibility, access ownership, and handover instructions. Use the agreed acceptance criteria to review each milestone.