"We didn't change anything, but conversion numbers shifted overnight." If you've experienced this — or if you're staring at mismatched numbers between Google Ads and GA4 — the cause might not be a broken tag. It might be your attribution model.
Attribution models determine which touchpoint gets credit for a conversion. Change the model, and the same set of conversions redistributes across campaigns, keywords, and channels — without a single tag change. It's one of the most impactful settings most advertisers never deliberately choose.
This guide explains what each model does, why it changes your numbers, how to pick the right one, and how to verify what's actually applied.
1. What attribution models actually do
A user might interact with your ads multiple times before converting: they click a brand search ad, come back via a display remarketing ad, then convert after clicking a non-brand search ad. Which of those three clicks "caused" the conversion?
Attribution models answer that question — and each model answers it differently.
| Model | Who gets credit | Best for |
|---|---|---|
| Last click | 100% to the final click before conversion | Simple reporting; matches older Google Ads defaults |
| First click | 100% to the first click in the path | Understanding what initiates customer journeys |
| Linear | Equal credit to every click in the path | Seeing the full funnel contribution |
| Time decay | More credit to clicks closer to conversion | Balancing awareness and closing touchpoints |
| Position-based | 40% first, 40% last, 20% split across middle | Valuing both discovery and closing |
| Data-driven (DDA) | Machine-learned credit based on actual conversion patterns | Google's default and recommended model |
Important: Google Ads deprecated first click, linear, time decay, and position-based models in 2023. New conversion actions default to data-driven attribution. If you still see the older models in your account, it's because they were set before deprecation — they still function but can't be newly selected.
2. Why switching models changes your numbers
When you change from last click to data-driven, the total number of conversions doesn't change — the same conversions happened. But the distribution shifts:
- Brand campaigns typically lose credit (they're often the last click, so they were over-credited under last click)
- Upper-funnel campaigns (display, video, broad-match search) typically gain credit (they assisted but never got the last click)
- Fractional conversions appear — a single conversion might show as 0.3 conversions on one campaign and 0.7 on another
This is not a measurement error. It's a more accurate picture of how your campaigns work together. But if you're not expecting it, it looks like something broke.
The GA4 difference
GA4 uses its own attribution model, and it defaults to data-driven for its paid channels and last click for organic. This means:
- The same conversion can show different campaign credit in Google Ads vs. GA4
- GA4's "Key events" report and Google Ads' "Conversions" column may disagree on which campaign drove a conversion — even when the underlying event is the same
This is not a bug. It's two systems applying different attribution logic. See Why Google Ads and GA4 counts don't match for the full breakdown.
3. How to check which model is applied
In Google Ads
- Go to Goals → Conversions → Summary
- Click on a specific conversion action
- Under Settings, find Attribution model
- It shows the currently applied model (e.g., "Data-driven")
Each conversion action has its own model setting. If you have five conversion actions, they can each use a different model — and frequently do, especially in accounts that have been running for years.
In GA4
- Go to Admin → Attribution settings
- Check the Reporting attribution model (applies to all reports)
- Check the Lookback window (how far back GA4 looks for touchpoints)
Unlike Google Ads, GA4 applies one model across all events. You can't set per-event models.
4. How to choose the right model
For most advertisers, the answer is straightforward: use data-driven attribution. It's Google's default, it's the most accurate for Smart Bidding, and it automatically adjusts credit based on your actual conversion data.
