Marketing Attribution: Measure ROI in Google Analytics

I have sat in a lot of meetings where someone presents a return on marketing spend and nobody in the room quite believes the number. Usually they are right not to. The number is not wrong because the platform lied. It is wrong because nobody in the room knows which rule was used to hand out the credit.

Attribution is not truth. It is a set of rules for deciding who gets the credit, and if you do not know which rules your account is running, every ROI figure you report to an owner or a board is guesswork wearing a suit. Here is the plain version, the settings that change every number, and a report you can build in an afternoon.

What Marketing Attribution Actually Means

A plain English definition you can repeat to your boss

Attribution is the process of deciding which marketing activity gets credit for an enquiry or a sale. Someone sees an ad, reads a blog post, gets an email, searches your brand name three days later and fills in a form. Attribution is the rule that decides how much of the credit each of those moments receives.

Google Analytics Help describes the mechanism as follows: "An attribution model can be a rule, a set of rules, or a data-driven algorithm that determines how credit is assigned to touchpoints along a user's path to completing important actions." The word doing the work in that sentence is rule. A rule is a decision someone made, and someone could have made a different one.

Attribution versus incrementality

These two get used interchangeably and they answer completely different questions. Attribution describes what happened along the path a customer took. Incrementality asks whether the sale would have happened anyway without that activity.

Attribution is cheap, always available and usually directionally useful. Incrementality needs a test, a holdback group or a controlled pause, and it costs money to run. Almost every argument about marketing spend at a small business is really an incrementality argument being conducted with attribution data, which is why it goes in circles.

Why Last Click Attribution Misleads Small Businesses

What last click overcredits

Last click gives all the credit to whatever touched the customer last. For most small businesses that is branded search or direct traffic, and both of those are usually results rather than causes. Someone types your business name because they saw your van, got a referral or read your post two weeks ago. Branded search gets marked as the hero.

The practical consequence is a budget conversation that makes no sense. The channel doing the closing gets credited and funded, and the channel that created the demand gets cut for appearing to do nothing.

What last click hides completely

Last click also hides the length of the journey. A decision that took eleven touches looks identical to one that took one, because only the final interaction is counted. In a business with considered purchases, such as trades work over a few thousand dollars or professional services, that is where most of the information lives.

And it hides the assist entirely. A channel that consistently introduces customers who convert a month later through a different route shows up as a cost with no return. I have seen businesses cancel a channel that was doing the heaviest lifting in the account, purely because last click could not see it.

The Attribution Models Available in Google Analytics 4

The list is shorter than most people assume, and Google has trimmed it over time. Google Analytics Help states: "There are 3 attribution models available in the Attribution reports in Google Analytics properties:"

Data driven attribution

Data driven is the one that uses your own account data rather than a fixed rule. The same Google Analytics Help page explains: "Data-driven: Data-driven attribution distributes credit for the key event based on data for each key event. It's different from the other models because it uses your account's data to calculate the actual contribution of each click interaction."

In practice this is a modelled distribution. It is more informative than a fixed rule for most accounts, and it carries the risk that comes with any model, which is that it looks more precise than it is.

Paid and organic last click

This gives full credit to the final paid or organic click before the conversion. It is the easiest model to explain to someone who is not a marketer, which is its real value. If you have to present a single number to an owner, this is the one they will understand without a lesson.

Google paid channels last click

This gives full credit to the last click that came from a Google paid channel, and ignores everything else. It is narrow by design, and it is useful mainly if you are trying to understand Google's own paid channels in isolation rather than the whole marketing picture.

The models Google retired in November 2023

Several models people still ask me about no longer exist. The same Google Analytics Help page states: "Note: The first click, linear, time decay, and position-based attribution models are no longer available as of November 2023." If a report template or a dashboard you inherited still references first click or linear, that template is describing a product that no longer exists.

