How to Capture UTM Parameters in Crisp Chats
Crisp conversations often begin after a visitor has browsed several pages. If your only record of campaign data is the URL that first brought them in, that context may already be gone by the time the chat starts. This guide explains a practical way to carry UTM parameters and related attribution details into each new Crisp conversation, then use that data in a CRM or reporting tool.
Why capture UTM parameters with chat conversations
Web analytics can show visits, but it does not automatically connect a visit to a chat, a lead, or a customer. When campaign data travels with the conversation, you can compare channels and campaigns on outcomes rather than traffic alone. The goal is not simply to store a long URL; it is to preserve the source, campaign, and click information that lets you answer which marketing activity produced a conversation and, later, revenue.
Four steps to get UTM data into Crisp
The workflow has four stages: install the attribution script on the website, let it classify and remember each visitor's source, pass the data through when a new chat begins, and send the resulting record to the systems where you report on leads and customers.
1. Install the attribution script
This workflow assumes you are using a third-party attribution tool. Account creation, trial length, and any commercial terms are governed by that provider, so check its current signup page before you begin. After registration, the tool supplies a short code snippet to add to your website. How you add it depends on your platform: many site builders let you paste the snippet into a settings area, while others work better with a tag manager such as Google Tag Manager. Setup guides are commonly available for WordPress, Wix, Squarespace, Webflow, and other major website platforms.
2. Classify and remember the visitor's source
An attribution tool of this kind observes how each visitor arrived and groups the visit into channels such as Paid Search, Paid Social, Display, and Organic Search. Rather than relying only on the current page address, it keeps that record in the visitor's browser. As a result, someone can browse several pages, leave the site, and return later, and the campaign details from the first arrival can still be available when a chat begins. The exact behaviour and the set of recognised channels vary by product, so verify them against the tool you choose.
3. Pass attribution data with each new chat
When a visitor opens a new chat, the attribution data is passed through with the conversation. Consider a gardening-services company. A prospect clicks a Google Ads link, lands on the homepage, browses service pages, and then starts a chat. Depending on how the ad was tagged, the conversation can carry values such as those in the example field map below.
Example field map from a tagged Google Ads visit
The following values illustrate how a single paid-search visit can be represented. They are an example of the data structure, not a claim about any particular account's results.
4. Send the data to your CRM and reporting tools
The destinations available for captured data depend on the tool and the connections it supports. Typical options include viewing attribution details inside the chat window, forwarding them to a CRM such as Salesforce, HubSpot, Pipedrive, or Zoho CRM, or passing them through an automation service like Zapier into a spreadsheet. Once the data is in a CRM or spreadsheet, you can build charts and dashboards in tools such as Looker Studio or Power BI. Check each connection against your own product setup before relying on it for team reporting.
Decision checklist: raw URL capture vs. persisted attribution
Before choosing a method, check the following. Does the method capture the UTM values only while they remain in the current URL? Does it remember the original source after the visitor navigates to another page or returns in a later session? Does it classify inconsistent source spellings into a single channel? Does it pass click identifiers that ad platforms can use for offline conversions? Does it record supporting details such as days to conversion or number of visits before conversion? A method that fails the persistence and classification checks will tend to produce incomplete or messy chat records even when the initial ad link was tagged correctly.
A useful failure distinction
Two very different faults can produce the same empty column in a report. One is a tagging fault: the visitor never carried campaign parameters, so no method could have recorded them. The other is a retention fault: the parameters were on the landing page, but the capture method read only the live address, so once the visitor clicked through to a second page the values were no longer visible to it. Retention faults call for a method that keeps the original values in browser storage; tagging faults call for fixing the ad URLs. Knowing which fault you have stops you from swapping tools when the real problem is a missing or inconsistent tag.
Four reports worth building
Once attribution data reaches a CRM or spreadsheet, the useful questions become concrete. A leads-by-channel report shows monthly volume by channel and is a quick way to spot a channel that supplies many conversations while receiving little budget, or the reverse. A leads-by-Meta-network report breaks results out by placement, so you can see whether Facebook, Instagram, Messenger, or another surface is actually starting the conversations your spend is meant to produce. A customers-by-Google-Ads-campaign report compares individual campaigns on customer creation rather than clicks. A revenue-by-keyword report uses keyword values added to Google Ads tracking templates to show which search terms are associated with revenue.
What to verify before you rely on the data
Confirm that the script is present on every page where a chat can start, that your ad URLs use consistent UTM values, and that the CRM or export destination receives the fields you intend to report on. Then compare a small set of known tagged visits with the conversations they produce. This check is more useful than assuming that installation alone guarantees complete attribution.
| Field | Example value |
|---|---|
| Channel | Paid Search |
| Channel Drilldown 1 | |
| Channel Drilldown 2 | Local Services Campaign |
| Channel Drilldown 3 | Garden Maintenance |
Illustrative example values from a gardening-services scenario; not a report of actual account performance.
| Report | Question it helps answer |
|---|---|
| Leads by Channel | Which channels produce the most leads each month? |
| Leads by Meta Ads Network | Which Meta placement produces the most leads? |
| Customers by Google Ads campaign | Which Google Ads campaign produces the most customers? |
| Revenue by Keyword | Which search keywords are associated with revenue? |
Report names describe the reporting questions illustrated in the source; exact field availability depends on the CRM or export destination.
Frequently asked questions
Why do UTM parameters sometimes disappear before a chat starts?
UTM parameters live in the page URL. When a visitor navigates to another page, those parameters are typically no longer in the address. A method that reads only the current URL can therefore miss the original campaign. Storing the values in the visitor's browser allows the original source to be remembered across later pages or return visits.
Can attribution work for visitors who did not arrive through a paid ad?
Yes, if the method classifies non-paid sources as well. Organic Search, Organic Social, Direct Traffic, and Referral can be recorded alongside paid channels, which lets reports cover every lead rather than only those with UTM parameters.
What are click IDs used for?
Click identifiers such as Google Click ID and Microsoft Click ID can be passed with a conversation and later returned to the ad platform when the person becomes a sales opportunity or customer. That supports offline conversion measurement.
What happens if UTM source values are inconsistent?
Raw values such as Google, google.com, and adwords can appear as separate sources in reporting. A classification layer can group them into a single channel, reducing manual consolidation.