Switch away from DDA only if:
- You have very few conversions (<50/month) and the model doesn't have enough data to learn — in this case, last click is more stable
- You need to match an external reporting system that uses a specific model (common in enterprise setups)
- Your business has a single-touchpoint buying journey (one click → one conversion, no assist paths)
Choosing by business model: B2C vs. B2B
The "just use DDA" recommendation assumes Google has enough data. When it doesn't — or when you need a rule-based model for transparency — the right choice depends on your buying cycle:
| Business type | Recommended model | Why |
|---|---|---|
| B2C (short cycle) | Linear | Purchase decisions are quick; every touchpoint along the short path matters roughly equally |
| B2B (long cycle) | Position-based (40/20/40) | Long sales cycles mean the first touch (discovery) and last touch (closing) are disproportionately important; middle touches maintain engagement but rarely change intent |
| B2C subscriptions | Time decay | Recent touches are stronger predictors of signup than older awareness touches |
| B2B with offline close | Last click (with offline conversion import) | When the real conversion happens in a sales call, the last online touch is the most actionable signal |
This matters most when DDA doesn't have enough data to learn — typically below 50–100 conversions per month. Above that threshold, DDA usually outperforms any rule-based model because it adapts to your actual conversion patterns rather than applying a fixed formula.
The "consensus across models" approach
Instead of asking "which model is right?", a more practical question is: do multiple models agree on the same conclusion?
Run the model comparison report in Google Ads (Tools → Attribution → Model comparison). If a campaign shows strong performance under last click, DDA, and position-based — you have high confidence that campaign is genuinely driving value. If a campaign only looks good under one specific model, its contribution is model-dependent and less reliable.
This approach is especially useful when making budget decisions. Rather than arguing about which model is "correct," look for campaigns where the signal is consistent regardless of model choice. Where models disagree, dig deeper before scaling or cutting.
What about cross-channel attribution?
Google's attribution models only credit Google touchpoints. If a user clicks a Meta ad, then a Google ad, then converts — Google gives 100% of the credit to its own touchpoint. To get cross-channel attribution, you need a tool outside of Google Ads (GA4 with cross-channel DDA, or a third-party attribution platform).
Meta's attribution model
Meta uses a different attribution framework entirely. Instead of modeling a full click path like Google, Meta uses attribution windows — fixed time periods after a click or view:
| Setting | What it counts |
|---|---|
| 7-day click (default) | Conversions within 7 days of an ad click |
| 1-day click | Conversions within 1 day of an ad click |
| 1-day view | Conversions within 1 day of viewing (but not clicking) an ad |
| 7-day click + 1-day view | Default combination for most campaigns |
Key differences from Google:
- Meta counts view-through conversions by default (Google Ads does not, unless you enable it)
- Meta's attribution is window-based, not path-based — it doesn't split credit across touchpoints
- Starting in 2026, Meta introduced engagement-through attribution, which counts conversions after a user engages with an ad (likes, comments, shares) even without clicking — this inflates conversion counts compared to click-only attribution
When comparing Google Ads and Meta performance, ensure you understand the attribution window each platform uses. A "conversion" in Meta with 1-day view + 7-day click is a fundamentally different measurement from a last-click conversion in Google Ads.
5. What happens to historical data when you switch models
A common concern: "If I change my attribution model, does it rewrite my past reports?"
The answer depends on the platform:
Google Ads: Yes — retroactive. When you change the attribution model on a conversion action, Google recalculates credit distribution for past conversions. Your historical reports will show different per-campaign numbers than they did before the switch. The total conversions remain the same, but how they're distributed across campaigns changes, even for past dates. This can be disorienting if you've already reported those numbers to stakeholders.
GA4: No — prospective only. Changing the attribution model in GA4 applies from the change date forward. Historical reports continue to show the old model's distribution. This means you'll see a discontinuity in your reports at the switch date — a sudden shift in how credit is distributed, not because anything changed in your campaigns, but because the attribution logic changed.
Meta Ads: Meta doesn't offer model switching in the same way — you choose an attribution window, not a model. Changing the window (e.g., from 7-day click to 1-day click) affects reporting retroactively for the selected date range.
Practical advice: Before switching models, export key reports under the current model. After switching, you can compare the old export against the new (retroactive) numbers to understand exactly how credit shifted. This comparison is the single best way to calibrate your intuition about what the new model is doing.
6. Impact on Smart Bidding
Your attribution model directly affects how Smart Bidding learns. Under last click, the algorithm only sees the final click as "valuable." Under DDA, it sees the full chain and can bid on upper-funnel keywords that assist conversions — even if they never get the last click.
