Marketing Attribution Models: What to Trust When You Sell on Calls

Marketing attribution models are the rules that decide which touchpoint gets credit for a conversion: the first click, the last click, every touch equally, or a split learned from your data. If you sell on sales calls, do not trust any single platform's number. Trust booked calls and closed revenue in your CRM, tied to the ad that started them and checked against what the buyer says brought them in.

What is a marketing attribution model?

A marketing attribution model is a rule for splitting the credit for one conversion across the touchpoints that came before it. Take a typical high-ticket journey: a buyer clicks through from your Reel, comes back a week later through a Google search ad, clicks two nurture emails, then books a call. First-touch gives the Reel all the credit. Last-touch gives it to the second email. Linear splits it four ways. None of them is wrong. Each one answers a different question.

Attribution modeling in marketing is the practice of applying one of these rules to your journey data so you can decide where the next dollar goes. Every model falls into one of three families:

• Single-touch: one touchpoint gets all the credit (first-touch, last-touch, last non-direct click).
• Rule-based multi-touch: credit is split by fixed weights that someone chose (linear, time-decay, position-based, W-shaped).
• Data-driven: an algorithm learns the split by comparing paths that converted with paths that did not.

Here is what none of them does: prove that an ad caused the sale. Attribution assigns credit. It does not tell you what would have happened without the ad. That takes a lift test, covered below. Keep that line clear and half the arguments about attribution disappear.

Marketing attribution models compared

These are the eight models you will meet in ad platforms, analytics tools and agency reports, and the situation where each one quietly misleads you.

Notice what no row can fix. Every model redistributes credit among the touches it can see. A podcast interview, a referral from a past client or a YouTube video watched on a TV never enters the path. That is why the setup further down adds a second source of truth.

Four of these models are gone from Google's tools

If an older guide tells you to switch GA4 to linear or time-decay, it is out of date. Google removed the first click, linear, time decay and position-based models from GA4 as of November 2023. GA4 now offers three: data-driven, paid and organic last click and Google paid channels last click. Google Ads offers only last click and data-driven, data-driven is the default for most conversion actions, and conversion actions that used the retired models were upgraded to data-driven. Want a linear or position-based view today? Build it from your own CRM or touchpoint export.

Which attribution model does each ad platform use?

Each platform measures from its own point of view. Know the defaults before you compare anything.

Meta: 7-day click, 1-day view

Meta's default attribution setting counts a conversion if it happens within 7 days after someone clicks your ad or within 1 day after someone views it without clicking. In Meta's Marketing API reference, the default window is defined as 7d_click plus 1d_view.

The view-through day is where results inflate. A warm lead who was going to book anyway scrolls past your ad at breakfast and books at lunch. Meta counts that booking. Nobody lied: the ad was seen. But it did not cause the call. The warmer the audience, the bigger the effect, which is why a retargeting campaign can look like the best campaign in the account.

Meta also changed what counts as a click. On March 3, 2026, Meta announced that click-through attribution for website and in-store conversions would include only link clicks. Conversions that followed a share, a save or another non-link action moved into engaged-view attribution, renamed engage-through attribution, and the engaged view for video ads dropped from 10 seconds to 5. If your Meta-reported conversions shifted this year with no change in your funnel, start there.

One more trap: Meta's reporting can date a conversion to the day of the ad impression rather than the day of the conversion. The API calls this the action report time. Line up a Meta week against a CRM week and the same bookings can land in different weeks.

Google Ads: data-driven by default

Google Ads uses data-driven attribution by default for most conversion actions, and the only alternative left is last click. Data-driven spreads credit across your Google Ads interactions using your account's own conversion paths. It sees Google touches. It does not see your Meta ads, your emails or your podcast. Google Ads also records conversions against the date of the ad click in its standard columns; separate by conv. time columns show the day the conversion actually happened.

GA4: three models, and no view of other platforms' impressions

GA4's models all exclude direct visits from credit unless the whole path was direct, so paid and organic last click works as a last non-direct click model. GA4 sees sessions, not impressions on other platforms, so a Meta view-through conversion is invisible to it. And a paid click that lost its UTMs or click ID on a redirect shows up as direct or organic. If your landing pages bounce through redirects, read how GA4 now flags landing pages that strip ad parameters before you trust any channel report.

Why don't Meta, Google Ads and GA4 numbers add up?

Because each one answers a different question with different rules. Add the platform numbers together and you will count the same booked call two or three times.

• Each platform credits itself. Meta attributes the booking to Meta. Google Ads attributes it to Google. Neither knows the other touched the buyer.
• Different windows. Meta's default stops at 7 days after a click. GA4 and Google Ads run on their own lookback settings.
• View-through. Meta counts some conversions from people who never clicked. GA4 cannot see those at all.
• Different dates. Google Ads books conversions to the click date, Meta can book them to the impression date, and your CRM records the day the call was booked.
• Lost signal. Consent choices, ad blockers, cross-device journeys and stripped parameters remove touches from some tools and not others. Google fills part of the gap with modeled conversions, which are estimates, not observed events.
• Different definitions. A lead in Meta is whatever event you told the pixel to fire. In the CRM it is a contact. On the sales floor it is a qualified call. Same word, three numbers.

