Google Shopping Attribution: How Conversion Modeling Works and What to Actually Trust

Attribution Models Available in Google Ads

Attribution determines which ad clicks (or ad exposures) receive credit for a conversion. If a customer clicks a Shopping ad on Monday, then searches again and clicks a text ad on Wednesday before buying, which ad gets the conversion credit?

Google Ads currently offers these models:

Model	How Credit Is Assigned	Best For
Last click	100% credit to the final click before conversion	Simple setups; easy to understand
Data-driven	Machine learning distributes credit based on contribution	Accounts with sufficient conversion volume (recommended)
First click	100% credit to the first click in the path	Awareness campaign measurement
Linear	Equal credit distributed across all clicks	Limited use cases; mostly legacy
Position-based	40% to first, 40% to last, 20% distributed to middle	Legacy; rarely optimal
Time decay	More credit to clicks closer to the conversion	Legacy; not recommended for Shopping

The Problem with Last-Click for Shopping

Shopping ads frequently appear earlier in the purchase journey than text ads. A customer might click a Shopping ad to discover a product, research it via branded search over several days, then convert on a branded text ad. Under last-click, the Shopping ad gets zero credit — making it appear less valuable than it is. This systematically under-bids Shopping campaigns relative to their true contribution.

How Data-Driven Attribution Actually Works

Data-driven attribution (DDA) uses machine learning to analyze your actual conversion paths across customers who did and didn't convert. It assigns credit to each touchpoint based on how much that touchpoint increased the probability of conversion.

The Shapley Value Foundation

Google's DDA uses a version of Shapley values — a concept from cooperative game theory. In simple terms: for each conversion, Google looks at all the different combinations of ad touchpoints that occurred in customer journeys leading to conversion, then figures out how much each individual touchpoint contributed by comparing paths that included it vs paths that didn't.

The result is a credit distribution that reflects actual causal impact rather than just sequence (last click) or arbitrary rules (time decay).

Minimum Data Requirements

DDA requires sufficient data to build a reliable model. Google's minimum requirements are:

• At least 300 conversions in the past 30 days (for the conversion action using DDA)
• At least 3,000 ad interactions in the past 30 days

If you fall below these thresholds, Google automatically defaults to last-click. This is why DDA is less useful for small advertisers — the model simply doesn't have enough data to be meaningful.

What Changes When You Switch to DDA

When you switch from last-click to DDA:

• Campaigns that serve early in the funnel will see their conversion values increase
• Campaigns that primarily close sales (branded, remarketing) may see their conversion values decrease
• Smart Bidding will relearn over 4-6 weeks based on the new attribution signals
• Your reported ROAS may change — but actual revenue shouldn't change, just how credit is distributed

Cross-Device Attribution and Modeling

A major challenge in Shopping attribution is cross-device journeys — a customer browses on mobile, then converts on desktop. Without cross-device stitching, the mobile click looks like it generated zero conversions and the desktop ad looks like it generated the whole conversion from a single click.

How Google Stitches Cross-Device Journeys

Google can stitch cross-device journeys for users signed into their Google account. When a user is logged in on both devices, Google can see that the mobile shopping click and the desktop purchase are the same person.

For users who aren't logged in (or who use different browsers/devices without a Google account), Google uses statistical modeling — essentially estimating the cross-device attribution based on patterns from logged-in users and anonymized signals. This modeling is where reported conversions can diverge from actual purchases.

Enhanced Conversions Help

Enhanced Conversions improve cross-device attribution by using hashed customer data (email addresses) to match ad clicks to purchases more deterministically. If a customer provides their email at checkout, Google can hash it and match it to the same email used to click the Shopping ad on another device.

Setting up Enhanced Conversions is one of the best investments you can make for tracking accuracy. It's available for web and app conversions and is especially impactful for stores with high mobile browse / desktop purchase patterns.

Store Visit Conversions Explained

Store visit conversions measure how many people visited a physical store after clicking a Shopping ad. This is a modeled metric, not a direct measurement — and understanding how it works prevents you from either over-relying on it or dismissing it entirely.

