How to Measure Website Conversions When Customers Use Multiple Channels Before Buying
A four-part plan for lead-gen, B2B and service businesses: capture sources, connect CRM outcomes to GA4 and Ads, ask buyers, and read each tool's limits.
How do you measure conversions across several channels?
Do not look for one tool that tells the whole truth. Build a simple measurement plan in four parts: capture the source of every lead consistently, connect what happens in your CRM back to GA4 and your ad platforms, ask buyers how they found you, and read each data source for the question it can actually answer. Then compare the views monthly and act where they agree.
This guide is for lead-generation, B2B and service businesses, where a visit becomes an enquiry and the enquiry becomes revenue weeks or months later, often on a call. Online stores have a different problem (the sale happens on the site); for that, see our guides to ecommerce attribution and choosing an ecommerce attribution model.
Key takeaways
- GA4, ad platforms and your CRM will always disagree, because each sees a different part of the journey.
- Consistent lowercase UTMs and CRM source fields fix more problems than any attribution model.
- Send CRM outcomes (qualified, won) back to Google Ads so bidding learns from revenue, not form fills.
- Ask buyers how they heard of you; it reveals channels no tracking can see.
- GA4 has three attribution models and a lookback window of at most 90 days.
- Consent Mode modelling needs volumes most small sites do not reach.
Why the numbers never match
A typical service buyer might see a LinkedIn post, search your name a week later, read three pages, leave, come back from a colleague's forwarded link, submit a form, and sign a contract after two calls and a proposal. Each system records a slice of that.
GA4 sees website sessions it was allowed to measure. It credits only touchpoints within its lookback window, and a forwarded link or typed address arrives as Direct. Ad platforms see their own ads and claim conversions they can connect to them, so platforms can each claim the same lead. Your CRM sees what was captured on the form and what sales recorded, which is often "Website" for everything unless you capture more. The buyer remembers the colleague's recommendation, which none of the systems saw.
None of these is wrong. They answer different questions, and the plan below is about making each one as accurate as it can be and then reading them together.
Part 1: Capture the source of every lead
Most attribution problems are capture problems. Google's guidance is that you should always use utm_source, utm_medium and utm_campaign, and that parameter values are case sensitive, so utm_source=google and utm_source=Google are reported separately (Google Analytics Help, UTM parameters). GA4 also supports utm_id, utm_source_platform, utm_term and utm_content, among others. Write a convention, put it in a shared sheet and use it for every link you control.
| Parameter | Convention (example) | Notes |
|---|---|---|
| utm_source | linkedin, google, newsletter, partner-name | Where the click came from. Lowercase, no spaces. |
| utm_medium | paid-social, cpc, email, referral, organic-social | The type of channel. Keep a fixed list so GA4 channel groups work. |
| utm_campaign | 2026-q4-audit-offer | Date or quarter plus a short name, so campaigns sort and do not collide. |
| utm_content | carousel-a, footer-link | Which ad or link, for comparing creatives. |
| utm_id | Campaign ID from your sheet | Lets you join cost or campaign data later. |
Pro tip
Tag the links people forget: email signatures, proposal PDFs, social bios, partner directories and links you share in messaging apps. Do not add UTMs to internal links on your own site; that overwrites the real source of the session.
CRM fields to add
UTMs only help if they reach the CRM with the lead. Store them in hidden form fields on submission (and, for first touch, in first-party storage from the first visit, subject to your consent setup). A minimal field schema:
| CRM field | How it is filled | Why you need it |
|---|---|---|
| First-touch source / medium / campaign | Captured on first visit, submitted with the form | What started the relationship |
| Last-touch source / medium / campaign | Captured on the converting session | What prompted the enquiry |
| Landing page | URL of the first page in the converting session | Which content brings leads |
| Click IDs (gclid and others) | Read from the URL when present | Lets you send outcomes back to ad platforms |
| Self-reported source | Visible form field (see Part 3) | What the buyer remembers |
| Lead stage and outcome | Updated by sales | Qualified, proposal, won or lost, with dates and value |
Part 2: Connect CRM outcomes back to GA4 and Google Ads
If Google Ads only knows about form fills, its bidding optimises for form fills, including spam and poor-fit enquiries. Sending later outcomes back changes what the system learns from.
In GA4, Google provides recommended lead events: generate_lead for the enquiry, then qualify_lead, disqualify_lead, working_lead, close_convert_lead and close_unconvert_lead for later stages (Google, GA4 recommended events). Marking generate_lead as a key event is the minimum; sending qualified and won stages from the CRM (usually server side) lets you see which channels produce good leads, not just many.