The fix is not a better model. It is one ledger. Make the CRM the source of truth for booked calls, shows, closes and cash, and use platform numbers only for in-platform decisions: which ad set, which creative, which bid. For a week-by-week way to read Meta against what actually happened after the click, use our weekly Meta diagnosis method.

How to attribute a funnel that closes on sales calls

This is the setup to build when the sale happens on a call. The platform does not matter (GoHighLevel, ClickFunnels, Webflow, WordPress or custom code). The chain does: click, landing page, application, calendar, CRM, close. Break one link and every report after it is fiction.

1. Capture the source on landing. Tag every ad URL with UTMs and the platform's own IDs. On Meta, dynamic URL parameters such as {{campaign.id}}, {{adset.id}} and {{ad.id}} fill themselves in at click time. Keep the gclid and fbclid the platforms append. Store everything in a first-party cookie the moment the page loads, and save first touch separately so a second visit never overwrites it.
2. Carry it through the form and the calendar. Pass the stored values into hidden fields on the application and into the booking, so the CRM contact holds first-touch and last-touch source, campaign, ad ID and click ID. Test the handoff between form and calendar with a real booking.
3. Name events by business meaning. In our builds, Meta's standard Lead event fires only on a submitted application. Lighter actions, like an opt-in or a quiz start, are custom events. A booked call fires Schedule. Qualified, showed, closed and cash collected live as CRM pipeline stages.
4. Attribute calls and revenue, not leads. Report cost per booked call, show rate, close rate and cash collected by campaign and by ad, straight from the CRM. A campaign with cheap leads and empty calendars is not your best campaign.
5. Ask the buyer. Put a required How did you hear about us? question on the application. Self-reported attribution catches what pixels miss: podcasts, YouTube, a friend's recommendation, a post seen last month. When tracking says Meta and the buyer says a podcast, both are usually true. The podcast created the demand and the ad captured it.
6. Send outcomes back to the platforms. Send booked calls and closed deals to Meta through the Conversions API, with the same event_name and event_id as the browser pixel so Meta can deduplicate them. Meta's server event spec includes an action source for conversions made over the phone and rejects events with an event time more than 7 days in the past, so push closes when they happen, not in a month-end batch. On Google Ads, enhanced conversions for leads imports offline conversions matched on hashed email or phone, plus the gclid where you have it.
7. Reconcile weekly. Line up platform-reported conversions against CRM booked calls by campaign. The gap is your measurement error. When it suddenly widens, a link in the chain broke.

This is the scoreboard to optimize against. On Dillon Kivo's high-ticket book-a-call funnel, fed by Meta ads we ran, the number that mattered was booked calls: they went from 44 to 94 in 30 days with unique visitors up only 10%, and landing page conversion went from 5.3% to 10.3%. If Meta is your main channel, our Meta ads management runs on this same setup.

Which attribution model should you use?

Pick by situation, not by what sounds sophisticated.

• If you run one paid channel and sell on calls, keep the platform default for bidding and make budget decisions on CRM booked calls and cash by ad. The model barely matters.
• If you run Meta and Google together, never add their conversions. Use GA4 data-driven for the channel mix and the CRM for the truth.
• If your sales cycle runs for weeks, compare first-touch and last-touch side by side in the CRM. First-touch shows what creates demand. Last-touch shows what converts it.
• If you want one blended number for a call funnel, use a W-shaped view with milestones at first touch, application and booked call.
• If you are about to cut a channel, run a holdout first. Attribution cannot tell you what happens without it.

When attribution is not enough: lift tests and marketing mix modeling

Attribution answers who touched the sale. Incrementality answers whether the sale would have happened anyway. When the budget decision is big, you need the second answer.

• Holdout and lift tests. Split the audience or a set of regions into a group that sees your ads and a group that does not, then compare outcomes. Google Ads offers Conversion Lift by user or by geography, and the geography version supports offline data like closed deals, though it is not available on every account. The do-it-yourself version: pause a channel in a few comparable regions for a few weeks and watch booked calls in the CRM.
• Marketing mix modeling. MMM uses spend and outcomes over time, across channels, to estimate what each channel contributes, without tracking individuals. Google's Meridian is an open-source MMM, and Meridian GeoX lets you feed experiment results into the model as priors. We broke down what changed in Google's Meridian GeoX rollout.

MMM earns its place when you run several channels at once, have enough spend history for the model to tell them apart, and the budget calls are big enough that a wrong split costs more than the analysis. One channel at modest spend? Skip it. CRM attribution, self-reported source and an occasional holdout will tell you more.

Leads in the ad account, empty calendar in the CRM? Get your funnel and tracking rebuilt so every booked call and closed deal traces back to the ad that started it.