How Store Visit Measurement Works

1. A user clicks a Google Shopping ad and Google records the click with a timestamp and the user's approximate location
2. If the user has location history enabled on their Google account (opt-in), Google's location data can detect when that user subsequently visits a store location you've verified in Google Business Profile
3. Google doesn't report individual visits — it aggregates data across many users and uses modeling to extrapolate total store visits from the sample of users with location history enabled

What Makes Store Visit Estimates Reliable vs Unreliable

• More reliable: Dense urban areas with many customers, large stores, high ad spend (more data)
• Less reliable: Rural areas, small stores, low traffic (estimates based on thin data are less accurate)
• Not reported: If your store visit data falls below Google's minimum threshold for statistical significance, it won't appear in reporting at all

Using Store Visit Data Practically

Treat store visits as a directional signal, not an exact count. If your Shopping campaigns show 200 attributed store visits per month, the real number might be anywhere from 150-300. What you can reliably use it for:

• Comparing periods (did store visits increase after a campaign change?)
• Comparing campaigns (which campaign type drives more in-store traffic?)
• Calculating total ROAS including offline revenue

Privacy-First Conversion Modeling

Privacy regulations (GDPR, CCPA, Apple's App Tracking Transparency) and third-party cookie deprecation have significantly reduced the observable portion of conversion paths. In response, Google has expanded its use of modeling to fill the gaps.

What Is Modeled Conversion Data?

When Google can't directly observe a conversion (because the user declined cookies, used Safari's ITP, or is in a jurisdiction with strict privacy rules), it uses statistical models to estimate whether a conversion occurred based on similar observable signals.

In Google Ads and GA4, you'll see two types of conversion data:

• Observed conversions: Directly tracked via conversion tags, Enhanced Conversions, or Google Analytics
• Modeled conversions: Statistically estimated for users who couldn't be tracked

The proportion of modeled vs observed varies significantly by market. In Germany (very strict privacy enforcement) or Safari-heavy audiences, 20-40% of reported conversions may be modeled. In markets with lower privacy adoption, it may be 5-10%.

Consent Mode and Modeling

If you use a cookie consent platform (which you should for EU traffic), implementing Google's Consent Mode V2 is critical. Consent Mode tells Google which users have consented to tracking, allowing Google to use modeling for non-consenting users rather than dropping the data entirely.

Without Consent Mode, Google has no signal for non-consenting users. With Consent Mode, Google can model their behavior from consenting users' patterns. This typically recovers 15-30% of reported conversions that would otherwise be missed.

Why Your Numbers Don't Match GA4

Nearly every advertiser notices that Google Ads conversion numbers don't match GA4 transaction numbers. The gap can be 10-30% in either direction. Here's why:

Google Ads Reports More Conversions Than GA4

Common reasons Google Ads is higher:

• Cross-device modeling: Google Ads includes modeled cross-device conversions; GA4 may not stitch the same journeys
• Attribution window differences: Google Ads uses a 30-day click / 1-day view window by default; GA4 uses session-based attribution by default
• Modeled conversions: Google Ads includes statistically estimated conversions from non-tracking users; GA4 only reports observed sessions
• Duplicate tracking: If you have both Google Ads conversion tags and GA4 goals imported as conversions, you may be double-counting

GA4 Reports More Transactions Than Google Ads Conversions

Less common but possible reasons GA4 is higher:

• Non-paid traffic: GA4 counts all transactions; if you have significant organic, email, or direct traffic converting, these won't appear in Google Ads
• Attribution cutoff: A customer clicked an ad 35 days ago but converted today — past Google Ads' attribution window but counted in GA4

Using Attribution Data to Make Better Decisions

Set Your Attribution Window Thoughtfully

Google Ads defaults to a 30-day click attribution window. For most Shopping campaigns, this is appropriate — most customers decide within 30 days. But if your product has a longer purchase cycle (home improvement, automotive), consider extending to 60 or 90 days. If your product is impulse-purchase (fast fashion, consumables), a shorter window (7 days) may be more accurate.

Don't Optimize on Modeled Data Alone

If your account relies heavily on modeled conversions (30%+ of reported conversions), be more conservative with ROAS targets. Modeled conversions are estimates — they'll be wrong in some proportion of cases. A conservative buffer in your targets accounts for this uncertainty.

Use Conversion Value Rules for Higher-Margin Products

Attribution tells you which clicks converted — but it doesn't weight high-margin product conversions more heavily than low-margin ones by default. Use conversion value rules in Google Ads to assign higher value to conversions for your hero products. This gives Smart Bidding more accurate profit signals, not just revenue signals.

Keeping your GMC account healthy ensures your product listings remain active — and your attribution data stays meaningful. A suspended account or a wave of disapprovals will distort your conversion data for weeks. Run the GMCUnbanned free scan to check your account health proactively.