GA4 attribution: what Google Analytics can and cannot tell you about sales that close by phone or in person

GA4 attribution divides credit for key events, such as form submissions and call clicks, among the channels that preceded them. It does not see sales that close on the phone or in person, and cannot take them more than 72 hours late. Use GA4 for what happens on the site; measure closed revenue by matching sales to leads.

Google Analytics 4 is the first place most businesses look when someone asks which channels are working, and its attribution reports look like an answer. There is a model selector, a channel table, a conversion path report and a column of credited key events with decimals in it. The decimals are the giveaway. GA4 is dividing something, and it is worth knowing exactly what before taking the division to a budget meeting.

For an online store whose sales happen at checkout, the thing being divided is close to revenue. For a business whose sales close on a phone call, in a showroom or after a quote, it is not: it is form submissions and call clicks, some of which become customers weeks later and most of which do not. This post covers how GA4 attribution works, what it is right for, and where its view stops.

What is GA4 attribution?

GA4 attribution is how Google Analytics assigns credit for each key event (the actions you mark as conversions, such as a form submission or a purchase) to the channels a visitor came through beforehand. It applies one reporting model across the property, and the model decides how the credit is split.

Three things define what GA4 reports for any channel: the key events you have marked, the reporting attribution model that divides credit for them, and the lookback window that limits how far back a touchpoint can be and still receive credit. Each is a setting in the property, and each changes the numbers in every report that shows credited key events.

GA4 uses the word key event where older tools said conversion. In GA4 they are events a business has flagged as important, nothing more. A key event can be worth a lot, like a purchase, or nothing at all, like a brochure download. GA4 counts both the same way unless you give them a value.

GA4 attribution models compared

GA4 offers three reporting attribution models: data-driven, which is the default; paid and organic last click; and Google paid channels last click. First click, linear, time decay and position-based were removed in November 2023, so any guide that still describes them in GA4 is out of date.

The general case for and against each family of models, including the ones GA4 removed, is in the attribution models comparison. The GA4-specific point is that data-driven is the only multi-touch option left, and it is a model you cannot inspect. Its weights come from Google's machine learning on your property's paths. That is defensible for steering spend inside Google's systems and hard to defend as a figure in a revenue report, because nobody can show their working.

One rule applies to all three models and changes how the Direct row reads.

In practice, a customer who clicked a paid search ad in March and typed your address in to book in April is credited to paid search. That is a reasonable choice, and it is a choice: GA4's Direct row is a floor, not a measurement of how many customers came straight to you. A large Unassigned row usually means campaign tags that do not follow the channel rules; UTM parameters covers the naming that keeps traffic out of it.

GA4 lookback windows and what changes retroactively

GA4's lookback window is 30 days by default for acquisition key events (first visits and first opens), with 7 as the alternative, and 90 days by default for all other key events, with 30 or 60 as alternatives. A model change rewrites history; a lookback change does not.

Two consequences follow. First, switching from data-driven to last click and back again will change last quarter's numbers in place, so a screenshot from before the switch will not match the report after it. Write the model on every exported figure. Second, 90 days is a ceiling. A customer whose first click was four months before the form submission is outside the window for that touch, however much it mattered. The longer argument about window length is in attribution windows.

Why GA4 misses sales that close by phone or in person

GA4 only knows about events sent from a tagged site or app, or pushed to it later with the visitor's GA identifier. A sale agreed on a call, in a showroom or after a site visit happens outside that view, and can only be added back within 72 hours of the date it happened.

For an online store, the purchase is a key event on the confirmation page, and GA4 attribution is attribution of revenue. For a business that sells by phone or in person, the last thing GA4 sees is the lead: a `generate_lead` event on a form, a tap on a phone number. What happens next, whether the lead picked up, got a quote, signed, paid and how much, happens in a CRM, a booking system or a sales spreadsheet.

• Calls that never touched the site. A caller who dials from a Google Business Profile listing, a Local Services ad, a van or a printed card produces no GA4 event at all.
• Leads that close weeks later. The key event GA4 credits is the form fill on day 1. The sale on day 40 is invisible unless you send it.
• Leads that never close. GA4 counts a spam submission, a duplicate and a tire-kicker as one key event each, worth exactly the same as the lead that became your largest job.
• Revenue. Unless the key event carries a value, GA4 has no revenue to attribute; a static value per lead is the same number for every lead, so it is a count with a currency sign.

The obvious fix is to send the closed sale back into GA4 as an event. Google provides a way to do that, and its limits are the reason it rarely works for sales that take weeks.