Setting Up Attribution in GA4 Step by Step

Choosing your reporting attribution model

Find the attribution settings in your property admin area and check which reporting model is selected. Most accounts I look at are on data driven by default and nobody has ever confirmed it. Confirm it, and write down what it is, because a number you cannot explain the source of is a number you cannot defend.

Setting the key event lookback window

This is the setting almost nobody checks, and it quietly changes every report. Google Analytics Help describes what it controls: "The key event lookback window determines how far back in time a touchpoint is eligible for attribution credit." The same page gives the defaults: "For acquisition key events (first_open and first_visit), the default lookback window is 30 days. You can switch to 7 days if you have different attribution needs. For all other key events, the default lookback window is 90 days. You can also choose 30 days or 60 days."

Now think about your own sales cycle. If most of your customers take eight weeks to decide, a thirty day window is deleting the first half of the journey. The clicks do not disappear from the platform, they simply stop earning credit, and the channels that started the conversation look weaker than they are.

Linking Google Ads and Search Console so the data is complete

Link both. An unlinked Google Ads account turns your paid traffic into something close to anonymous referral traffic inside Analytics, and an unlinked Search Console leaves you unable to see the organic queries that started the journey. Both links are free, both take minutes, and both are routinely missing.

Once they are linked, check that costs and conversion values are flowing. An attribution report without cost data can tell you what happened but not what it was worth.

Reading the Attribution Reports Without Fooling Yourself

Conversion paths and how many touchpoints actually happen

Open the conversion paths view and look at the length of the top paths rather than the summary. If your best performing path is routinely three or more touchpoints, you have a considered purchase and any single touch report is misleading you by definition.

This is also the fastest way to settle an internal argument. When someone insists their channel drives sales, the paths view shows whether that channel appears early, late, or not at all.

Fractional credit and what the decimals are telling you

If your report shows 3.4 conversions rather than 3, nothing is broken. Under data driven attribution a single conversion can be split across several interactions, so each interaction shows a fraction of it and the fractions add to one.

The decimals are a feature, not a rounding problem. They are the model telling you that no single channel deserves the whole credit, which is usually the most honest thing in the report.

Connecting Attribution to Real Money

Conversion values and the ROAS columns that matter

Conversion counts are a means to an end. What you need is value, and Google Ads exposes a metric built exactly for this. Google Ads Help describes the control exactly as: "Attribution model" setting: For website and Google Analytics conversion actions, you can choose how much credit each of a customer's clicks gets for each conversion. The same page describes the metric that turns clicks into a business number: "Conversion value per cost ('Conv. value / cost') estimates your return on investment. It's calculated by dividing your total conversion value by the total cost of all ad interactions."

Set a realistic value against each conversion action before you read any return figure. A lead form worth several hundred dollars and a newsletter signup worth nothing cannot both count as one conversion, or the ROAS column becomes decoration.

Blended cost per lead across every channel

Total marketing spend divided by total leads, with no attribution model in the middle. This is the number I would take to an owner, because it cannot be gamed by a platform's counting rules and it survives every argument about who deserves credit.

Track blended cost per lead monthly alongside your attributed numbers. When they diverge sharply, the divergence is the finding. It usually means one channel is claiming credit for demand another channel created. A wider ROI analysis is the way to resolve that properly when it matters.

The Limits of Attribution You Must Explain Honestly

Direct traffic and the unattributable gap

Direct traffic is the bin everything unclassifiable falls into. Some of it is people who typed your address. Much of it is people where the tracking broke, the referrer was stripped, or they clicked from an app that passes no information. Do not read direct traffic as a channel with a real volume of loyal visitors.

The bigger issue is the gap. Some portion of your conversions came from activity nothing can see, including word of mouth, a van with your number on it, a conversation at a kids birthday party. That portion is real and no report will ever contain it.

Consent, cookies and modelled conversions

Consent choices, blocked cookies and privacy settings all reduce what gets recorded. Google fills some of the hole with modelling, which is an estimate rather than a count. Treat modelled conversions as a range, not a precise figure, and say so when you present them.

