Understanding Attribution in Google Ads: Data-Driven, Last Click & Co.

One conversion, three truths: Your Google Ads account shows 100 conversions, GA4 shows 72, your shop backend 85. All three numbers are "correct" – they just answer different questions. If you don't understand this, you'll draw wrong conclusions and shift budget into the wrong campaigns.

This article explains how attribution works in Google Ads today, why data-driven attribution is the standard, why GA4 fundamentally shows different numbers, and what this concretely means for your bidding and budget decisions.

Why Attribution Decides Your Budget

Attribution answers a simple question: Which touchpoint gets credited with the conversion?

That sounds academic, but it's hard-nosed budget control. An example:

Customer journey of a buyer:
1. Click on generic search ad "running shoes cushioning" (research)
2. 3 days later: click on Shopping ad (comparison)
3. 5 days later: click on brand ad "yourshop running shoes" → purchase €120

With last click, the brand campaign gets the full €120. The generic campaign that brought the customer in the first place sits at €0 – and looks unprofitable.

The consequence of wrong attribution: You cut the budget of the generic campaign because it "doesn't perform." Three months later, everyone wonders why brand conversions are collapsing too. The mistake wasn't in the campaign but in the measurement model.

This is exactly why the attribution model is one of the most consequential settings in the entire account – even though it's just a dropdown in the conversion settings.

The Attribution Models in Google Ads Today

Google has heavily thinned out the field in recent years. First click, linear, time decay, and position-based were abolished in 2023. Two models remain:

| Model | How it works | Status | |-------|-------------|--------| | Data-driven attribution (DDA) | Distributes conversion value across multiple ad clicks based on data | Default for most conversion actions | | Last click | The last Google Ads click gets 100% | Selectable, sensible only in special cases |

Data-Driven Attribution Explained

Using machine learning, DDA compares the conversion paths of buyers with those of non-buyers and calculates from this what contribution each click actually made. Simplified:

What the model observes:
Paths WITH a click on the generic ad → convert at 4%
Paths WITHOUT this click → convert at 2%

Conclusion: The generic click significantly increases
the conversion probability → it receives a corresponding
share of the conversion value.

The result is fractional conversions: A campaign can be credited with 0.4 conversions from one journey. That's why you see decimal numbers like "23.7 conversions" in reports – that's not a bug, it's DDA.

Important to know:

• DDA only evaluates Google-owned touchpoints (Search, Shopping, YouTube, Display – depending on measurability). Your newsletter, SEO, or Meta Ads don't appear in the model
• DDA is a black box: you cannot inspect or recalculate the weighting
• DDA used to require high minimum data volumes – today it's available to practically all accounts; with very little data, Google falls back on more conservative methods

When Last Click Still Makes Sense

• You deliberately want to evaluate only the closing click (e.g., for certain controlling views)
• Very short, single-step journeys (local business, one search, one call)
• You need comparability with historical last-click data

For most accounts: leave DDA active. It's the foundation on which Smart Bidding works best.

Conversion Windows: The Underrated Lever

Besides the model, the conversion window (lookback window) decides how long after a click a conversion is still counted:

| Setting | Default | Adjustable to | |---------|---------|---------------| | Click-through conversion window | 30 days | 1-90 days | | Engaged-view / view window | shorter (1-3 days typical) | limited adjustability |

Why this matters:

• Window too short: With long decision cycles (B2B, high-priced), late conversions fall through the cracks → campaigns look worse than they are
• Window too long: With impulse purchases, Google still attributes conversions weeks later that have little to do with the ad anymore

Practical recommendation: Look at the days to conversion report in Google Ads (attribution reports). If 95% of your conversions happen within 14 days, a 30-day window is fine. If 20% come after day 30, you're throwing away data – extend the window.

And: The conversion window directly affects Smart Bidding, because what isn't counted, the AI can't learn from.

Google Ads vs. GA4: Why the Numbers Never Match

The most common question in every reporting meeting: "Why does GA4 show fewer conversions than Google Ads?" The answer: Both systems measure different things – by design.

| Difference | Google Ads | GA4 | |------------|-----------|-----| | Perspective | Ads-centric: "What did my ads contribute?" | Cross-channel: "Which channel was involved?" | | Attribution logic | DDA/last click only across Google Ads clicks | DDA across all channels (organic, email, referral …) | | Time of counting | Conversion is assigned to the click date | Conversion is assigned to the conversion date | | View conversions | Can be included (e.g., engaged views) | No | | Counting method | "Every" or "One" per conversion action | Event-based, own counting logic |

The most important example to understand:

June 1: Click on Google ad
June 8: User returns via newsletter and buys

Google Ads: Conversion on June 1, attributed to Google Ads
            (last ADS click counts, other channels invisible)
GA4:        Conversion on June 8, shared between email + paid search
            (cross-channel attribution)

Both systems are right – from their respective perspectives. Deviations of 15-40% between Google Ads and GA4 are normal and no indication of broken tracking. It only becomes suspicious with deviations well beyond that, or if one of the systems suddenly drops to zero – then it's worth looking at your conversion tracking.

