Website-to-CRM Integration: Enquiries, UTM Tags and Acceptance Tests
A website-to-CRM integration is ready to launch when every accepted enquiry can be found and traced to the record a manager receives. Test delivery retries, CRM failures and attribution too. A “Thank you” message in the browser should confirm only the stage that the server has actually completed.
The business outcome is straightforward: the enquiry reaches the correct pipeline with its form data and source, gets an assigned owner and survives a temporary failure. For the developer, this means agreeing several separate rules before connecting the API.
The workflow below is intended for service and business websites. Use it to prepare requirements and acceptance tests. Check the exact fields, permissions and limits against your CRM documentation and account settings.
To review the integration with your contractor, download the acceptance template: RU — CSV · EN — CSV. It contains 22 scenarios covering fields and attribution, repeated enquiries, routing, failures, mobile forms and analytics checks. Each has space for the actual result, redacted evidence, an owner and the check date.
Decide what the CRM should create
Start with the sales team's workflow. For one business, a new enquiry becomes a lead; for another, it becomes a deal and a linked contact. The word “enquiry” on a website does not automatically determine the CRM object type.
Record the enquiry's path: which form is submitted, which object is created, which pipeline it enters, who receives a notification and what the manager does next. If “Request a quote” and “Become a partner” forms are handled by different teams, they need different routing rules.
Assign an owner for reference data. Services, sources, stages and employees have identifiers. Matching by a text label without checking it can break after a rename. Changes in the CRM require an integration update procedure and a control test.
System design and exchanges with other services are covered in the CRM architecture guide. This article addresses a specific task: reliably delivering a website enquiry and checking how it is processed.
Map the fields
For each field, define its source, transformation and destination. Review the resulting record with a manager: a technically successful request can create an object that is difficult to work with.
| Data | Source | What to agree |
|---|---|---|
| Name and contact details | Form fields | Required fields, format and behaviour when missing |
| Service or product | Form and current page | CRM reference data or a separate field |
| Comment | Visitor input | Length limit and safe display |
| Enquiry page | URL of the page containing the form | Which parameters are retained and which are excluded |
| Form identifier | Website configuration | Distinguish header, service-page and partnership forms |
| UTM parameters | Campaign landing and agreed storage | First or last touch and retention period |
| CRM source | Integration rule | Match the available source reference data |
| Accepted enquiry identifier | Website server | Lookup, retry and link to a CRM object |
| Responsible manager | Routing rule | Fallback when an employee is unavailable and assignment checks |
Do not send the entire page URL to the CRM without reviewing it. Its parameters may contain data that sales does not need. Agree on an allowed set, such as campaign tags and a page identifier. Also define what service logs retain, who can access them and for how long.
Check CRM requirements before development: mandatory fields, text length, connection permissions, supported objects and automatic actions. For example, Bitrix24 documentation for crm.item.add describes permission and required-field checks; automation rules can run after saving. Route test enquiries to an agreed test workflow.
Preserve UTM tags and enquiry attribution
UTM parameters help associate an enquiry with an advertising campaign. Their values need to travel from the landing page to the form and then to the intended CRM fields. If a visitor browses another page before submitting, reading only the current URL may be insufficient.
Before implementation, choose a model: retain the first known campaign touch, the last, or both in separate fields. Define what happens on a later direct visit and when saved data expires. These are your analytics rules; no single option suits every business.
Do not attribute an enquiry to a previous campaign if your rule no longer considers that campaign current. Do not label an unknown source “organic” merely because UTM tags are missing. The visitor may have followed an untagged link or arrived with limited referral information.
Metrica classifies sources using tags and other information about a visit. The CRM receives the data your integration provides. Their reports can therefore differ, particularly when they use different attribution rules. To reconcile them, use the same test visit with known parameters and a clearly defined attribution rule.
A practical test: open a tagged URL, navigate to a service page, submit the form and check every agreed value in the CRM. Repeat without tags, with another campaign and on a return visit. Each result should follow the chosen rule.
Separate enquiry acceptance from CRM delivery
One option for a custom form is to save the enquiry reliably on your server before delivering it to the CRM. This requires storage, a retry handler and delivery monitoring. It separates the wait for an external system from the response shown to the visitor.
If you use an embedded CRM form or a connector, establish which functions the supplier provides and which you can verify. A successful connection does not establish that a queue, logs or retries are available.
The getting-started guide for amoCRM describes embedded forms, processing in Unsorted and manager access to that section; it also documents API integration for a custom form. During acceptance, check where the enquiry appears and whether the correct employee can access it. Behaviour in a particular account depends on its configuration.
For a custom handler, agree on delivery states:
| Delivery status | What has happened | What to check |
|---|---|---|
| Accepted | The server has validated and saved the data | An enquiry identifier exists |
| Awaiting delivery | The CRM has not yet confirmed object creation | The enquiry remains in the queue |
| Delivered | The CRM returned success with an object identifier | The record exists and its fields are correct |
| Requires manual review | Automatic retries do not resolve the error | An owner is assigned and the reason is recorded |
The interface confirmation must match the actual state. If an enquiry has been reliably accepted for later delivery, acknowledge acceptance. If it has not been saved, a success message hides the loss. Tie the analytics event to the agreed stage too, so that button clicks are not counted as accepted enquiries.