In Google Ads, enhanced conversions for leads is Google's upgraded form of offline conversion import. Your site sends hashed lead information (such as the email address) from the form through the Google tag or Tag Manager. When the lead later qualifies or closes, you upload the conversion with the same hashed data, and Google matches it. Google advises including the GCLID whenever possible, and it is required if you are not using a tag to collect user-provided data (Google Ads Help, enhanced conversions for leads).
The 2026 change. Google's Help page says that starting 15 June 2026, offline conversion imports and enhanced conversions for leads uploads are migrated to the Data Manager API and blocked in the Google Ads API, except for developer tokens allowlisted for legacy access. Google's listed routes now include Google Ads Data Manager (in the interface or by API) and supported integrations. If your CRM connector, agency script or custom integration was built on the Google Ads API, ask whoever maintains it to confirm it has moved. This is a recent change and Google may revise it, so check the page before building anything new. Once outcomes flow back, use them to diagnose and improve lead quality.
Worth noting
Personal data is involved here. Hashing is not anonymisation, so say in your privacy notice that lead details may be shared with advertising platforms for measurement, and follow the consent rules that apply where your visitors are.
Part 3: Ask buyers how they found you
Self-reported attribution means asking the buyer directly, usually on the enquiry form or in the first call. It catches what tracking cannot: a recommendation over dinner, a podcast, a private community, a post forwarded in a messaging app. It also has limits. People forget, compress a long journey into one memorable moment and pick the first plausible option. Treat it as a different lens, not a correction.
- Use an open text field ("How did you hear about us?") or a short list plus "Other, please tell us". Open text gives richer answers; a list is easier to report on. Many teams use a list and review the "Other" answers monthly.
- Keep the list short and in plain words buyers would use ("Google search", "LinkedIn", "A colleague or friend", "Podcast or event").
- Randomise or alphabetise options so the first one is not over-chosen.
- Make it optional if it hurts form completion; test it rather than assume. Our guide to improving enquiry form conversions covers how to test form changes.
- Ask again on the first sales call and record the answer in the CRM. Spoken answers are often more detailed.
- Group free-text answers into categories monthly. Do not let the raw text pile up unread.
Part 4: Read each source for what it can and cannot say
GA4 currently offers three attribution models: data-driven, paid and organic last click, and Google paid channels last click. First click, linear, time decay and position-based were removed in November 2023 (Google Analytics Help, attribution models). For key events other than first visits, the lookback window defaults to 90 days and can be set to 30 or 60, so 90 days is the maximum (Google Analytics Help, attribution settings). If your sales cycle is longer, early touchpoints simply fall outside GA4's view.
| Source | Good for answering | Cannot tell you |
|---|---|---|
| GA4 | Which channels and pages precede enquiries; on-site behaviour; trends over time | Touchpoints older than the lookback window; journeys on devices or browsers it could not link; visitors it could not measure under consent rules; revenue unless you send it |
| Google Ads, LinkedIn and other ad platforms | Which of their own campaigns, keywords and audiences produce leads (and, with imports, good leads) | Anything outside their platform; they may each claim the same lead |
| CRM | Which sources produce pipeline and revenue; lead quality; time to close | Anything not captured on the form or entered by sales; it is only as good as Part 1 |
| Self-reported | Word of mouth, communities, podcasts, events, private sharing | Precise credit across many touches; small differences between channels |
Key takeaway
Act confidently where at least two views agree. Where they disagree, look for the reason (missing UTMs, a channel invisible to tracking, platform over-claiming) before moving budget.
Consent and the gaps it leaves
Where visitors decline analytics or advertising cookies, some or all of their journey is not measured. Google's consent mode passes visitors' choices to Google tags through settings such as analytics_storage and ad_user_data; in basic mode tags do not load until consent is given, while advanced mode sends cookieless pings (Google, consent mode overview). Google's EU user consent policy applies to end users in the European Economic Area, the UK and Switzerland (Google EU user consent policy).
Modelling can estimate some of the missing data, but GA4's behavioural modelling has thresholds: at least 1,000 events a day with analytics storage denied for at least 7 days, and at least 1,000 daily users with it granted on at least 7 of the previous 28 days. Google says meeting these does not guarantee eligibility (Google Analytics Help, behavioural modelling). Most small lead-generation sites will not reach them, so plan on the CRM and self-reported data carrying more weight. Consent and privacy rules differ by country; take advice for the markets you serve.