This matters for strategy:
- Last click + Target CPA tends to under-bid on broad-match and upper-funnel terms (they look like they have zero conversions)
- DDA + Target CPA distributes bids across the full funnel, often improving total conversion volume at the same CPA
If you switch models, give Smart Bidding 2–4 weeks to recalibrate. The numbers will look unstable in the first week — that's the algorithm adjusting to the new credit distribution, not a bug.
7. Verifying attribution is working correctly
Check 1: Compare models side by side
In Google Ads, use Tools → Attribution → Model comparison to see how conversions redistribute under different models. If all models show nearly identical numbers, your conversion paths are short (single-touch), and the model choice doesn't matter much.
Check 2: Look for fractional conversions
Under DDA, you should see fractional conversion counts (e.g., 2.4 conversions instead of 2 or 3). If every conversion action shows only whole numbers, either:
- The model is actually set to last click (check each action individually)
- Your conversion paths are all single-touch (the model has nothing to distribute)
Check 3: Verify the conversion path report
In Google Ads → Attribution → Top paths, check that multi-touch paths exist. If all conversions show a single touchpoint, DDA has nothing to work with and effectively behaves like last click.
Check 4: GA4 attribution analysis
GA4 provides its own attribution reports that complement Google Ads:
-
Advertising → Attribution → Conversion paths: Shows the sequence of channels that led to conversions. Use this to see whether your conversions are truly multi-touch (making model choice important) or mostly single-touch (making it irrelevant).
-
Advertising → Attribution → Model comparison: Compare how key events are distributed across channels under different models. If a channel's credit changes dramatically between models, its value is uncertain — investigate further before making budget changes.
-
Advertising → Attribution → Attribution paths → Path length: Check the distribution of path lengths. If 90%+ of conversions happen in 1 touchpoint, attribution model choice doesn't matter for your business — the models only diverge on multi-touch paths.
Note: GA4's attribution reports include all traffic sources (organic, social, referral, paid), not just Google Ads. This gives you the cross-channel view that Google Ads attribution reports can't provide.
Check 5: After changing models, compare the before and after
Run a report for the same date range with the old model and the new model. The total conversions should be identical; only the per-campaign distribution should change. If total conversions shifted, something else changed at the same time (tag, consent mode, conversion window).
Frequently asked questions
Q. We switched to data-driven and our brand campaign's conversions dropped by 40%. Is that right? A. Likely, yes. Brand campaigns are typically the last click in a multi-touch path. Under DDA, they share credit with the upper-funnel touchpoints that introduced the user. The total across all campaigns should be the same.
Q. Can I use different attribution models for different conversion actions? A. In Google Ads, yes — each conversion action has its own model setting. In GA4, no — one model applies to all events.
Q. Does the attribution model affect the raw data / tag firing? A. No. Attribution is applied in reporting, not at the tag level. Your tags fire and send data identically regardless of the model. The model only changes how credit is distributed in reports and how Smart Bidding interprets conversions.
Q. We have very few conversions. Should we still use DDA? A. With fewer than ~50 conversions per month, DDA may not have enough data to learn meaningful patterns. In that case, last click is simpler and more stable. Google will sometimes fall back to last click automatically when data is insufficient.
Conclusion: the model isn't just a reporting setting — it shapes your bidding
Attribution model choice is one of those settings that sits quietly in the background until someone asks "why do our numbers look different?" Understanding what it does, checking what's currently set, and deliberately choosing the right model for your business turns a source of confusion into an informed decision.
The key habit: whenever conversion numbers shift unexpectedly, check the attribution model before assuming a tag is broken. And when comparing platforms (Google Ads vs. GA4, or Google vs. Meta), remember that each system attributes conversions with its own model — mismatches are expected, not errors.
ConversionOK verifies that the conversion signals your site sends actually reach the ad platform — the foundation that attribution models build on. If the tag doesn't fire or the data doesn't arrive, no attribution model can save you. Start with the signal, then worry about credit.