For a temporary CRM outage, define retries with delays and a limit. For a permanent field error, notify the responsible owner. If access has been revoked or a required field has changed, endlessly repeating the same request will not resolve the failure.
Keep connection secrets on the server. Send only the data the form needs to the visitor's browser. Limit integration permissions to its tasks and include credential replacement and access revocation procedures in the handover.
Distinguish a delivery retry from a new enquiry
These are separate scenarios. A delivery retry occurs when the website receives no response and sends the same accepted enquiry again. A new enquiry occurs when an existing customer requests another service or returns later with a separate request.
Each accepted enquiry needs a stable identifier. A retry uses the same identifier, and the system checks which object it already maps to. This protection is called idempotency: repeating an operation does not create an additional result. Agree its implementation for your API and website storage.
Finding a person by phone or email addresses a different task: linking to an existing contact. Bitrix24 provides a communication-match lookup, but your workflow determines what happens next. One customer can have several independent enquiries that must survive contact merging.
For example, a customer presses submit twice because the response is slow. That is one enquiry with possible repeat delivery. A week later, they request another service using the same phone number. That is a new enquiry from an existing customer. Checking only the phone number risks confusing these situations.
Pay particular attention to a timeout after the CRM has created the record. The CRM may have saved the object even though the website received no response. Before retrying, reconcile the result using the enquiry identifier or an agreed matching attribute. An API method's successful response does not itself provide this protection.
Integration acceptance tests
Ask the contractor to run tests using agreed test data and provide the results. Review the website interface, the enquiry's server-side state and the CRM record together.
| Scenario | Expected result |
|---|---|
| Valid enquiry from each form | Correct object, fields, pipeline and owner |
| Missing required contact | A clear error; no false success confirmation |
| Double click and repeated request | One accepted enquiry without an extra record |
| New enquiry from an existing customer | Contact linked as agreed; the new enquiry is retained |
| Tagged visit followed by navigation across pages | CRM tags follow the chosen attribution model |
| Temporary CRM outage | The accepted enquiry is retained and delivered after recovery |
| CRM response lost after record creation | Retrying does not create a second object |
| Required field changed or access revoked | The owner can see the error and arrange reprocessing |
| Spam and invalid format | Invalid input is rejected under the agreed rule; a valid enquiry passes |
Test forms on mobile devices, on service pages and in modals. Identical appearance does not imply an identical handler. List every form and result in the report so that an untested variant is visible.
First, add your project conditions to the CSV: fields, attribution rules, CRM object and test environment. “Not run” means the check has not been performed. After a test, record the actual result and a link to redacted evidence, such as test request and CRM record identifiers. Empty cells and unavailable diagnostics do not confirm successful delivery.
Test failures in an agreed test environment or using a stub; the template does not include changes to production credentials or real enquiries. If an embedded form prevents you from reproducing a scenario, mark it “Not checked” and describe the limitation. Use “Not applicable” only with an explanation of why the scenario does not apply to the chosen workflow. Repeat checks after changes to the form, CRM fields or connection method.
Recovery requires an owner who can find undelivered enquiries, understand the cause and trigger a retry. Test that procedure before launch, including a scenario where the primary owner has left or is unavailable.
What to include in the implementation brief
Provide the form list, field map, attribution rules, CRM object, routing and acceptance scenarios. Add the access procedure, test environment and support owner. This gives the contractor concrete behaviour to estimate.
The broader project scope is covered in the CRM development budget guide. For website-to-CRM integration, request separate estimates for intake, delivery, retries and failure diagnostics: a single line labelled “CRM” can hide substantial differences in scope.
If your CRM also sends orders to 1C and receives invoice or payment data, define that exchange separately. Fields, exchange directions and acceptance scenarios are covered in the CRM and 1C integration guide.
If you need automation on top of existing data, the next step may be connecting an AI agent to CRM through MCP. First, check that incoming enquiries are delivered and linked to customers correctly.
Discuss the implementation scope and workflow configuration through the CRM implementation service. Attach a completed field map and examples of existing failures to your enquiry.
FAQ
Can you receive enquiries only by email?
Yes, if that workflow suits your team. You must separately define how an enquiry moves into the CRM, who receives ownership and how processing is checked. Email can provide a backup notification, but receiving a message does not prove that a CRM record was created.
Should you retain every independent enquiry from an existing customer?
Retain independent requests according to your sales workflow. Handle a technical retry of one enquiry using its identifier. A matching phone number helps find a contact; it does not prove that the request has already been processed.
Why does Metrica record a conversion with no enquiry in the CRM?
Check when the goal fires: on a click, server acceptance or CRM confirmation. Find the enquiry identifier and delivery state, then review API errors, field settings and access permissions. Until you establish the cause, the systems' figures cannot be treated as interchangeable.
Can you use an existing connector?
Yes, if it covers the required fields, routing, retries and diagnostics. Apply the same acceptance checks. If its functions are insufficient, compare extending the connector with a custom server handler, including support work and ownership.
Sources
- Yandex Metrica: processing source tags.
- Bitrix24: finding communication matches.
- Bitrix24: creating a CRM object with crm.item.add.
- amoCRM: Unsorted and website forms.
Documentation was rechecked on 11 October 2026. The delivery design and acceptance template are recommendations for project planning. The CSV contains blank test results, rather than a report of CRM testing. The capabilities of a specific form or connector must be confirmed separately.