So a closed sale can reach GA4 only if three things are true: the lead came through your website, you captured the GA client ID with it, and you send the event within three days of the date you want on it. A sale entered in the CRM a week after it was agreed cannot carry its real date, and a sale that closes on day 40 is a day-40 event in GA4, not a property of the day-1 lead. Phone leads from outside the site have no client ID to send. For most businesses that sell offline, GA4 is the wrong ledger for closed revenue, and the ad platforms' own offline imports, covered in offline conversions, are a better destination for the matched sale.

Worked example: GA4 key events against matched closed revenue

Put GA4's credited key events by channel beside closed sales matched to their leads for the same month. The two tables rank channels differently, and the difference is the cost of optimizing for form fills rather than for customers.

Here is an invented home-remodeling company's month. On the left, GA4's data-driven key events (form submissions and call clicks) by default channel group. On the right, the 80 jobs that closed that month, matched by phone number and email to the lead each customer first came in through. The channel names are not a perfect match, because GA4 groups by traffic source and the lead file records how the customer arrived, so read them as pairs.

1. Paid social is the most flattered channel. It has 21.6% of key events and 5.6% of closed revenue. Each closed sale took 13.3 key events, against 6.4 for paid search. GA4 is not wrong that the ads produce form fills; the form fills just close less often and at smaller jobs ($6,500 average against $8,500).
2. The second-best revenue source is missing from GA4. Business Profile callers who never visited the site closed 14 jobs worth $126,000, 17.9% of revenue. GA4 has no row for them because they never fired an event, so any budget split based on GA4 alone treats the listing as worth nothing.
3. Organic search earns the most per lead. $171,000 from 95 key events is $1,800 per key event, against $1,336 for paid search and $488 for paid social. GA4's credited key events would put paid search well ahead; revenue puts them almost level.
4. Direct and Unassigned are not the same thing as Direct / Unknown. GA4's 40 are visits GA4 could not credit to a channel; the right-hand 11 are sales with no matching lead at all. Both deserve a row, and neither should be spread across the others. Unattributed revenue covers how to shrink the second one honestly.

A budget review that read only the left-hand table would move money toward paid social and away from the listing. The right-hand table says the opposite. Neither table is fake; they count different things. The point is to know which one you are holding.

How to measure closed revenue by channel alongside GA4

Keep GA4 for site behavior and lead volume, and measure closed revenue by matching each sale to the lead it came from in your own records. Capture the source on every lead, including calls, then join sales to leads on phone number and email.

1. Record the source on every form lead. Put UTM values and the click ID in hidden fields so the source travels with the lead into the CRM, not only into GA4. Web form leads covers the fields.
2. Give calls a source. Tracked numbers per channel, or at minimum a separate number on the Business Profile listing and on print, so a call is a lead with a source. Call tracking to revenue shows how to get from a call log to closed revenue.
3. Export closed sales with the customer's phone and email. The sale record is the ledger: date, amount, customer contact.
4. Match sales to leads on normalized phone and email. A sale whose phone number matches an earlier lead inherits that lead's source. A sale with no match goes in Direct / Unknown, not into the nearest channel. Lead source tracking sets this up as a monthly routine.
5. Report the two side by side. GA4's key events show which channels bring people to the form; matched revenue shows which bring customers. A channel high on the first and low on the second has a lead-quality problem, not a volume problem.

CloseRev does the matching in steps four and five from two CSV files you export, closed sales and leads. It matches each sale to leads on normalized phone and email and, by default, credits it to the source of the earliest matched lead; last touch is an option. Only high-confidence matches count automatically; medium- and low-confidence ones stay reviewable, and sales with no match stay in Direct / Unknown. It does not read GA4 data, so the key-event column stays a GA4 export.

What GA4 attribution is good for

GA4 attribution is good at what happens on your site: which channels bring visitors who submit forms, which landing pages convert, how paths change after a campaign starts. It is the right tool for lead volume and on-site conversion, and the wrong one for closed revenue.

• Landing page and form performance. Key events per session by landing page, before and after a change.
• Early warning. A drop in key events from one channel shows up in GA4 days before it shows in closed sales.
• Online purchases. Where the sale happens at a tagged checkout, GA4's attributed revenue is real revenue, subject to consent and the model.
• Modeled gaps. GA4 fills in some unobserved activity with modeling; modeled conversions explains what that does to the numbers, and why they should be labeled as estimates.

None of that needs GA4 to know about the sale. It needs GA4 to be honest about the lead, and the sales record to be honest about the customer. Put the two side by side and each one covers the other's blind spot.