Practical rule:

• Bidding and optimizing: by Google Ads numbers (that's what Smart Bidding optimizes on)
• Comparing channels and evaluating budget mix: in GA4
• Truth about business success: backend/CRM (orders, revenue, margin)

Practical Consequences for Bidding Strategies

Attribution isn't a reporting topic – it directly feeds your bids:

1. Smart Bidding Optimizes on Attributed Conversions

Target CPA and Target ROAS work with the conversion values that the attribution model assigns. If you change the model, the numbers the AI learns from change – therefore:

• Model change = data change: After switching (e.g., last click → DDA), expect 2-4 weeks of fluctuations, no hectic target adjustments
• Upper-funnel campaigns benefit from DDA: Generic and Shopping campaigns finally get credited with the value they contribute – Smart Bidding bids there again

2. Evaluate Brand Campaigns Honestly

Under last click, brand campaigns almost always look brilliant (CPA €5!) and generic campaigns expensive (CPA €80!). Under DDA, this gets put into perspective. Before you shift budget from generic to brand, check in the attribution report (Tools → Measurement → Attribution) how many brand conversions had a generic first contact.

3. Read Reports with Time Lag in Mind

Because Google Ads assigns conversions to the click date, the last 7-14 days in your reports are systematically incomplete – late conversions are still trickling in. If you finalize last month's performance evaluation on the first of the month, you're evaluating too early. Rule of thumb: know your account's conversion lag and delimit reporting periods accordingly.

4. Conversion Values Instead of Conversion Counting

DDA unfolds its value above all with values: If every conversion has the same value (or none at all), the model can distribute, but Smart Bidding can't distinguish between valuable and cheap conversions. Especially in lead gen, it pays to pass leads with different values.

Typical Misunderstandings

"DDA invents conversions"

No. DDA distributes the same total amount of conversions differently – across more clicks. The sum across the account stays (almost) the same, only the distribution across campaigns changes.

"GA4 shows less, so Google Ads is inflated"

Not necessarily. GA4 only sees what happens on the website with the GA4 tag, attributes cross-channel, and assigns to the conversion date. Google Ads also counts modeled and late-attributed conversions on the click date. Different questions, different answers.

"I'll switch to last click, then the numbers will match my gut feeling again"

Understandable – last click feels "more honest" because a conversion belongs to exactly one click. But you'll then systematically optimize for the last step of the journey and starve everything that creates demand. The feeling improves, the results usually don't.

"Attribution solves my tracking problem"

The other way around: Attribution is only as good as the data foundation beneath it. If conversions are missing due to ad blockers, ITP, or consent rejections, even the best model distributes incomplete data. Fix the measurement basis first – we show how in our tracking service.

"The backend shows different numbers, so everything is broken"

The backend counts orders, Google Ads counts attributed ad conversions, GA4 counts cross-channel events. Three systems, three definitions. It's only broken when it can't be explained.

Checklist: Setting Up Attribution Cleanly

• [ ] Check the attribution model per conversion action (Goals → Conversions → Settings) – when in doubt, DDA
• [ ] Adapt the conversion window to your real conversion lag ("days to conversion" report)
• [ ] Check the counting method: purchases = "Every", leads = "One" (don't count multiple form submissions twice)
• [ ] Cleanly separate primary vs. secondary conversion actions (only primary ones feed into bidding)
• [ ] Pass conversion values – even for lead gen (estimated lead values are better than nothing)
• [ ] Document the GA4 deviation once (e.g., "GA4 typically shows ~25% less") – then nobody panics in reporting anymore
• [ ] When changing model or window: note the date and don't jump to conclusions for 2-4 weeks

Conclusion

Attribution is not a numbers game for analysts – it determines where your money flows. The most important points:

1. DDA is today's standard – sensible for almost all accounts because it reflects the contribution of the entire journey and feeds Smart Bidding better
2. The conversion window must match your buying cycle – otherwise data is missing or too much is attributed
3. Google Ads ≠ GA4 is normal – different perspectives, different counting logic. Bid by Google Ads, compare channels in GA4, find truth in the backend
4. Measurement basis first, then model: incomplete tracking makes any attribution worthless

If you're unsure whether your account optimizes on the right data, get in touch – an attribution and tracking check uncovers the typical problems quickly.

---

FAQ on Attribution in Google Ads

Which attribution model should I use in Google Ads?

For the vast majority of accounts: data-driven attribution (default). It distributes conversions across all participating ad clicks based on data and is the best foundation for Smart Bidding. Last click only makes sense in special cases – such as very short, single-step journeys or when you deliberately want to evaluate only the closing click.

Why does GA4 show fewer conversions than Google Ads?

Because both systems measure differently: GA4 attributes cross-channel (email, SEO, referral also get shares), assigns the conversion to the conversion date, and doesn't count view conversions. Google Ads attributes only across Google Ads clicks, assigns to the click date, and uses modeling. Deviations of 15-40% are normal and not a tracking error.

What is the right conversion window?

That depends on your buying cycle. Check the "days to conversion" report: If almost all conversions come within 7-14 days, the 30-day standard window is more than enough. For B2B or high-priced products with long decision paths, extending to 60-90 days can make sense so late conversions aren't lost.

Does changing the attribution model change my performance?

Not the real performance – but the measured distribution. After a switch (e.g., last click → DDA), conversions shift between campaigns: brand loses optically, generic and Shopping campaigns gain. Smart Bidding then needs 2-4 weeks to adjust to the new data. During this time: no hectic budget or target changes.

Why does my campaign have 12.4 conversions – decimal numbers?

That's data-driven attribution: A conversion is distributed proportionally across multiple participating clicks. If your campaign "contributed" to a journey, it gets credited with e.g. 0.4 conversions. Across the entire account, the shares add up to whole conversions again.