Common leaks that send leads to Direct or the wrong source
- Third-party booking, scheduling or payment steps. A visitor who leaves for another domain and returns can start a new session credited to that domain. GA4's unwanted referrals setting tells Analytics not to show listed domains as traffic sources (Google Analytics Help, unwanted referrals).
- Untagged links in email, PDFs, social bios and messaging apps.
- UTMs on internal links, which overwrite the original source.
- Inconsistent case and spelling (LinkedIn, linkedin, li) splitting one channel into several.
- Forms that do not pass hidden fields to the CRM, or a CRM that overwrites the first-touch source on each new form fill.
- Phone and messaging enquiries with no source recorded. Ask on the call, and log it.
A monthly reconciliation routine
Measurement only helps if someone looks. A one-hour monthly review is enough for most small teams.
- Pull leads, qualified leads and won deals by first-touch and last-touch source from the CRM.
- Pull the same period's key events by channel from GA4, with the attribution model noted.
- Pull conversions from each ad platform, noting which are imported qualified or won conversions.
- Group self-reported answers into channels and compare with the CRM sources.
- Note where the views agree, where they disagree and the likely reason.
- Fix one capture problem a month (a missing UTM, a broken hidden field, a referral leak).
- Record decisions and what you expect to change, so next month's review can check them.
Hypothetical example: an accounting firm with four channels
This example is hypothetical. A small accounting firm runs Google Ads, posts on LinkedIn, gets partner referrals and speaks at local business events. GA4 shows most enquiries as Organic Search or Direct, and Google Ads reports more conversions than the firm's actual qualified leads.
The firm adopts a lowercase UTM sheet, adds hidden first- and last-touch fields and a "How did you hear about us?" field to its form, and starts recording the answer again on discovery calls. It sends qualified and won stages back to Google Ads through a supported Data Manager route, and switches bidding to optimise for qualified leads once enough have been imported. After three months the CRM shows paid search producing many enquiries but few qualified ones, while self-reported answers credit events and partners for a large share of won work that GA4 had labelled Direct. The firm moves some budget towards events, keeps search for high-intent terms only, and keeps checking each month whether the pattern holds.
When this plan is more than you need, or not enough
If you get a handful of enquiries a month from one or two channels, start smaller: consistent UTMs, a generate_lead key event and a self-reported field. With low volume, statistical attribution and conversion modelling will not have enough data to learn from, and your time is better spent on the site itself; our guides on CRO for low-traffic websites and on why a website gets traffic but no leads cover that.
At the other end, if you have many channels with significant spend, several sales teams and long enterprise cycles, this plan is a foundation rather than the finish. You may need a data warehouse, marketing mix modelling or dedicated attribution software, and a person who owns the data. Those tools still depend on clean capture, so Part 1 comes first either way.
Common mistakes
- Expecting one tool's report to be the truth.
- Switching attribution models to make a favoured channel look better.
- Optimising ads for form fills when sales outcomes are available.
- Leaving the CRM source as "Website" for every lead.
- Adding a self-reported field and never reading the answers.
- Assuming Consent Mode modelling fills the gaps on a small site.
- Building on the Google Ads API for conversion uploads without checking the 2026 Data Manager change.
Conclusion and next steps
Multi-channel measurement for a lead-gen business is mostly plumbing and habit, not clever modelling. This month, write a UTM convention and add first-touch, last-touch and self-reported fields to your form and CRM. Next, send qualified and won outcomes back to GA4 and Google Ads through a currently supported route. Then hold a monthly review where the views are read side by side. If you want an outside check of how your site and forms capture sources, a CRO audit is one way to get it.
Not sure what your data can tell you?
ZSpace Labs is a remote-first technology studio that works with clients globally on CRO audits and website and CRM integration work. If useful, we can talk through how your UTMs, GA4 setup and CRM source fields fit together and what they can and cannot tell you.
Sources
Google Analytics: attribution models; attribution settings and lookback windows; UTM parameters; identify unwanted referrals; behavioural modelling for consent mode; GA4 recommended events.
Google Ads and consent: enhanced conversions for leads and the Data Manager API change; consent mode overview; EU user consent policy.
Accessed 14 October 2026. Google changes these products and policies often; check the current pages before implementing.
Common questions.
Usually because the visit that produced the lead carried no campaign information. Common causes are links without UTM tags (email signatures, PDFs, messaging apps), the visitor typing the address after hearing about you elsewhere, and redirects through booking or payment tools that start a new session. Tag every link you control, list third-party domains as unwanted referrals, and use a self-reported field for the rest.