None of this is a reason to distrust the data entirely. It is a reason to stop treating it as exact to the decimal and start reading it as a direction with a known margin of error.

A One Page Monthly Attribution Report for a Small Business

Five numbers that fit on one screen

Total spend for the month. Total leads from every source combined. Blended cost per lead. Attributed conversions by channel, listed as shares rather than raw counts. And the conversion value per cost figure for your paid channels. Five numbers, one screen, no scrolling.

Add one line of commentary: what changed this month and why. That line is what makes it a report rather than a data dump, and it is the only part nobody can generate automatically.

What to change based on what the numbers show

If blended cost per lead is rising while attributed conversions hold steady, you are paying more for the same result and the problem is usually in campaign targeting rather than in measurement. If attributed conversions fall but leads hold steady, you likely have a tracking break, not a performance problem.

If one channel's attributed share drops while total leads rise, check your lookback window before you change any budget, because that pattern is often a settings artefact. And if nothing moves for three consecutive months, you have a business problem rather than a marketing one, and no report will fix it.

When to Bring In Help With Attribution

Signs your measurement cannot currently be trusted

You cannot name your reporting model. Your Analytics and Ads numbers differ by a wide margin and nobody can explain why. Nobody has checked the lookback window. You have conversions recorded with no value against them. Or your reports have been rebuilt so many times that no two months are measured the same way.

Any one of those is survivable. Two or three together mean you are making budget decisions on a number you cannot defend, and that gets expensive at exactly the moment you scale spend.

What a consultant actually sets up

The work is unglamorous and specific. Confirm the reporting model and the lookback window against your actual sales cycle. Link the platforms. Assign meaningful values to conversion actions. Build one monthly report with a fixed definition of every metric and a written note of which source is authoritative when they disagree.

That is the shape of Google Analytics consulting services in practice, and if you are still on a legacy property the sequence usually starts with a GA4 migration so there is a stable history to compare against. A wider marketing audit covers the same ground across every channel if the measurement problem turns out to be one symptom of several.

Before your next owner or board report, check two settings: your reporting attribution model and your key event lookback window. Everything else in this article follows from those two, and both take about five minutes.

Frequently Asked Questions

What is marketing attribution in simple terms?

It is the rule you use to decide which marketing activity gets credit for a sale or enquiry. Google's own definition is that an attribution model "can be a rule, a set of rules, or a data-driven algorithm that determines how credit is assigned to touchpoints along a user's path to completing important actions." The important word is rule, because attribution is a bookkeeping decision rather than a measurement of truth.

Which attribution model should my small business use in GA4?

Start with data driven for your reporting. GA4 offers three models, being Data-driven, Paid and organic last click, and Google paid channels last click, and data driven is the one that uses your own account data to weight each interaction. If you want a simple, explainable number for a monthly owner report, run Paid and organic last click alongside it and show both.

What is a lookback window and why does it matter?

It sets how far back in time a click or visit can still earn credit for a conversion. In GA4 the default for acquisition key events is 30 days, and for all other key events it is 90 days. If most of your customers take eight weeks to decide, a 30 day window is silently throwing away the first half of the journey.

Why does my GA4 report show decimals instead of whole conversions?

Because credit is being shared. When you use data driven attribution, a conversion can be split across several interactions, so each one shows fractional credit that sums to 1.0. That is the model working as intended, and it means no single channel is being handed all the glory.

Can marketing attribution tell me if a channel actually caused the sale?

No, and this is the most important limitation to be honest about. Attribution describes what happened along the observed path. Proving that a sale would not have happened without a specific ad is a different question, answered with incrementality testing or holdback groups rather than with an attribution report.

My Google Analytics and Google Ads numbers do not match. Which is right?

Usually both, for different questions. The platforms count on different clocks, different attribution models and different lookback windows, and Google Ads also applies its own conversion settings. Decide which source you will report from, write it down as the standard, and stop switching between them month to month.